Why I'm Leaving My Company Immediately (Stability AI) w/ Emad Mostaque | EP #93
Peter H. Diamandis2024-03-29
peter diamandis#longevity#xprize#abundance
12K views|3 months ago
💫 Short Summary
The video discusses the importance of decentralized intelligence and governance in AI technology, emphasizing the need for collective intelligence and distributed data sets. It highlights concerns about centralized control, the impact of AI on society and democracy, and the potential risks associated with advanced technology. The speaker advocates for democratized and decentralized AI models, focusing on transparency, accessibility, and ethical practices. Additionally, the video addresses the need for a collaborative approach to AI development, setting norms for its use, and empowering individuals through technology for the betterment of society.
✨ Highlights
📊 Transcript
✦
Concerns about centralized AI threatening freedom and liberty are shared, highlighting the importance of decentralized intelligence.
00:00The CEO decided to step down from a unique company involved in global leaders and policy debates, focusing on developing generative AI models.
Despite resource challenges and competitive offers, the company made significant progress in AI development.
The importance of collective intelligence and ethical AI practices is emphasized for the future of technology and humanity.
✦
Importance of governance in AI and technology management.
04:17Speaker's role as a founder and strategist rather than a CEO, emphasizing the need for different skill sets in leadership.
Evolution of technology like stable diffusion and Sora as the next big thing.
Concerns about governance in organizations like OpenAI and questioning who should manage technology impacting humanity and education.
Speaker reflecting on past experiences as a CEO and the emotional toll of stepping down.
✦
Importance of decentralized data standards and ownership in a national model tied together by web3 framework.
09:08Concerns about consolidation in the tech industry and control of infrastructure by private companies with unclear objectives.
Acknowledgment of good intentions of tech leaders like Satia from Microsoft, but worry about consolidation of power and talent.
Highlighting potential risks associated with the current structure of technology companies.
✦
Emotional relief after stepping down as CEO of a moonshot company.
14:24Focusing on strengths and values is crucial for personal growth and success.
Challenges of growing a company rapidly and pivoting towards VC funding while maintaining the mission.
Generative AI has potential to make a significant impact on society.
Democratized and decentralized AI can empower individuals to reach their full potential.
✦
Key components discussed are accessibility and governance of technology.
18:22Centralized governance presents challenges in decentralizing technology, emphasizing the necessity for every country to have an AI strategy.
The potential impact of AI is likened to fire and electricity, highlighting the importance of establishing norms for its use.
Transparency is emphasized, along with the need to find a balance between top-down and bottom-up approaches in AI development.
The CEO of Google views AI as powerful as fire and electricity, signaling its profound influence on future society.
✦
Importance of Democratic Governance and Advanced Technology in Healthcare.
21:55Emphasis on the need for democratic governance to prevent top-down control in governments.
Urgency of implementing advanced technology and diagnostic centers like Fountain Life for comprehensive health screenings and early disease detection.
Fountain Life offers advanced therapeutics to add healthy years to life.
Encouragement for viewers to visit FountainLife.com for more information and membership opportunities.
✦
Democratized and decentralized AI is becoming more accessible and focuses on governance.
23:40Technology is now available to everyone without the need for giant supercomputers.
Graduates are advised to include local culture and knowledge in models.
Collecting quality datasets is crucial for training models at a lower cost.
Leading models in image and language may cost under $10,000 to train by the end of next year.
✦
Importance of capturing and preserving cultural data and stories in national models.
26:31Emphasis on the value of focusing on data sets over models and utilizing AI to enhance data quality.
Mention of ventures like OTOY and Render Network aiming to provide accessible, high-quality assets for 3D models.
Discussion on the concept of a 'hollow deck' reliant on data and the need for a commons of data to train specialized graduates.
Importance of establishing standards around monetization and intellectual property rights.
✦
Future stability and leadership within the organization discussed.
32:31Advice given for new leadership and plans for setting up new companies highlighted.
Mention of being a founder and majority shareholder, potential influx of money, and the role of the organization in the media industry.
Emphasis placed on the importance of having the right leadership for success in the organization.
✦
Importance of power dynamics and imbalance in founder-led companies.
34:26Emphasizing the need for team members to step up with their own vision.
Concerns raised about over-reliance on the founder's vision and its impact on team development.
Highlighting the importance of stability, revenue growth, and governance considerations before making major changes.
Discussion on a unique scenario where the CEO steps down to focus on broader issues, highlighting challenges and responsibilities of visionary CEOs.
✦
Viome utilizes advanced mRNA technology and AI to analyze microbes and cells, offering personalized nutrition recommendations.
38:41Viome provides tests for gut, oral, and cellular health that are tailored to individual genetics.
The data collection and AI engine at Viome improve daily, positively impacting health.
Users receive precision supplements and probiotics based on their specific needs.
Viome is offering up to 40% off tests with the code 'moonshots' for those looking to join the journey of health discovery and improve health from the inside out.
✦
Implications of AI Technology on Speech and Democracy.
40:42Discusses the potential for emotionally intelligent AI and the power of persuasive speech.
Explores the impact of AI on political manipulation and centralized government control.
Highlights the need for a more representative democracy and the concern about AI's ability to enhance persuasive speech.
Emphasizes the importance of understanding the influence of AI on communication and democracy.
✦
Manipulation and control of technology by organizations like YouTube, Google, and meta for engagement and advertising purposes.
45:00Concerns raised about the spread of extreme content and lack of morality in these companies and governments.
Emphasis on the need to protect individual liberty, freedom, and agency in the face of advancing technology.
Debate on whether to control technology or be controlled by it, highlighting the importance of setting defaults and standards to safeguard democracy and personal freedoms.
✦
Future of AI: Collective intelligence with amplified human intelligence.
47:52Preference for distributed system with data reflecting local cultures and promoting transparency.
AI should enhance human potential and coordination, not control.
Emphasis on collaboration in AI development and setting up AI champions in every nation.
✦
Importance of implementing organizations for data governance and model creation.
51:01Need for generative AI infrastructure company to support sectors like healthcare and finance.
Highlighting the concept of a web3 type protocol for intelligence, focusing on identity attribution and verifiability.
Plan to launch a company for every major sector, bringing together experts to solve infrastructure challenges.
Emphasis on financial, human, and political capital to drive global change and vision for a coordinated protocol to standardize sector advancements.
✦
Importance of Decentralization and Reframing Concepts in Technology.
55:00Bitcoin's success attributed to lack of centralized power.
Setting up technology properly and giving away power to avoid reliance on individuals emphasized.
Building a successful company with a diverse team and cutting-edge models challenges notion of impossibility in competing.
Advocacy for the use of generative AI in healthcare and excitement expressed about the future of AI.
✦
The speaker shares his transition from building financial systems to developing NLP tools for medical research.
57:51He discusses the need for improved access to information and personalized care in the healthcare system.
The speaker advocates for an open infrastructure for cancer patients that utilizes AI models for medical advice.
He envisions a future where technology enhances healthcare decision-making and empowers individuals with comprehensive resources.
✦
Efforts to improve data quality for enhanced models and saving lives.
01:01:05Launch of stable health infrastructure to provide knowledge on longevity and health.
Development of an open infrastructure to address conditions like autism and cancer, giving agency and compensation to individuals.
Transition from analog to digital models to revolutionize understanding and treatment of various conditions.
Initiative to attract top talent in different areas to drive innovation and improve millions of lives.
✦
Empowering children's learning and creativity through AI and language models.
01:04:25Emphasizing the importance of a transformative educational system that respects children's rights and agency.
Highlighting the potential of decentralized intelligence for global collaboration and impact.
Discussing the collaborative nature of open-source models in education.
Advocating for a shift away from outdated industrial-age teaching methods.
✦
The potential development of intelligence and education as a human right is discussed, focusing on its impact on national leaders and corruption.
01:08:28Empowering people with technology and education is highlighted as a key strategy to combat corruption.
The debate on technology distribution, particularly in China and other nations, is examined.
Global coordination, standards, and a decentralized approach to intelligence augmentation are emphasized as crucial.
Uniting efforts in the AI field is deemed necessary to achieve a significant breakthrough and avoid a centralized solution.
✦
The battle against powerful centralized AI systems striving for AGI.
01:11:48Concerns of organizations consolidating too much power and creating unbreakable norms.
Race between human intelligence, artificial general intelligence, and collective intelligence.
Urgency in upgrading the human operating system to prevent centralized systems from taking away freedom and liberty.
Collaboration among experts in various fields is essential to address challenges posed by advanced AI.
✦
Advocacy for Decentralized Intelligence
01:14:54Emphasizes the need for collective intelligence and distributed data sets to benefit society.
Expresses concern about the potential dangers of artificial general intelligence (AGI) and argues against a centralized approach to its development.
Highlights the importance of setting norms and involving diverse voices in discussions surrounding AGI.
Proposes a more distributed and collaborative approach to AI development to avoid placing too much trust in a single entity.
✦
Importance of building a decentralized AI system for safety, better benefits, and achieving goals.
01:18:16Leaders are focused on advancing technology to improve health, education, and overall capabilities of their people.
Emphasis on creating a system where the people of the country collaborate to develop technology for the benefit of all, without political or external influences.
Need for coordination and a few giant supercomputers instead of many.
Mention of working on a white paper and challenges faced in previous roles.
✦
Tribute to real heroes who can carry the message forward.
01:20:47Expressing gratitude for those who continue to inspire and make a positive impact.
Highlighting the importance of recognizing and honoring those who make a difference in the world.
Encouraging viewers to reflect on the contributions of everyday heroes in society.
Emphasizing the power of individual actions in creating positive change.
00:00these organizations are telling you that
00:02they're building something that could
00:03kill you and something that could remove
00:05all our freedom and
00:08liberty and they're saying it's a good
00:09thing you should back them cuz it's cool
00:11they don't care about the revenue they
00:14have the political power people are
00:16scared of them power should not be
00:18invested in any one individual if I can
00:20accelerate this over the next period I
00:22don't have to make an impact I should
00:23not have any power whereas again you see
00:26everyone else trying to get more and
00:27more
00:28power the only way that you can beat it
00:32to create the standard that represents
00:34humanity is decentralized intelligence
00:37it's collective intelligence the data
00:40sets and Norms from that will be ones
00:42that help children that help people
00:45suffering that reflect our moral
00:47upstanding and the best of us and
00:50gathers the best of us to do
00:51[Music]
00:54it a week ago EOD stock was on my stage
00:58at abundance 360 talking about the
01:00future of open- source ai democratized
01:03decentralized ai the day after a360 he
01:07stepped down as CEO of stability now 5
01:09days later I've sat down with iMat to
01:12talk about why he's stepping down what
01:14he's doing next the future of AI he
01:16takes the gloves off he talks about the
01:18dangers of centralized Ai and the
01:21potential for decentralized democratized
01:24AI to be the only Avenue that truly
01:26uplifts all of humanity all right uh if
01:29you like this episode please subscribe
01:32let's jump in if you're a mood shot
01:34entrepreneur this is an episode you're
01:36not going to want to miss all right now
01:38on to emod good morning emod good to see
01:41you my friend that's as always pizza so
01:44you and I were on stage literally last
01:47week at the 2024 abundance Summit
01:50talking about uh the whole open- Source
01:55AI movement you were beginning to talk
01:57about decentralized AI you talking about
02:00where stability was the speed of the
02:02development of the different uh products
02:05and the day after the abundance Summit
02:08was over uh the news hit that you had
02:12stepped down as CEO of stability and
02:16stepped off the board um so let's begin
02:19with the obvious question uh why what
02:22happened um and I I have huge respect
02:26for you and I know a lot of the issues
02:28in the past but I'd like you to have a
02:29chance to share with entrepreneurs out
02:33there and folks interested in AI exactly
02:36your side of what's
02:38Happening yeah thanks I think that uh
02:40Elon Musk once characterize being a CEO
02:44as uh staring into the abyss and chewing
02:47glass uh because you are looking at a
02:50very uncertain future having to make
02:52decisions and the chewing glass is all
02:54the problems that come to you all the
02:55time and it's required to steer the ship
02:58when things are incredibly be uncertain
03:01and stability it's a pretty unique
03:04company at a unique time like we hired
03:06our first developer and researcher two
03:08years ago and then in those two years we
03:11built the best models of almost every
03:13type except for large language image
03:15audio 3D Etc and had over 300 million
03:19downloads of the various models we
03:21created and supported which was a bit
03:23crazy and then generative AI is crazy in
03:26terms of usually in a startup you don't
03:29have to deal with global leaders and
03:31policy debates about the future of
03:33humanity and AGI and everything else at
03:37the same time with building code yeah at
03:40the same time as building code yes and
03:42um especially be building code a take
03:44fraction of the resources of our
03:47competitors um like we had uh certain
03:50teams who offered triple their entire
03:53packages to move to other companies I
03:56was grateful that only a couple of
03:58researchers before
04:00N I was going to leave left for other
04:02companies and that was just startups no
04:05one left for another big company which I
04:07think is Testament to their kind of
04:08loyalty in the mission but you know what
04:11we've seen over the last year is or last
04:13half year in particular is the question
04:15of governance in AI is something that's
04:17incredibly
04:19important and who manages owns controls
04:24this technology and how is it
04:25distributed so we saw you know
04:28everything from kind of open AI
05:29because we've had this
05:31conversation uh because it was there was
05:34a lot of pressure asking whether for you
05:36to step down as
05:38CEO um and I think Founders typically
05:43want to see themselves or or feel they
05:45need to be the CEO and I I've heard you
05:48say recently you know that you view
05:50yourself more as a founder and
05:52strategist than a CEO is that a fair
05:55assessment yeah I think everyone's got
05:56their own skill sets right so I'm
05:59particular great at taking creatives uh
06:01developers researchers others and
06:03achieving their full potential and
06:04designing systems but I should not be
06:07dealing with you know HR and operations
06:11and business development and other
06:12elements they're far better people than
06:14me to do that so now for example our
06:17most popular thing stable diffusion and
06:19comf UI the system around it is the most
06:21widely used image software models in the
06:24world there are great media CEOs that
06:26can take that and amplify that to make
06:28hundreds of millions of Revenue so
06:29should come in and me on that so why now
06:33pal what is there anything that
06:35specifically tipped for you that has I
06:40mean because you know it has you know
06:43you have done an extraordinary job this
06:45has been your
06:46baby I mean how you have to
06:50feel a a whole slew of emotional
06:53elements and and I've had to step down
06:56as CEO uh on two occasions over the 27
06:59companies
07:00I've had to sell a company for pennies
07:02on the dollar and it takes an emotional
07:06um hardship on you yeah I've um had
07:10calls for me to step down as CEO since
07:13the well since 2022 you know uh but I
07:17always thought you know what's best for
07:18the company in the mission and where I
07:20look at the world right now there's a
07:21few things say the company has momentum
07:24it has spread it's turning into a
07:26business like last year I said let's not
07:28enter into large new contracts because
07:31technology isn't mature yet and our
07:33processes aren't mature yet and you have
07:34to deliver so we did a lot of
07:36experimental things we're setting up and
07:37again now it's ramping on a business
07:39side on a technology side the technology
07:41is maturing diffusion Transformers such
07:44as stable diffusion three and Sora are
07:46going to be the next big thing and again
07:47stability has got a great place there
07:50but I think there's also the macro on
07:52this so if you look at the open aai um
07:55CEO for thing you know Sam mman said the
07:58Bor conf fire anytime this is the
08:00governments of open Ai and then they
08:02fired him and then he is back on and he
08:07appoints himself back on the board
08:09there's clearly no governance opening I
08:10mean I respect the people on the board
08:12greatly you I think there're some great
08:13individuals but who should manage the
08:16technology that drives humanity and
08:18teaches every child and manages our
08:20government who's really leading on that
08:22that can build these models and do those
08:24things as you know I've always wanted to
08:26build the science models and the the
08:29health teams done that I want been doing
08:30the education work and then my concept
08:33of a national model for every country
08:35owned by the people of the country all
08:38tied together I think it needs to be by
08:39a web3 not crypto or necessary token
08:43framework that's something that's a
08:45brand new kind of Challenge and one that
08:47I think there's only a window of a year
08:49or two to do if you have highly capable
08:52models let's put aside AI for now which
08:54you can discuss later really
08:57accelerating then no one will be able to
08:59keep up with that unless you build in a
09:01decentralized manner and distributed
09:03manner for data talents distribution
09:06standards and more so there's only a
09:08small window of time here to do that and
09:10realistically again successful companies
09:12and these things are all great but geni
09:16is a bit bigger than the classical knobs
09:18just like the whole life cycle of the
09:20company was a lot faster than the
09:21classical knobs so that's why I felt you
09:23know now is the right time to make that
09:26change and hopefully play my part in
09:28making sure technology is distributed as
09:31widely as possible and governed properly
09:34as pretty much I think I'm the only real
09:35independent agent that has built
09:38state-of-the-art models in the world
09:39right now yeah we've seen a lot of
09:43turbulence with open AI we just saw
09:47Mustafa um from inflection become part
09:51of Microsoft um and I am curious I mean
09:55you you had a a now famous conversation
09:58with Satia
10:00uh a couple of days after stepping down
10:02um was that investigatory on your part
10:05or was that just a touch base with an
10:08old friend that's just ch trolling
10:11actually you know like to let off some
10:12steam that picture was I think from year
10:15or two year or so ago
10:19okay but you know um I think Sati is an
10:22amazing CEO and uh you know he responds
10:25again like the top CEOs incredibly
10:26quickly when you message um and he's got
10:30a great vision but there is again this
10:32concern about consolidation in Tech we
10:34didn't take money from any trillion
10:36dollar companies at stability you know
10:39we remained and retained full
10:41Independence you know um and you know
10:45the jman of some of the elements that we
10:47could have taken very big checks and
10:48other things um and even though you have
10:52good intention R room that companies are
10:54like slow Dumb AIS that over optimize
10:57the various things that's not certainly
10:59in the best interest of humanity when
11:01you have
11:02infrastructure it's like this is the
11:05airports the railways the roads of the
11:07future AI is an infrastructure G is an
11:10infrastructure but it should be and
11:12should it be Consolidated under the
11:14control of a few private companies with
11:16unclear objective functions again the
11:18people in the companies may be great I
11:20don't think so and this is a key concern
11:23and part of that was that commentary
11:25again he's doing an amazing job he's
11:26consolidating a lot of power for the the
11:29good of the company and also he has a I
11:31think genuinely good heart of Mission to
11:33bring technology to the
11:34world but it is a bit concerning right
11:38especially with the new types of
11:39structure you're speaking of Sam in this
11:41case Satia here oh SAA like like again I
11:45think uh the commentary was always
11:46interesting like you know Satia playing
11:484D chess you know assembling the AI
11:50Avengers yeah I mean he is he's building
11:52an amazing massive Talent covering the
11:54bases and Microsoft is doing incredibly
11:56well here right if you asked who's doing
11:59gen to people would say Microsoft um but
12:02there is has to be concerns about
12:03consolidation of talent and power and
12:05reach before I get to your vision going
12:08forward because it's so important and
12:10what you're doing next um I just again
12:14as a Founder as a CEO of a moonshot
12:19company we have a lot of those listening
12:22here can I ask how are you feeling right
12:25now because the decision to step down
12:28has to have huge emotional are you
12:30feeling relief are you feeling anxiety
12:34um what's what's the feeling after
12:36making that momentous
12:39decision uh I was a big feeling of
12:43relief um you know
12:46because there's the there's that
12:48Japanese concept of iy I I know iy and I
12:51love it yes yeah do what you good at do
12:54what you like and do what you believe
12:55you're adding value and other people do
12:57too you know like realistically again I
13:01think I was an excellent research leader
13:03strategist other things but I didn't
13:05communicate properly or hire the right
13:07other leaders in certain other areas of
13:09the company and they're better people to
13:10do that and so I wasn't doing what I was
13:13best at a lot or I could have the most
13:15measurable value and it was tying down
13:18you know there's a lot of Legacy you
13:19know technical organizational other debt
13:22especially when you grow so so so fast
13:25and you know we were lucky that we had
13:26high retention in the important areas
13:28and we could execute in spite of all of
13:30that in spite of going the big company
13:32route I think you know moonot Founders
13:34you have to do it because you don't have
13:35the resources at the start and you have
13:37to guide the ship you know as it goes
13:39out from Port but there does come that
13:41transition point there and there is a
13:44competing thing where you typically take
13:46on VC money which has its own objective
13:48function versus your overall mission so
13:51again if you look at the generative AI
13:53world right now how many
13:55credible intelligent independent voices
13:58are there there theyve had the ability
14:01to build models and design things and
14:03make an impact you know there's not many
14:05so I was like that's where I can add my
14:07most leverage and also the design space
14:10is again is unprecedentedly huge because
14:13the entire Market has just been
14:15created like where does geny not fit and
14:19where does it not touch and what needs
14:21to be built there we need to actually
14:22have the agency to go and build that and
14:24so I felt tired relieved um I felt that
14:29now there's a million options I want to
14:32rather than taking a long break get on
14:34with things I've just done the first
14:35thing in kind of web 3 and I've got a
14:37whole bunch of other things we're going
14:38to discuss kind of coming and catalyze
14:42stuff that can make an exponential
14:44benefit because you know like massively
14:46transformative purpose here is I want
14:49every kid to achieve their potential and
14:51give them the tools to do that and and I
14:53love you for that um because you've been
14:55true to that that vision and I know on
14:58the heels of your announcement you've
15:00been reached out to by national leaders
15:02by corporate you know by CEOs and major
15:06investment groups and um you have a lot
15:08of opportunity ahead of you so let's
15:12talk about where you want to go next um
15:15you mentioned publicly and you discussed
15:18on our abundant stage the idea of of uh
15:21democratized and decentralized AI um
15:25let's define that first what is that why
15:27is it important and what do you want to
15:29do there yeah I think that when I said
15:33I'm going to move to do my part in
15:35decentralizing people like isn't that
15:37just open source you give the technology
15:39right and then anyone can use it but it
15:41isn't a decentralizing AI has a few
15:44important components one is availability
15:46and accessibility everyone should be
15:48able to access this technology the foods
15:51the labor and there's some very
15:52interesting political and other elements
15:54around that number two is the governance
15:57of this
15:58technology you have centralized
15:59governance because the models are the
16:01data there's a recent data bricks thing
16:04uh model where they show that you have
16:05massive improvements from data we all
16:06know that you know who governs the data
16:09that teaches your child or manages your
16:11health or runs your
16:13government that's an important question
16:15I think too few are asking and we need
16:16dat transparency and other things like
16:18that um so accessibility you know you've
16:22got the um governance aspect of that and
16:25then finally you have how does it all
16:27come together is it a single package or
16:30is it a modulized
16:33infrastructure that people can build on
16:35and is available kind of everywhere you
16:37know does it require monoliths and
16:39Central servers where if it goes down
16:41and you have an outage on gp4 you're a
16:43bit messed up or someone can attack and
16:45corrupt it I think that those are kind
16:48of the key elements that I was looking
16:49at when I was talking about
16:50decentralizing Ai and you know I've come
16:53up with an infrastructure to do that I
16:56hope um as well so if if you don't mind
16:59let's double click on it even further so
17:02um you mentioned we don't have long to
17:05get there if that's a true statement why
17:09don't we have long to get there and then
17:12what does getting there look like if you
17:15had all the capital available and If the
17:18right national leaders were hearing
17:20about this because a lot of this is
17:23supporting um supporting the populace of
17:27a Nation um to have have ai that serves
17:30them versus uh top down what's it look
17:35like 25 10 years from now yeah I think
17:39you'll have both proprietary and open
17:41source Ai and they'll work in
17:43combination the Practical example I give
17:46is that sayi is like graduates right
17:48very talented slightly of enthusiastic
17:50graduates you got those and
17:52Consultants um but I believe you know on
17:55stage with mat Freedman last week at
17:56a360 when we're there he said it's like
17:58we discovered this New Concept what do
18:00you call it AI Atlantis Atlantis yes yes
18:03with 100 billion graduates that will
18:04work for free yes I love that analogy it
18:07was it was a brilliant analogy yeah we
18:09need to figure out how to say Atlantis
18:11you know um but but there's a few things
18:14here first of all is the defaults you
18:16know once a government Embraces
18:18centralized technology it's very
18:20difficult to decentralize it and every
18:22country needs an AI strategy a year ago
18:26one year ago was GPT 4 yeah crazy how
18:30crazy is that you know at the AI safety
18:33Summit uh in the UK the king of England
18:37came on stage or came via video call and
18:40he said that this is the biggest thing
18:42since fire you know and that was like
18:45what six seven months later where are we
18:47going to be in here yeah I think he took
18:49that from from uh uh from uh uh the
18:53found the CEO of Google um AI is as
18:57powerful as as fire and electricity yeah
19:00yeah yeah I've heard the same from like
19:02Jeff Bezos and a bunch of others you
19:03know uh not Kindle Fire proper fire you
19:06know um but then if you think about it
19:09Norms are going to be set in this next
19:11period like you know I'm in California
19:13La at the moment if you don't set Norms
19:15on rights for actors and the movie
19:18industry then you could have a massive
19:20disruption just occurring as fulllength
19:22Hollywood features come in a year or two
19:24generated you know if you don't have
19:26Norms around open models and earn sh and
19:29governance by the people it'll be top
19:30down governance because governments
19:32can't allow that to be out of control if
19:34they don't have a reasonable
19:36alternative um and I think the window is
19:39only a year or two because every
19:40government must have a strategy by the
19:42end of the year and so I think if you
19:44provide them a good solution that has
19:46this element of democratic governance
19:48and others that will be immensely
19:49beneficial I think also it's urgent
19:53because we have the ability to make a
19:55huge difference you know as we kind of
19:57May probably discuss later having we W
19:59the knowledge of cancer longevity autism
20:01at your fingertips we have the
20:03technology for that right now we have
20:04the technology that no one ever ever
20:06alone again on those things or to give
20:08every child a superior education
20:10literally in a couple of years like
20:13there is an urgency both from there's a
20:15small window but also from we must do
20:17this now because it can scale and make
20:20that impact we have dreamed of for so
20:21long the enabling technology is finally
20:23you know it's finally good enough fast
20:25enough and cheap enough everybody I want
20:26to take a short break from our episode
20:28to talk about company that's very
20:29important to me and could actually save
20:32your life or the life of someone that
20:34you love company is called Fountain life
20:37and it's a company I started years ago
20:38with Tony Robbins and a group of very
20:41talented Physicians you know most of us
20:43don't actually know what's going on
20:45inside our body we're all optimists
20:48until that day when you have a pain in
20:50your side you go to the physician or the
20:52emergency room and they say listen I'm
20:54sorry to tell you this but you have this
20:56stage three or four going on and you
20:59know it didn't start that morning it
21:01probably was a problem that's been going
21:03on for some time but because we never
21:06look we don't find out so what we built
21:09at Fountain life was the world's most
21:12advanced diagnostic Centers we have four
21:14across the us today and we're building
21:1720 around the world these centers give
21:19you a full body MRI a brain a brain
21:22vasculature an AI enabled coronary CT
21:25looking for soft plaque a dexa scan a
21:28Grail blood cancer test a full executive
21:30blood workup it's the most advanced
21:33workup you'll ever receive 150 GB of
21:37data that then go to our AIS and our
21:39physicians to find any disease at the
21:42very beginning when it's solvable you're
21:45going to find out eventually might as
21:47well find out when you can take action
21:49Fountain life also has an entire side of
21:51Therapeutics we look around the world
21:53for the most Advanced Therapeutics that
21:55can add 10 20 healthy years to your life
21:57and we provide them to you at our
22:00centers so if this is of interest to you
22:03please go and check it out go to
22:06Fountain
22:08life.com Peter when Tony and I wrote Our
22:11New York Times bestseller life force we
22:14had 30,000 people reached out to us for
22:17Fountain life memberships if you go to
22:19Fountain life.com back/ Peter we'll put
22:21you to the top of the list really it's
22:24something that is um for me one of the
22:27most important things I offer for my
22:28entire family the CEOs of my companies
22:31my friends it's a chance to really add
22:34decades onto our healthy lifespans go to
22:38fountainlife
22:40decomp it's one of the most important
22:42things I can offer to you as one of my
22:44listeners all right let's go back to our
22:46episode so the let's talk about the
22:48objective function of democratized and
22:49decentralized AI um is it that the
22:53compute is resident um in countries
22:56around the world is it that the models
22:59are owned by the citizens of the world
23:02is it that data is owned um and how do
23:06you get there from here I think that um
23:09you can think of the supercomputers like
23:11universities M you don't need many
23:13universities honestly if someone's
23:15building good quality models that's one
23:16of the things is iability and we did the
23:19hard task we could have just stuck with
23:20image we said no we're going to have the
23:22best 3D image audio biomedical all these
23:25models and no one else managed that
23:27apart from open AI to Dee in fact I
23:29think we have more modalities than open
23:31AI again kind of want to kind of
23:33describe this is
23:35accessibility and governance and a few
23:38of these other factors so I think what
23:40it means is that this technology is
23:42available to everyone but you see now
23:45that you don't necessarily need giant
23:46supercomputers to even run it you know
23:49we show showed you a language model
23:50running on the laptop stable M2 will run
23:53on a gigabyte on a Mac mare faster than
23:56you can read you know we're writing some
23:57poems ious things you know um we see
24:01stable diffusion now at 300 images a
24:03second on a consumer graphics card our
24:06video model was like 5 gabt of vram this
24:09really changes the equation because in
24:11web 2 all the intelligence was
24:13centralized on these giant servers and
24:14big data now you have big supercomputers
24:17I think you'll need less with better
24:19data training these graduates that can
24:21go out and customized to each country
24:23but they must reflect the culture of
24:25that country like the Japanese stable
24:27diffusion model we had if you typed in
24:29salary man it gave you a very sad person
24:31versus the base model giving you a very
24:33happy person right so you must have
24:36graduates that reflect the local culture
24:38and then reflect the local knowledge and
24:40then Global models again that reflect
24:45our Global Knowledge and can be accessed
24:48by anyone but who decides what goes in
24:50there these are some very important
24:52questions and who vouches for the
24:55quality as well what's your advice to a
24:58a national leader because you know we're
25:00now starting to see Ministers of AI in
25:03different nation states and what what's
25:06your advice to them right now in this
25:09area I think my advice to them would be
25:12to start collecting the data sets that
25:15they would teach a graduate that was
25:17very smart through school and kind of
25:19other things this is National broadcast
25:21data this is the curriculum this is
25:24their accounting legal and others and
25:26note that those data sets are
25:27infrastructured
25:29they will enable the local populace and
25:32others to create these models because
25:34models are just data wrapped in
25:35algorithms with a bit of compute that's
25:37the recipe compute algorithms and data
25:40and it's not going to be as hard as you
25:43think to train these models but you have
25:45to build them to good standards so by
25:46the end of next year probably year after
25:49I would estimate that a l 70b LEL model
25:53or a stable diffusion model so these are
25:55two leading models in image and language
25:58will cost about under $10,000 probably
26:01even $1,000 to train and then it comes
26:04all about the data and then it becomes
26:06about the
26:08standards you know it's it's it's
26:10interesting um there is so much
26:13knowledge in the world that
26:16will vaporize sublimate over the decade
26:19ahead as people die you know uh cultural
26:23data locked up in people's minds and
26:25stories and so forth that's never been
26:27recorded it's an interesting time to
26:29actually capture that data and
26:31permanently store it into uh the
26:34national
26:35models yeah and again I think people
26:38over focus on the models versus the data
26:40sets I mean is data set yeah yeah yeah
26:42with the exponential compute you can
26:45recalibrate and improve the data as well
26:47so right now a lot of the improvements
26:49and models are actually synthetically
26:51improving data and data quality um as
26:54you said there's so much that can be
26:55lost but now we can actually capture
26:57this and the concepts and the other
26:59guidance and have cross checks like you
27:02can deconstruct laws you know you can
27:04translate between contexts you can make
27:08expert information available to everyone
27:10because again you have this new
27:12continent of AI Atlantis and all these
27:14graduates soon to be Specialists that
27:16are on your phone and that's incredibly
27:20democratizing you know um because
27:23otherwise the knowledge is throughout
27:25history knowledge has always been gatee
27:26kept always
27:29I want to get to uh health and education
27:31next but before we go there I know you
27:34were meeting with uh a mutual friend uh
27:38Jules herbach the other day um and uh
27:42and stability announced a deal with otoy
27:44Endeavor and render Network um are you
27:47still an adviser to that Venture yeah no
27:50this is part of the whole thing it's the
27:52first of many web 3 kind of elements
27:55there I think web 3 is 95% % I say 90%
27:59I'll be
28:00generous um speculative and rubbish but
28:05there is that 51% of genuine people that
28:07have been thinking about questions of
28:09governance coordination and others and
28:12have built things that are proper so
28:14ottoy is the bridge to the creative
28:16industry that's why with ar Emanuel and
28:19you Eric Schmidt and others are on the
28:21board um and the render network is a
28:23million gpus largely from creative
28:26professionals that are available and so
28:28the first uh thing I announced there was
28:31the initial 10 million gaps 250 million
28:34of distributed compute to create the
28:37best 3D data sets like at stability we
28:40funded and worked with Allen institution
28:42and others on obverse Excel which was 10
28:44million high quality 3D assets we're
28:46going to a billion distributed you don't
28:48need giant supercomputers but then that
28:51is a community good that is owned by the
28:52people of the network and accessible to
28:55non academic and others as well why
28:58because you need high quality assets to
29:00create better 3D models we have a new 3D
29:02model tripol that can generate a 3D
29:04image from a 2D image in 0.5 seconds and
29:07that 3D model feeds into better 3D
29:09assets and then what does that mean it
29:11means we're heading towards the hollow
29:12deck without the data you're not going
29:14to get there and Jules Jules wants the
29:17hollow deck for sure yeah so you know
29:19Jules and I are on the same page L you
29:21know and you're not going to get there
29:23without again a Commons of data that can
29:26train the graduates that then becomes
29:28specialized with Star Trek or you know
29:31Star Wars or any of these other IPS and
29:33then also setting standards
29:35around monetization IP rights all sorts
29:39of other things um and so a network like
29:41render is really good for that but you
29:44know I've been talking to a lot of
29:45people in web 3 about the different
29:46elements of the stack uh because what I
29:49basically see is that we have the
29:51opportunity to build almost a human
29:53operating system uh models and data sets
29:56for every nation every sect coordinated
29:59through proper web 3 principles again
30:02not speculative tokens or anything like
30:04that you know making it so that every
30:07child in the world or adult can create
30:10anything they can imagine they can be
30:12protected against the harms and they
30:14have access to the right information at
30:16the right time to thrive and again
30:18that's infrastructure for everyone it's
30:20a common good access to gpus has been
30:24sort of The Limited fuel um do you think
30:27decentralized GPU structures like render
30:31is part of that future is an important
30:33part of that future I think so right now
30:36it's far more efficient to train models
30:39on these again big supercomputers the
30:41university but the rate of exponential
30:43growth there again is insan last year to
30:46train llama 2 cost $10 million in a year
30:49it'll cost
30:50$10,000 a, times improvement from
30:53algorithms data super compute
30:56speeds and that's crazy if you think
30:58about it right um so I don't think this
31:00will be the limiting factor I think the
31:02GPU overhang for language models
31:04probably lasts until the end of the year
31:06but then there's plentiful Supply
31:08because what you have is NVIDIA makes
31:10amazing gpus at an 83 or 87% margin
31:14right but the actual calculations aren't
31:16complicated like we took Intel gpus and
31:19we ran the stable diffusion 3 diffusion
31:21transforma training so this is the same
31:23technology that's used in Sora and
31:25stable diffusion 3 is multimodal so it
31:27can train Sora model with enough compute
31:29and I think us and them were the only
31:30people kind of doing this I think M
31:32pixel as well um and then it ran faster
31:36than the Intel gpus than the Nvidia
31:39gpus but we know that it can run even
31:42faster because it's not optimized for
31:44either it's still running fast so what
31:46you'll see is a commoditization of the
31:48hardware once the architectures get
31:50stabilized because gp4 is just a
31:53research artifact St diffusion was just
31:56a research artifact you're not going
31:58engineering phase yet and you've got to
32:00the point whereby this runs on MacBooks
32:03it runs on other things so I think it's
32:05a shortterm phenomenum of the next year
32:08because people were taking a point in
32:10time and extrapolating it without taking
32:12into account efficiencies
32:15optimizations and the fact that models
32:17that work on the edge and can go to your
32:19private data will be more impactful than
32:22generalized intelligence everyone's over
32:24indexing on generalized intelligence and
32:26building AI God
32:28versus Amplified human intelligence
32:30shall we
32:31say before we leave stability it's now
32:35in the hands of uh of the chair and your
32:40uh past CTO what's what do you imagine
32:42the future of stability is going to be
32:45going forward I know that you're not
32:46involved anymore it's under different
32:48leadership and there's a what's your
32:50advice to them or where do you think
32:52they're going to go you know I can get
32:55the very basic advice I didn't want any
32:56conflicts or anything because I'll be
32:58setting up lots of new companies and you
33:01know being a founder and a shareholder
33:03and against ability I'm a Founder
33:05shareholder in fact you're still the
33:07majority shareholder I think as of right
33:09now yeah just about just about yeah that
33:12will change I'm sure money will come in
33:14like we saw Co here yesterday on I think
33:1710 or 20 million of Revenue run rate
33:19they're raising at five billion money
33:21it's plentiful cing yes yeah with the
33:23right leadership I think that um it can
33:27again have an amazing part to play in
33:29media and that's what I've suggested to
33:30it um and again there's a great team
33:33that continues to ship great models so
33:34last week there was an amazing code
33:36model next week amazing language audio
33:38and other models are coming out so you
33:40know you continue shipping and great
33:41products around that too um so that was
33:45kind of my advice to them let's focus on
33:47media and take that forward but you know
33:50I'm not the expert on the business side
33:52of things I did the best I could my
33:53expertise on setting this up and taking
33:55it not to
33:5610 and uh yeah Z 0 to one is is
34:01definitely um a role that that you've
34:05played here and allow it allow someone
34:08else to take it the rest of the way uh
34:11but the area that I know if I actually
34:15there something I wanted to kind of
34:16discuss here I think is quite important
34:18please for again the founders listing
34:19and the new shop
34:21companies um there is an imbalance of
34:24power when you
34:26have very visionary very highly
34:28competent leaders there like what I
34:31found at stability is that everyone
34:33would be waiting for me no matter how
34:35competent because I was the one that
34:36could see around the corners and I was a
34:39bit good at everything even if I hired
34:41people that had built billion dooll
34:42startups or were leaders and research at
34:44Google or kind of whatever um because
34:47you have the outsize thing what kind B
34:49says you have to speak last in some
34:51cases and some people they're meeting CU
34:53otherwise everyone just does everything
34:54you say and they also wait on you now
34:57now what I find and what I told the team
34:59is that you're flat as a power Dynamic
35:01you're all on the same page you're all
35:02kind of relatively equal owners and
35:05it'll be interesting see how it evolves
35:06from that uh given that there's actually
35:09a business and again I think this is
35:10something that you probably had a
35:11challenge with than other Founders here
35:14whereby they put more on your plate
35:17because you are so Visionary and because
35:18you're like up there in the future and
35:21they're always waiting on you so you're
35:23always like well my schedule's
35:24completely packed my schedule now is
35:26actually quite free which is also quite
35:27LS
35:28I've had a chance to speak with you
35:29every day for the last few days so
35:31that's been a pleasure to have extra
35:32time on your schedule you know so so we
35:35we do have a world of uh Visionary
35:39founder-led CEO companies right so
35:41you've got musk and you've got Bezos
35:43historically and you had Steve Jobs and
35:45you have and and that's both powerful
35:48and dangerous the the power is the
35:51ability for that because we don't ever
35:53have a company that is pre-existing a
35:56new CEO comes in has the same um uh both
36:02hutzpah and and also the power of their
36:05of their Vision
36:08um the danger there you're concern
36:12you're saying is not not allowing your
36:15team to step up with their own Vision or
36:18being over overly indexed on your on
36:20your vision yeah I think that that can
36:23be the issue and that's why I wanted
36:24stability to again reach the point of
36:27spread and revenue rate increase and
36:29other things before I did anything um
36:32and I felt again this external pressure
36:34in that if nobody or there's very few
36:37people in the world actually thinking
36:38properly about governance and spread and
36:40others and a very small window giving
36:42the pace of this to make a difference in
36:43the dent I believed I had a reasonable
36:45approach to that but I couldn't while
36:48remaining CEO of this company and again
36:50it's a pretty unique scenario because
36:52you've never seen a sector move this
36:54fast that has such wide reaching human
36:56implications
36:58and regrettably there's too few people I
37:02think with the right alignment and
37:05approach in this area I've been very
37:08disappointed like usually what happens
37:10is you have power maximization equations
37:12and this is what we're seeing from the
37:13industry
37:14consolidation how many people want to
37:18genuinely bring this technology to kids
37:20in Nigeria or to the global south or to
37:24help those leaders build their own
37:26models you know and believe all in a
37:28positive SU game that was actually my
37:30biggest surprise from the discussion
37:31silic Valley almost entirely they all
37:34believe
37:35in uh flat or negative sum you know zero
37:39sum or negative sum things where there
37:40has to be a winner everyone's a winner
37:42in
37:43this and again I was just very
37:45disappointed seeing that I've been
37:47asking you for a while to write and
37:50distribute your vision white paper cuz
37:53I've heard you describe it in detail and
37:54it's brilliant and I still hope that the
37:57world will it soon enough everybody I
37:59want to take a break from our episode to
38:00tell you about an amazing company on a
38:02mission to prevent and reverse chronic
38:04disease by decoding your biology the
38:07company is called viome and they offer
38:10cuttingedge tests and personalized
38:12products that help you optimize your gut
38:14microbiome your oral microbiome and your
38:17cellular Health as you probably know
38:19your microbiome is a collection of
38:21trillions of microbes that live in your
38:24gut and mouth and these microbiomes
38:26influence every everything your
38:28digestion immunity mood weight and many
38:30other aspects of your health but not all
38:33microbes are good for you some can cause
38:36inflammation toxins and actually lead to
38:38chronic diseases like diabetes heart
38:41disease obesity and even cancer viome
38:43uses Advanced mRNA technology and AI to
38:47analyze your microbes and your cells and
38:50give you personalized nutrition
38:52recommendations and products designed
38:54specifically for your genetics
38:56specifically for your biology you can
38:58choose from different tests depending on
39:00your goals and needs ranging from
39:03improving your gut health your oral
39:04health cular function or all of them
39:07I've been using viome for the past 3
39:09years I can tell you that it has made a
39:11huge difference in my health and because
39:13of the data they collect and the AI
39:16engine they've built it gets better
39:18every single day I love getting Health
39:21scores and seeing how my diet and
39:23lifestyle affects my microbiome and my
39:25cells and I love getting precision
39:27supplements and probiotics tailored for
39:30my specific needs if you want to join me
39:32on this journey of Discovery and improve
39:34your health From the Inside Out viome
39:37has a special offer for you for a
39:39limited time you can get up to 40% off
39:42any viome test using the code moonshots
39:46just go to vom.com back/ moonshots and
39:49order your test today trust me you
39:51wouldn't regret it all right let's go
39:52back to our episode before we jump into
39:55into health and education let's talk
39:57about governance a second because we've
39:58seen governance
40:01um complicate this what is the right
40:04governance structure for this super
40:05powerful technology we have
40:07representation of democracy that I think
40:09can be improved by this like I don't
40:10think democracy survives this technology
40:12and it's current for it will either
40:14improve or it will end I don't see
40:17anything else like um yesterday they
40:19what does end what does end mean here a
40:22a benign dictatorship a driven by an AI
40:25Overlord yeah like yesterday there were
40:27was a announcement of an app called Hume
40:31which had emotionally intelligent speech
40:33and it can understand your emotions and
40:35talk with emotion you and I have to
40:37discuss this yes you know where that's
40:39going rightful speech it's incredibly
40:42powerful and governments have a tendency
40:44I mean if Government but but say it say
40:47it here it's important for you to State
40:49what what it means because we've
40:51discussed it but help people here under
40:54be ready for this democracy is is all
40:57about representation and you see the
40:59questions of deep fakes and things
41:00speech is one of the most impactful
41:02elements there but now you can't believe
41:04anything you see here everything so One
41:08path that we have is a 1984 on steroids
41:12panoptica you know where life is
41:15gamified and you listen to whatever the
41:17government says and they're incredibly
41:18convincing and you're happy and you've
41:20always been happy and you've always been
41:21at war with Eurasia you know propaganda
41:24on steroids the other part that you have
41:26is things like like citizen assemblies
41:28consultative democracy the ability to
41:31take right now you can take any of the
41:33bills in Congress and completely
41:35deconstruct them and find what the
41:38motivations are you know you can check
41:40laws against the constitution in a
41:42second seconds this is incredibly
41:45powerful empowering technology from a
41:47democratic perspective so I see two
41:49routes unfortunately because I think
41:51that once the thing goes It goes really
41:54fast centralized government control
41:56increasing because the governments want
41:58to protect themselves as an organization
42:00and you know every party says the other
42:02party is crap we've seen the increasing
42:04polarization in America already and you
42:07know fundamentally come on you can do
42:08better than those two leaders that are
42:10currently competing I'm saying this is
42:13clearly our system is sclorotic across
42:15this like democracy is the worst of all
42:16systems except for everyone else we can
42:18have a better
42:20democracy where it's actually
42:21representative and empowers the people
42:25or we will have the end of democ where
42:27it is a 1984 panopy in my opinion
42:30because the momentum will go there of
42:32course you will start using this
42:33technology you're already seeing it
42:35being used but not at scale and not
42:37intelligently yet which is
42:40scary I I I think we finally have the
42:42technology for a direct uh democracy
42:46versus a representative democracy right
42:47where I can have my my desires directly
42:52represented on any specific law or but I
42:56think the point point you've made before
42:58is that
43:01speech um if you look back to everybody
43:05from Hitler to some of the most
43:07persuasive
43:09politicians um is is a powerful tool and
43:13AI can become the most persuasive
43:16speaker out there it can take anyone's
43:19speech and make it far more persuasive
43:20like I think my voice is a bit whiny I
43:22can remove the wine you know I can go in
43:24a very polish British accent and other
43:26things like that right you know must
43:28fight them on the hills and the harriers
43:30and whatever public speaking makes a big
43:32difference someone took Hitler's
43:34speeches and put them through an AI and
43:36took them into English because when
43:39we're not in the German context and we
43:41listen to them it sounds like he's
43:42shouting like what this crazy thing you
43:44hear him in English it is very different
43:46in his own voice just like someone took
43:49Javier M's one in the United Nations and
43:52put him into English again he sounds a
43:54bit sh shouty but then he sounds very
43:57reable when it's in English and you can
43:59take the phenomes of Obama's best
44:01speeech and a bit of trumpism and a bit
44:03of church chill and you'll have full
44:05modulation wave control over all of this
44:07people are already using this technology
44:10that everyone should have a passcode
44:12with their loved ones because people are
44:14getting calls from their
44:15mother saying help I'm in an emergency I
44:18need to send money right now and you
44:21cannot tell it and it pulls at the
44:22emotional strengths and if you look at
44:24something like us radio and you know one
44:27side of the political divide is taking
44:29over imagine if you're hearing optimized
44:31speech every single
44:33day that will have a huge impact and
44:35then they control the visuals and they
44:36control the other things we're not set
44:39up for defenses yeah if it's if it's
44:41optimized speech for you specifically
44:43for you right for the kids you have the
44:46age group they have where you live your
44:48historical background and so forth and N
44:50of one persuasive speech coming at you
44:54um the brain is not set up for enses
44:58we're not and you know we can take this
45:00as an example of the YouTube algorithm
45:02like YouTube as an organization is not
45:04an evil
45:05organization but it's an organization
45:07optimized for engagement which optimized
45:09for more extreme content so there's some
45:11dark place in YouTube which then
45:12optimize for Isis so Isis video spread
45:16viral I don't know sometimes viral is
45:18good sometimes viral is bad that one was
45:19bad because it was extreme and they
45:23didn't understand why and if you look at
45:25it are two of our largest gen of are
45:28companies are Google and meta and their
45:32business is advertising their business
45:34is manipulation and they are both aoral
45:37companies because why would you expect a
45:39company to have morality our governments
45:42are also aoral and again you can view
45:45these things as slow Dumb AI so you can
45:46see the way they will optimize unless we
45:49do something about it and they will have
45:51full control like again you put on your
45:53Vision Pro headset with your spatial
45:56audio that is full sensory control not
46:00full but you know what I mean a level we
46:02full immersion full immersion full full
46:05immersion and so we have to be aware of
46:08this and there's obviously other tools
46:10that can be Ed like in the wake of the
46:11Arab Spring you know governments
46:13targeted everyone that was in social
46:15media we can do that on a again hyper
46:16personalized basis like we need to set
46:21some defaults and standards here to
46:23protect democracy but again why
46:26democracy
46:28we're not really trying to protect
46:29democracy you know again people have
46:31different definitions that what we're
46:32trying to protect is individual liberty
46:34freedom and
46:35agency education should be about
46:38enhancing the education of every child
46:39is not you know healthc care is sickcare
46:42our government should uplift us but how
46:44many people believe our governments do
46:46that rather than put us down because
46:48they couldn't encapsulate and cater to
46:51the Brilliance of each
46:54individual because they didn't have the
46:56tools until out so that's why I said
46:58which way one man yes which waynet
47:00agency or massive control these are the
47:04two ways do we control the technology or
47:06do these organizations control the
47:08technology that controls us you know
47:11when we were on the stage at at the
47:13abundance Summit we talked about a
47:15future of digital superintelligence
47:17right and a future in which we've got AI
47:22a billion times more capable than a
47:24human which looking at it just from a
47:27ratio of neurons is the ratio of a
47:30hamster to a human um y
47:34uh do you believe that someday we could
47:38have a a benign super intelligence that
47:43is supporting
47:44Humanity yes and I believe that it
47:47should be a collective
47:49intelligence that is made up of
47:52Amplified human intelligence as
47:53amplifying all of us Pilots that contain
47:56our Collective knowledge and culture and
47:59the best of us and data sets that are
48:01built from helping and augmenting US
48:03versus a collected intelligence and AGI
48:06that is topped down and designed to
48:08effectively control us again if you look
48:10at open AI statements on the road AGI
48:12they say this technology will end
48:14democracy end capitalism and maybe kill
48:17us all I don't like that I I remember I
48:20remember seeing that you you you texted
48:22me you said read this does this sound
48:25the same as it does to me yeah so what
48:27i' prefer instead is for this to be
48:28distributed like if you have data sets
48:30from Nations that are built on enhancing
48:32the capability of the nation that
48:34reflect the local cultures and you push
48:36for data transparency on models which I
48:38believe we must have you know especially
48:40language models then you're more likely
48:42to have a positive thing and again the
48:44human Collective can achieve anything
48:46from splitting the atom to go to space
48:48if we put our minds to it but we have
48:50laed in coordination
48:53mechanisms they've not been good enough
48:55so if you create the human Colossus and
48:57every single person has an AI That's
48:59just looking out for them to enhance
49:01their potential and coordination AIS
49:04that is a far more positive view of the
49:06future and that is the AGI that is a
49:09general intelligence that's the hive
49:11mind general intelligence not a b style
49:13hi mind but one that's really thinking
49:15again every child should achieve their
49:17potential versus this embodied concept
49:20of an AGI that's a very Western concept
49:22you see that as well like you know you
49:23look at the Japanese concept of a robot
49:25the robot is your equal and your helper
49:28you look at the Western concept of the
49:29robot it's Terminator and Skynet and all
49:32of that and again I think this is again
49:34where cultural norms become very
49:36interesting and what do we want to build
49:37do we want to build AI God or do we want
49:40to build that AI helper that helps us
49:42and we help it you
49:44know um those listening now I mean you
49:48can see yod's Brilliance and why I'm so
49:52enamored with the way you think about
49:54this because it's there are very few
49:56indviduals who are looking at this from
49:59uh an objective function of what's best
50:01for Humanity what's best for every
50:03nation state uh out there let's talk
50:06about your going forward future um are
50:10you going to build something in the uh
50:12in the decentralized um uh side of of AI
50:17the democrac size of AI is there a
50:19company there or Fund in your future for
50:22that yes so uh you know doing the white
50:25paper finally getting there with a bit
50:27of help from AI uh I I can't I I can't I
50:29can't wait to uh to help uh broadcast
50:32that white paper yeah but look I think
50:35the basic thing is this what I want to
50:36do is set up a AI champion in every
50:39nation with the brightest people of each
50:41Nation working with the organizations of
50:43reach Nation to help guide them through
50:45this next period because there will be
50:47massive job displacement from the
50:48graduates going massive uplifts and
50:51productivity from the technology being
50:53implemented and again that organization
50:55can help govern
50:57and create these data sets and these
50:59models that are so important and I Every
51:01Nation should have that but then I also
51:03believe that every sector should have a
51:05generative AI first infrastructure
51:07company that builds this and helps the
51:09healthcare companies finance companies
51:11and others through that and to
51:12coordinate all of that you need to have
51:15a web3 type protocol what is the
51:17protocol for intelligence so what is a
51:19web what is a web three type protocol
51:21Define that for folks listening again
51:24people talk about web 3 it's not about
51:25the tokens or the coins or anything like
51:27that what about three protocol is is
51:30that everyone should have like AIS first
51:32of all aren't going to have bank
51:34accounts they're going to need some way
51:36to pay each other or exchange value and
51:39again web 3 done a lot of work in that
51:41there needs to be some sort of identity
51:43attribution and other format because
51:45you'll have this Mass influx of
51:47information and so again web three
51:49concepts are very useful there there
51:52needs to be an identity concept because
51:53you'll have real and digital people web
51:56through concepts are very useful there
51:58so data ass estation all these other
52:00things verifiability so when I look at
52:03it if you've got sectorally my plan is
52:05to launch or Mr company for every major
52:07sector and we can talk about health and
52:09education and bring the smartest people
52:10in the world to solve that challenge of
52:12the infastructure for the future Every
52:14Nation but you need to have some sort of
52:16coordinating protocol for all of that
52:18that becomes a standard and that's the
52:20substrate for this Amplified human
52:22collective
52:24intelligence and and is that where you
52:26want to play
52:27and focus your energy next yeah it's
52:29setting up these organizations and
52:31bringing the brightest smartest people
52:33that really want to make a difference
52:35there because there's massive network of
52:37fols in doing this but again I just need
52:39to be the founder and architect I don't
52:41want to run the day-to-day of any of
52:42these things um and then because the the
52:46most scarce Talent there's three types
52:48of capital as I view it there's
52:50Financial Capital human capital and
52:53political capital and in order to affect
52:55a change in the world you actually need
52:56all three but the financial Capital
52:58actually comes with the people capital
53:00and the political capital and the
53:03smartest people in the world in every
53:04sector from Healthcare to education to
53:06finance to
53:08agriculture almost all believe that geni
53:11is the biggest thing they've ever seen
53:12in the last year everyone's asking you
53:15all the smartest people Keo what's next
53:17right and you know many of the smartest
53:18people in the world so I want to create
53:20organization that they can come the
53:22chefs and the cooks the thinkers and the
53:25doers and think what is the future of
53:26Finance that's the future of education
53:29and then the national champions that
53:31should be owned by the people of each
53:32country become the distribution for the
53:35amazing infrastructure that they build
53:37and there's a vice a nice kind of Vice
53:39Versa but then again you need the
53:40coordination function so I'm trying to
53:43bring together people in each of these
53:44and you know there'll be public calls
53:46and things like that to build that
53:48infrastructure the future because as
53:49mentioned AI is an infrastructure where
53:51it should be I maybe it's the rocket
53:53ship of the mind
53:54right I love that I love that analogy my
53:57friend it it it is the most important
53:59infrastructure that Humanity will have
54:01going forward across everything it does
54:04and I look forward to helping you build
54:05it exciting right like again I think one
54:08of the things I got I got lots of
54:10messages they're like I'm so sorry for
54:12your loss it was like my dog died after
54:15I left the CE I was like what is that
54:18you know they're like it's nice it's
54:21nice that people care right but I'm
54:22generally excited kind of about what's
54:24next like you know again it was like
54:26staring into the abyss and chewing glass
54:28every single day and that's not what I'm
54:31best at or where I could have the most
54:33impact but I want it to be a point
54:35whereby if I can accelerate this over
54:37the next period I don't have to make an
54:39impact I should not have any power on
54:42this whereas again you see everyone else
54:45trying to get more and more power I want
54:46to make sure it's set up properly but I
54:48want to give it all away because power
54:50is obligation it's dragging and again it
54:53should not be invested in any one
54:55individual we should not have to rely on
54:57anyone being nice or good or for this
55:00technology I was talking to uh Michael
55:02sailor during the abundance Summit uh
55:04that evening and you know talking about
55:06the fact that because Satoshi uh when he
55:10set it up uh did not retain any power
55:13and did not trade on the founding blocks
55:16and so forth that that's the reason it's
55:18been able to succeed because there
55:21wasn't that centralized power and you
55:23know Bitcoin had been tried he said
55:25Bitcoin had been tried many times before
55:27but because it didn't have that uh
55:30initial anonymity and and the and the
55:33dissolution of founding power that
55:35that's the reason it didn't succeed yeah
55:38I mean again I think you need to have it
55:40accelerate and you see this with
55:41movements right the movement starts but
55:43then it goes once you've got the DNA and
55:46the story there right you know you see
55:48the prophets you see the leaders you see
55:50the others but then it's about setting
55:52the framework correctly and reframing
55:54the concept
55:56this technology is not Beyond look
55:58stability is a company that started two
56:00years ago above a chicken shop in London
56:03right you know my first 20 employees I
56:06went to the job center and I said bring
56:07me people that have overcome adversity
56:10and I will train them young graduates
56:12and six of them are still at stability
56:14you know um like because it was a
56:16program and they're doing things from
56:17cyber security to running super
56:19computers we only had like 16 17 phds
56:22yet we built the state-of-the-art models
56:24in every modality we built mind reading
56:26models like mind ey you know I remember
56:28that contributed to all these things yet
56:32you're told it's impossible to compete
56:34we have shown it's not impossible to
56:36compete that's a reframing the reframing
56:38is data versus models it's you don't
56:40need giant supercomputers for everyone
56:43you just need to have a trusted entity
56:45to build it right yeah you know and so I
56:49hope to kind of convey this and then
56:50figure out this organizational structure
56:52that can proliferate so I can take a
56:54holiday so before go further let's talk
56:57about one area of your next chapter in
57:00life that uh we both have as a passion
57:03uh which is the use of generative Ai and
57:05health it's an area that you've given a
57:06huge amount of thought to uh and I think
57:09uh you're excited about can you share
57:11what your vision is there yeah so I got
57:14into AI 13 years ago gosh I was a
57:17programmer before for 23 years building
57:19L scale systems as a hedge fund manager
57:21and other things when my son was
57:23diagnosed with autism and then I built
57:25an NLP to analyze all the clinical
57:27literature and then looked at
57:29biomolecular pathway analysis of
57:31neurotransmitters gab toate in the brain
57:33to repurpose drugs for him and he went
57:35to Mas school which was great and equals
57:36one and then I was one of I was lead
57:39architect on the one of the co AI
57:41projects of the United Nations La of
57:42Stanford and others and then because I
57:44didn't get the technology I was like oh
57:46we got to build it
57:47ourselves but what is health you know
57:51again I think we've had this discussion
57:52a lot Healthcare is sickcare we don't
57:55have all the information that we should
57:56have at our fingertips Health assumes
57:59ergodicity a thousand tossers of the
58:01coin the same as a coin tossed a
58:03thousand times but we are all individual
58:06and across the world there are amazing
58:07data sets that could be better because
58:10when you write down a clinical trial or
58:12your own kind of experiences you lose so
58:15much information at the same time you
58:18don't have all the information on cancer
58:20autism multiple sceris at your
58:22fingertips and the comprehend
58:23authorative and Upstate way so when I
58:25look at the health operating
58:27system we're going to build a gp4 open
58:30for
58:32cancer and it's going to mean that
58:34nobody is alone again on that journey
58:36and loses that agency because they know
58:38comprehensive Authority update all their
58:40knowledge but AI models today already
58:44outperform human doctors and empathy so
58:46they're not going to be a learn on that
58:47anymore yeah can I just double click on
58:49what you just said because it's really
58:51important I've have so many people
58:53because of my role as chairman of
58:54Fountain life who reach out say I just
58:57got diagnosed with this cancer or my
58:59brother my sister or my wife and and
59:03there is they're left with this
59:05decimating use and they're left Googling
59:09um but a model that's able to have the
59:14most Cutting Edge information and then
59:16incorporate all their medical data and
59:18give them advice in empathic fashion how
59:21far is that see couple of years if we
59:24focus maybe even like next year and
59:27that's amazing because for all of these
59:29topics that again we will have diagnosis
59:32that is superior we will have research
59:34augmentation because again even
59:35researchers don't have all that
59:37knowledge at their fingertips and again
59:39this is public infrastructure and a
59:40public good you know from primary care
59:43all the way through that what is the
59:45open infrastructure of the future where
59:47this technology can come again to your
59:49own data as well you have uh things like
59:51Melody and other things around
59:53homomorphic encryption Federated
59:55learning that were trying to figure out
59:56how to preserve privacy we can run a
59:59language model on a smartphone right now
01:00:01that can analyze all your data and then
01:00:03just feedback stuff to a global
01:00:05Collective but people are people so when
01:00:07I look at Healthcare I see amazing data
01:00:09sets that we can activate by taking the
01:00:11models to the data an infrastructure
01:00:13that we can build like we had check's
01:00:15agent with Stanford the top x-ray ad
01:00:18Radiology model to build good standard
01:00:21things across the entire gamma of
01:00:23healthcare so we can actually get to
01:00:25healthcare versus sit so that we can
01:00:27make it so that everyone is empowered to
01:00:29make the best decisions either as
01:00:30experts or individuals and make it so
01:00:33nobody is alone again as well as
01:00:35increasing the data quality that will
01:00:37then feed better models that will then
01:00:39save lives save suffering and again
01:00:42increase our potential like you've got a
01:00:44longev book behind you right why don't
01:00:47you have all the latest knowledge of
01:00:49longevity at your fingertips at gp4
01:00:52level right now that will happen over
01:00:55the next year we will launch stable
01:00:56health or whatever decide to call it and
01:00:58there will be the smartest people in
01:01:00each of these areas working on that so
01:01:02again you never AR like doesn't matter
01:01:03if you're if you're with100 billion doar
01:01:05and your kid has autism
01:01:08ASD there's no cure there's no treatment
01:01:10there's nothing doesn't matter how rich
01:01:12you are yet with just a little bit of
01:01:15effort right now we can build it as an
01:01:17open infrastructure for the 5% of people
01:01:20in the world that know someone with
01:01:21autism the 50% of people in the world
01:01:24that receive a cancer diagnosis of them
01:01:26of someone they love and they feel that
01:01:28loss of agency so we're going to return
01:01:30agency to humanity that way and again it
01:01:33needs to be an open infastructure that
01:01:35they can then access private data sets
01:01:37and compensate them appropriately so
01:01:39everyone is incentivized we need that
01:01:41fast yeah and and that's a beautiful
01:01:44Vision it is again infrastructure and
01:01:47one of the things that's so beautiful
01:01:48about it is guess what all 8 billion
01:01:51people were all human or all running the
01:01:54same software and the the the
01:01:57breakthroughs and the knowledge
01:01:58accumulated in you know in Kazakhstan is
01:02:01going to be as useful in Kansas yeah but
01:02:05this is the thing operating system this
01:02:07is the biggest upgrade to the human
01:02:08operating system we can imagine because
01:02:11we're going from analog to digital text
01:02:14is black and white whereas this these
01:02:16models only understand context you know
01:02:19Daniel canaman just passed you know
01:02:21amazing kind of guy but you know he did
01:02:24have this concept of type one type two
01:02:25thinking and so we had one which is
01:02:28these big data things that can only
01:02:30extrapolate but now we have these models
01:02:31that understand context and so we have
01:02:33the missing parts of the brain and that
01:02:35will allow us to extrapolate allow us to
01:02:37have more rainbows you know have the
01:02:39context of each individual push
01:02:41intelligence to the edge and that's why
01:02:43again there is this imperative to do
01:02:45this now because there's a window on the
01:02:47freedom agency democracy side but the
01:02:50other imperative is no one should have
01:02:52to suffer as they're suffering
01:02:54now amazing and how much does actually
01:02:57need doesn't need that much which is the
01:02:59really amazing stuff this the total
01:03:01amount spent in generative I think I
01:03:03said at the conference is less than the
01:03:05total amount spent on the Los Angeles
01:03:07San Francisco
01:03:09Railway which hasn't even started yet
01:03:11and and in building stable Health again
01:03:14if if that's what it's called I mean the
01:03:15amount of capital required to build that
01:03:18is di Minimus compared to what spent on
01:03:20a single human trial of any any drug
01:03:24yeah it is but then you know you build
01:03:26it and you get to that 8020 incredibly
01:03:29quickly that will change hundreds of
01:03:31millions of lives and that will attract
01:03:32the smartest people in each of these
01:03:34areas thinking about what is the open
01:03:35infrastructure of multiple sclerosis of
01:03:37longevity of cancer and more but then
01:03:40you can amp that because the value is so
01:03:43so huge and you I hope to build a
01:03:46trusted organization as part of this
01:03:48whole human operating system upgrade
01:03:50yeah that's what I want to build I want
01:03:51to build human Os or at least catalyze
01:03:53it again I don't want to run or control
01:03:56own anything I want to check out how to
01:03:58give back that control because who
01:04:00should decide what cancer knowledge goes
01:04:02in there who should decide what
01:04:03education
01:04:04ET let's talk about the second half of
01:04:07your of your vision which is how we
01:04:10originally met um when you were one of
01:04:12the winners of the Global Learning X
01:04:14prize that uh Elon and Tony uh Robbins
01:04:18had had co-funded your your vision
01:04:21around education um speak to us about
01:04:25that yeah you know like we're deploying
01:04:27it um kind of the windows are kind of
01:04:29separate but every child my my entire
01:04:32operating system is like if you think
01:04:35about things in terms of the rights of
01:04:36child today they have no agency and so
01:04:38we must respect their rights climate
01:04:41everything becomes a lot simpler now
01:04:43that we have language models on a laptop
01:04:45like I said you can go to LM studio.
01:04:48download stable LM and it will run on
01:04:51your MacBook faster than you can
01:04:54read it's crazy we could have a gp4
01:04:57level AI from us or someone else on a
01:04:59smartphone or a tablet by next year One
01:05:02Laptop a child was too
01:05:03early you know now we have this
01:05:06transformative technology you have an AI
01:05:07that teaches the child learns from a
01:05:09child are you visual auditory dyslexic
01:05:11that's the best data in the world for a
01:05:13national model but also to teach these
01:05:15models how to be optimistic how to this
01:05:20really is this really is uh the young
01:05:23ladies Illustrated primer this really is
01:05:25is Neil Stevenson's Vision in that
01:05:28regard yeah but n shouldn't have had to
01:05:31find the primer she should have had it
01:05:33from day one as a human right as a human
01:05:37right yes our schools education system
01:05:41our child care mixed with the social
01:05:43status game mixed with petri dish you
01:05:46know they teach our kids not to have
01:05:49agency yes whereas they should be
01:05:51telling the kids yeah they should be
01:05:52teaching it's a relic of the Industrial
01:05:55age where everyone had to be counted and
01:05:57you can't measure what you can't manage
01:05:59so you manage the creativity and belief
01:06:01out of people everyone in the world can
01:06:04do anything why because even if you
01:06:06don't have that Talent you can convince
01:06:08someone else who does have that
01:06:10talent but they don't believe it so they
01:06:12can't do it so what happens if we have
01:06:14an entire nation of children that have
01:06:18this helper that brings the right
01:06:19information at the right time and tells
01:06:20them they can always believe that
01:06:22supports them entire
01:06:24world what can't you do you know then
01:06:27they have all of the councel knowledge
01:06:29of their fingertips and all of the
01:06:30engineering knowledge of their
01:06:31fingertips and it's a constantly
01:06:33learning adaptive and improving system
01:06:36again right now almost the entire AGI
01:06:39and AI debate is about these machine
01:06:42Gods train on giant supercomputers that
01:06:44bestow their beneficence down or may
01:06:46kill us or whatever what about that
01:06:48human operating system upgrade that is a
01:06:51decentralized intelligence where that
01:06:52kid in Mongolia or Malawi or wherever
01:06:55can make a real difference to humility
01:06:57some of the contributors to our open
01:06:58code basis for our models are 15 years
01:07:01old you they just taught themselves and
01:07:03just happen to be their Wizards you
01:07:05don't know in this new age right and
01:07:08again they should contribute to the
01:07:09whole because once something goes into
01:07:11this model of this system and again it
01:07:13needs the verification and other things
01:07:15that can be dynamic they can proliferate
01:07:17to everyone using that system do you
01:07:20think that once this capability is built
01:07:23it will run into blocks in different
01:07:27nations uh or do you imagine that this
01:07:29will become a again a a human right uh I
01:07:33and not all n i mean listen there's no
01:07:36greater gift and no greater asset you
01:07:39can give to a nation's populist than
01:07:42intelligence and education but I'm not
01:07:44sure every national leader wants to see
01:07:47that that's why I think again there is a
01:07:50gap here there is a year maybe where you
01:07:53can go to any national leader and say
01:07:56I will bring this technology to your
01:07:57people and I will empower the smart to
01:07:59your people and I want it to be eared by
01:08:00the people and what option do they have
01:08:03this is positive for them what happens
01:08:05is that a lot of the corruption in the
01:08:06world is because of local
01:08:08Maxima you know actually it's weird
01:08:10because unpredictable corruption is the
01:08:12worst predictable corruption is a bit
01:08:13like tax you know there's a good B book
01:08:15by fusser J at Harvard about this and
01:08:18then you have taxation kicking at
01:08:1914% um if you can show them something
01:08:22bigger and this is clearly big they will
01:08:24Embrace this technology set new norms
01:08:26and if you correct the same across all
01:08:28these countries with talented
01:08:29individuals in each of those groups and
01:08:31talented individuals in each of those
01:08:33sectors with a shared Mission even
01:08:34though they're separate
01:08:36organizations that's how you set amazing
01:08:38standards that's how you build a network
01:08:40effect and if you tie them all together
01:08:42with a intelligent protocol and again we
01:08:45talk about tokens or speculation or
01:08:46ramps or anything like that but taking
01:08:48the best of thinking around
01:08:50coordination that can work that can
01:08:53break this open you know um
01:08:56but it's not going to be everywhere and
01:08:58also when you look at the current debate
01:09:00the current debate is for example we
01:09:02can't let China have this
01:09:05technology and you're like what about
01:09:06the kids in China like well you know
01:09:09it's dangerous they can have AGI so
01:09:11under what circumstance would China ever
01:09:13have this technology never you know
01:09:15Pakistan when should they have thech
01:09:17never that's really what they kind of
01:09:19saying it's also self-defeating because
01:09:20China has 100 million people they can
01:09:22use to create data sets and two x of lot
01:09:24superc computers let's put side again
01:09:26it's a very Western oriented debate
01:09:28whereas actually if you go to these
01:09:29countries and you talk to the leaders
01:09:32and the family offices that have power
01:09:34and the people they will lead frog in
01:09:36the global South to a intelligence
01:09:39augmentation like they let frog to
01:09:41mobile they want to embrace this
01:09:43technology and again you can set Norms
01:09:45Now versus what's going to happen is you
01:09:48know they will get a centralized
01:09:49solution they'll adopt that instead if
01:09:51you don't right now for hundreds of
01:09:54millions billions of people
01:09:56that's why I think again it's a
01:09:59Crossroads um is there anybody else
01:10:02working towards this position that you
01:10:03know of no in the in the large AI yeah
01:10:07no certainly no credibility and again
01:10:10that's why I had to build these models
01:10:11you know and I had to kind of do this
01:10:14everyone's working on Tiny parts of this
01:10:16but they're expecting emergence build it
01:10:18and somehow it will spread and again
01:10:20this is why I found it passing in the
01:10:21web 3 Community there are good people in
01:10:23there and I hope to be able to unite
01:10:25them just like hope to unite the people
01:10:26in health and others again Peter you've
01:10:28seen people working on Tiny parts of
01:10:30this but this isn't a Manhattan Project
01:10:33where we're facing an enemy unless the
01:10:35enemy is ourselves you know but this
01:10:37does require this big Global coordinated
01:10:40push and that's why I've tried to design
01:10:42this system that I believe will work
01:10:44because it's all about the talent and it
01:10:46is multiplicative is the race against
01:10:51uh overly powerful centralized AI
01:10:54systems that achieve some version of AGI
01:10:57is that what we're racing
01:11:02against yeah again we're racing against
01:11:04ourselves like um humans could scale
01:11:08through stories you have organizations
01:11:09you know come and join abundance come
01:11:11and go to Oxford come and do this but
01:11:14then when we scaled through text text
01:11:17was a lossy information format and
01:11:18there's this poem by Ginsburg how about
01:11:21this carthaginian demon of disort of Moc
01:11:23that comes in Mallet comes in through
01:11:26the data loss our organizations are slow
01:11:28Dum AIS but now what's happening is
01:11:30they're configuring to achieve their
01:11:32thing of getting more and more power
01:11:34again corporations are technically a
01:11:36people under law but they're not fully
01:11:39formed people they eat our hopes and
01:11:40dreams so I believe the competition here
01:11:43is against those organizations
01:11:45consolidating too much power and
01:11:47creating Norms that are almost
01:11:48impossible to break so we're almost
01:11:50competing against ourselves and again
01:11:53the question is this do you believe it
01:11:55amp ifed human intelligence or do you
01:11:56believe in artificial general
01:11:58intelligence do you believe in
01:11:59collective intelligence or do you
01:12:00believe in collected
01:12:03intelligence who decides is this
01:12:05infrastructure or is this a product
01:12:09like so it's not like a Manhattan
01:12:12Project against you know the Soviets or
01:12:14anything like that but it this is
01:12:17require us all to come together or at
01:12:20least the smartest people in each of
01:12:22these areas from coordination to
01:12:24governance systems to healthcare to
01:12:26education with a blank slate of how do
01:12:28we upgrade the human operating system
01:12:30the time is now it's our last chance to
01:12:33do it and I and I I love you for it
01:12:36because I think you're you're right you
01:12:38were there when Elon beamed in on x x
01:12:43video over
01:12:44starlink from his airplane which is
01:12:47which was a fun moment and we were
01:12:49talking about uh the rate of growth and
01:12:53his his statement um cu you know Rick
01:12:55herw was there talking about his still
01:12:57his prediction of AGI by 2029 and Ray
01:13:01and and Elon saying we'll have AGI
01:13:04whatever that means by next year and and
01:13:07the intelligence of the entire Human
01:13:09Race by 2029 so I I am I am curious what
01:13:14just to close out what you think about
01:13:17that those timelines and that potential
01:13:21for a super intelligent uh AI system
01:13:24that is centralized because that's the
01:13:27people who are building that level of
01:13:29Power are building centralized systems
01:13:32they're building centralized single
01:13:34systems that again take our collective
01:13:36intelligence like all of YouTube In the
01:13:38case of open AI clearly and other things
01:13:41and they package it up sell it back to
01:13:43us but they don't give they don't
01:13:44care you know these organizations are
01:13:47trying to build a system that will take
01:13:48away our freedom Liberty and potentially
01:13:50kill us all let's be kind of fair about
01:13:52that direct and sell it to us uh on an
01:13:55incremental basis the selling to us is a
01:13:58complete Canad they don't care about the
01:14:00revenue of this again let's kind call a
01:14:03spade a spade they are telling you that
01:14:05they're building something that could
01:14:06kill you and something that could remove
01:14:10all our freedom and liberty and they're
01:14:12saying it's a good thing you should back
01:14:13them because it's cool it's not it's
01:14:15actually shameful if you think about it
01:14:18and we should not stand for it anymore
01:14:19this another reason I want to step aside
01:14:21to see because you can't say things like
01:14:22that I got canceled Sil Valley so many
01:14:24times but realistically it's ridiculous
01:14:27and it should not be stood
01:14:28for but they're going to do it anyway
01:14:31because they have the political power
01:14:32people are scared of
01:14:34them so there has to be an alternative
01:14:37and the alternative has to be
01:14:38distributed intelligence when I resigned
01:14:40I said you can't beat centralized
01:14:43intelligence with centralized
01:14:44intelligence you're not going to beat it
01:14:45with a stability this is a great
01:14:47organization it's going to do well the
01:14:49only way that you can beat it to create
01:14:51the standard that represents humanity is
01:14:54decentralizing intelligence it's
01:14:56collective intelligence and the data
01:14:59sets and Norms from that will be ones
01:15:01that help children that help s people
01:15:04suffering that reflect our moral
01:15:06upstanding and the best of us and
01:15:09gathers the best of us to do it because
01:15:12if you work in healthcare if you work in
01:15:14education if you work in finance if you
01:15:15work in any of these things there's no
01:15:17Organization for you to come and join or
01:15:19partner with on this there's no kind of
01:15:23centralized Mission I have looked I've
01:15:24want to help other people I don't want
01:15:26to do this myself and I don't want it to
01:15:28be about me very very quickly which is
01:15:29why I'm kind of getting it out there now
01:15:31I hope I can capitalize something that
01:15:33then people will take forward and time
01:15:36is now for that because AGI when it
01:15:38comes if it comes again there's various
01:15:39definitions of this why on Earth do you
01:15:42need any knowledge workers anything that
01:15:43can be done via a laptop doesn't need
01:15:47humans and so you have concepts of Ubi
01:15:50here you have concept I think when the
01:15:52AJ comes you don't need money money is a
01:15:54common story as a common good we HD
01:15:56towards a post capitalist Society yeah
01:15:59yeah I think the example I think you
01:16:00said was uh Star Trek versus Mad Max you
01:16:03know I'm like Star Trek versus Star Wars
01:16:06I think is AIT a better
01:16:08one you know and so you got the Sith
01:16:11Lords and all of that um but again if
01:16:14you kind of look at this I don't think
01:16:16we need money like it's cross contextual
01:16:18like bartering with our AI systems
01:16:20representing us or it's you don't need
01:16:23money because you're told what to do
01:16:25again our governments the definition of
01:16:28a government is the entity with Monopoly
01:16:30on political violence and an AGI can
01:16:32overtake any government that can then
01:16:34control the
01:16:35people because again listen to it
01:16:37whispering look at the kind of human
01:16:39thing so we have this opportunity to set
01:16:41Norms right now the way that the big
01:16:44labs are going to AGI is likely to kill
01:16:46us all Elon and I signed that six-month
01:16:49pause letter because even though people
01:16:50like emad you're an accelerationist you
01:16:52put all this open source AI out you have
01:16:54to think about the other and who's
01:16:55involved in that discussion and again if
01:16:57we build an AGI as a centralized thing
01:17:00is Windows or Linux safer as
01:17:04infrastructure our entire internet
01:17:06infrastructure is built on open open can
01:17:09be challenged open can be
01:17:10augmented a monolith is like to be crazy
01:17:14and the way that I put this is you and I
01:17:17both know so many
01:17:18Geniuses you know side effect of Genius
01:17:21is insanity honestly we're not ment
01:17:24Geniuses are not mentally
01:17:25stable why would you expect an AGI to be
01:17:28so and you're putting all your EGS in
01:17:29one basket versus creating a complex
01:17:32hierarchical system that is a hive mind
01:17:36that's intelligence that represents us
01:17:37all we should be working towards
01:17:39building that because it's safer it's
01:17:41better it achieves all the benefits that
01:17:44people are talking about and it's
01:17:45possible today do you think Elon shares
01:17:47in this vision of a de centralized AI do
01:17:49you think he would play in that area and
01:17:51do you think any of the national leaders
01:17:53that you've been speaking to would
01:17:55support that kind of a vision as
01:17:58well um yeah I can't speak veryy long
01:18:00I'll speak to him and see what he thinks
01:18:02and I'll get back to you you know he
01:18:04always says what he thinks um but you
01:18:06know he's immensely concerned he was one
01:18:08of the leaders in this area saying
01:18:09originally why Google you know now why
01:18:12Microsoft open AI like it can't be
01:18:14centralized but it's difficult this is a
01:18:16difficult question how many people have
01:18:18a feasible solution or you even thought
01:18:19about this properly you and I both know
01:18:21just not many and that's very sad should
01:18:23be everyone thinking about this
01:18:25on the leader side all the leaders I've
01:18:27met are super happy you know because
01:18:29they again leaders want Power they want
01:18:33control and all of this but genely like
01:18:36they want to see
01:18:38abundance they're not happy with where
01:18:40their countries are and embracing this
01:18:42technology they know that they can leap
01:18:43ahead and you know they will still have
01:18:46a say in all of this it's not like it's
01:18:48kicking them out or removing them there
01:18:50were still various kind of mechanisms
01:18:53there and ultimately improving the
01:18:55health education and capability of your
01:18:57people is not a bad thing I mean like
01:19:00obviously I haven't talked to the
01:19:01completely oppressive leaders you know
01:19:03maybe that'll be an interesting thing
01:19:05but honestly I don't want to even be
01:19:07talking to leaders I want to create
01:19:08again a system the people of the country
01:19:10coming together with the franchise
01:19:12system can then build this technology
01:19:14for the good of their people in the open
01:19:16and not be reliant on anyone politically
01:19:18or any other type of thing like that we
01:19:19don't need giant super computers for
01:19:21where we're going we need
01:19:22coordination need a few giant super
01:19:24computers yeah what's your timeline for
01:19:27putting out this white
01:19:29paper I'm working as hard as I can you
01:19:31know put it together I've held I've held
01:19:33I've held you to this a number of times
01:19:35I've said get the vision out there it's
01:19:37it's getting though we're about to go
01:19:39off this call to a 4our session to
01:19:41dictate all the various bits and pieces
01:19:43and again it was impossible when I was a
01:19:44CEO of stability there was always
01:19:46another fire there was always another
01:19:47thing I didn't have time to think you
01:19:50know and I hope people can take that
01:19:51white paper and make it better I I don't
01:19:53have all the answers I'm just trying to
01:19:54Capal something man I think after I
01:19:56heard you Ste down I wrote you a text
01:19:58saying
01:19:59congratulations yeah exactly not
01:20:01commiserations time time to time to feel
01:20:05Unleashed yeah um yeah uh Imad thank you
01:20:09my friend uh thank you for sharing uh
01:20:13where you are what led up to this where
01:20:16you're going next and uh and really
01:20:18pulling the gloves off on discussing the
01:20:20idea of centralized um closed AI system
01:20:24s and their dangers uh and the
01:20:27importance of of the vision that you
01:20:29portrayed because I'm I'm fully
01:20:32supportive and full believe that what
01:20:34you laid out um is probably one of the
01:20:38most sane visions of AI in the future
01:20:41that I've
01:20:42heard I hope other people agree you know
01:20:44and they can take it Forward other real
01:20:47heroes thank you thank you
01:20:53pal
🎥 Related Videos![What vaccinating vampire bats can teach us about pandemics | Daniel Streicker](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FlSPxeA6Z_m4%2Fhqdefault.jpg&w=3840&q=75)
![a16z Podcast | Things Come Together -- Truths about Tech in Africa](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fjbl5I7kYWHM%2Fhqdefault.jpg&w=3840&q=75)
![2024 TSCRS Applications of anterior segments diagnostic instruments in cataract surgery](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FWnUDk5vKuqE%2Fhqdefault.jpg&w=3840&q=75)
![a16z Podcast | The Infrastructure of Total Health](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FoLGRslHa5FE%2Fhqdefault.jpg&w=3840&q=75)
![The Robot Lawyer Resistance with Joshua Browder of DoNotPay](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FAmVdYPTdw2c%2Fhqdefault.jpg&w=3840&q=75)
![NES Controllers Explained](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FnAStgQzPrAQ%2Fhqdefault.jpg&w=3840&q=75)
![What vaccinating vampire bats can teach us about pandemics | Daniel Streicker](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FlSPxeA6Z_m4%2Fhqdefault.jpg&w=3840&q=75)
What vaccinating vampire bats can teach us about pandemics | Daniel Streicker
![a16z Podcast | Things Come Together -- Truths about Tech in Africa](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fjbl5I7kYWHM%2Fhqdefault.jpg&w=3840&q=75)
a16z Podcast | Things Come Together -- Truths about Tech in Africa
![2024 TSCRS Applications of anterior segments diagnostic instruments in cataract surgery](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FWnUDk5vKuqE%2Fhqdefault.jpg&w=3840&q=75)
2024 TSCRS Applications of anterior segments diagnostic instruments in cataract surgery
![a16z Podcast | The Infrastructure of Total Health](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FoLGRslHa5FE%2Fhqdefault.jpg&w=3840&q=75)
a16z Podcast | The Infrastructure of Total Health
![The Robot Lawyer Resistance with Joshua Browder of DoNotPay](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FAmVdYPTdw2c%2Fhqdefault.jpg&w=3840&q=75)
The Robot Lawyer Resistance with Joshua Browder of DoNotPay
![NES Controllers Explained](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FnAStgQzPrAQ%2Fhqdefault.jpg&w=3840&q=75)
NES Controllers Explained
🔥 Recently Summarized Examples![4 Steps to Master Any Complex Skill (quickly)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FpHN7BXpdAPQ%2Fhqdefault.jpg&w=3840&q=75)
![40 Years of Fitness Experience in Less Than 11 Minutes.](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2F2ZsHsX4hvlk%2Fhqdefault.jpg&w=3840&q=75)
![Gun Controlling Media Makes FATAL Mistake... They Have Tied Their Fate To Biden's & Gun Rights Win](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2F8L34kAwjriw%2Fhqdefault.jpg&w=3840&q=75)
![GET READY! Palantir Is Officially The Next Nvidia.](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FscpVS5--ikw%2Fhqdefault.jpg&w=3840&q=75)
![Abundant Thinking: The Hidden Key to Get Everything You Want (Audiobook)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Far9dcxJ11H4%2Fhqdefault.jpg&w=3840&q=75)
![The Coming Demonic Invasion (Revelation 9:12–21)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FDYzkCtbwHqU%2Fhqdefault.jpg&w=3840&q=75)
![4 Steps to Master Any Complex Skill (quickly)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FpHN7BXpdAPQ%2Fhqdefault.jpg&w=3840&q=75)
4 Steps to Master Any Complex Skill (quickly)
![40 Years of Fitness Experience in Less Than 11 Minutes.](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2F2ZsHsX4hvlk%2Fhqdefault.jpg&w=3840&q=75)
40 Years of Fitness Experience in Less Than 11 Minutes.
![Gun Controlling Media Makes FATAL Mistake... They Have Tied Their Fate To Biden's & Gun Rights Win](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2F8L34kAwjriw%2Fhqdefault.jpg&w=3840&q=75)
Gun Controlling Media Makes FATAL Mistake... They Have Tied Their Fate To Biden's & Gun Rights Win
![GET READY! Palantir Is Officially The Next Nvidia.](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FscpVS5--ikw%2Fhqdefault.jpg&w=3840&q=75)
GET READY! Palantir Is Officially The Next Nvidia.
![Abundant Thinking: The Hidden Key to Get Everything You Want (Audiobook)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Far9dcxJ11H4%2Fhqdefault.jpg&w=3840&q=75)
Abundant Thinking: The Hidden Key to Get Everything You Want (Audiobook)
![The Coming Demonic Invasion (Revelation 9:12–21)](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2FDYzkCtbwHqU%2Fhqdefault.jpg&w=3840&q=75)