00:01hi everyone and big thanks for taking
00:03the time on you know we've been trying
00:05to get you on the podcast since uh we
00:06started it two years ago so we are super
00:08pleased that we that we have you on and
00:10indeed on your xplatform how cool yeah
00:14uh it's pretty cool yeah I mean you have
00:16like lots of people from all around the
00:18world uh simultaneously do effectively a
00:22realtime podcast and uh it works pretty
00:25well very good well we have so much to
00:37uh love to kick off with with AI um now
00:41what's your take on where we are in the
00:45now wow that's a long answer
00:48um there's there's so much happening in
00:50AI is the fastest advancing technology
00:53that I've ever seen of any kind and I've
00:55seen a lot of Technology um you know
00:58barely a week goes by with without some
01:03so uh and if you look at the amount of
01:07uh AI Hardware the computers coming
01:10online that are dedicated to AI that is
01:12increasing what looks like at least by a
01:16factor of 10 every year if not every six
01:20months so when you combine the hardware
01:25um Coming online really order of
01:27magnitude increase every you know call
01:30at least every 9 months um and uh many
01:35many software breakthroughs
01:38uh if you look at that that curve it
01:40looks insane so I think we'll
01:44um my my my guess is that we we'll have
01:48ai that is smarter than any any one
01:51human probably to around the end of next
01:58AI the total amount of sort of sentient
02:02compute of AI I think will probably
02:04exceed all humans in 5
02:06years what what is the what is the race
02:09about just now is it algorithms is it
02:11people is it computing power what what
02:14is it about just now is it the supply of
02:19it yeah last year it was uh chip
02:22constraint um and the hardware
02:25deployment if you break it down into the
02:28three areas of people um data and
02:31hardware and starting with Hardware last
02:34year it was about a shift Supply people
02:36could not get enough um in video chips
02:39particularly um this year it's starting
02:41to transition to a voltage Transformer
02:44Supply so actually getting enough
02:47voltage Transformers uh put in place so
02:51my sort of very Niche joke is
02:52Transformers for Transformers because a
02:55lot of the AI That's run is called a
02:57Transformer so you need Transformers to
03:00um and then next in the if you look out
03:03a year or two or certainly three years
03:06um it's just electricity
03:10so that's those those constraints in the
03:16um so many of the smart world's smartest
03:18people are are doing AI people that
03:21would have done physics before in fact
03:23or had have done physics for example
03:25have moved into AI because it's just the
03:28fastest moving field
03:30so we're seeing a lot of the best
03:32talents a lot of the smartest humans
03:33going into Ai and then uh we see along
03:37with that algorithmic
03:39breakthroughs um and then then you start
03:42hitting the the wall with the the data
03:47the you know you can fit all books ever
03:52written um just the text the the text in
03:55compressed form uh on one hard drive or
04:01computer um so when you when you're
04:03looking at like so called tokens to
04:05train on yeah uh and you you still think
04:10of like all the books ever written in
04:11every in in all languages by All Humans
04:14sounds like a lot certainly it's far
04:15more than any one human could could ever
04:18read um it actually is a small it's a
04:21small number of train training tokens
04:23it's just not enough so then you you
04:26start having to look at all the videos
04:28of I created um you know all the
04:31podcasts all the everything
04:35um and and you start even running out of
04:38data there well hopefully they hopefully
04:41they will include this podcast uh that
04:43definitely will include this podcast
04:46what's the biggest challenge you have
04:49xai well xai is still relatively new so
04:56uh like the limiting factor right now is
04:59just training our Gro version 2 model
05:02which should be do we think better than
05:05GPD 4 um and that's we're hoping to
05:09complete that in May so that's that's
05:12training right now so it's just really
05:14we're just trying to get enough gpus
05:16online to train it fast enough to get
05:19that done in May um which I think
05:25then and and that's with uh roughly
05:30h100s uh and and doing I think very
05:33efficient training then the next step
05:36would be for GR 3 which would be I guess
05:40G55 or Beyond uh would you know requires
05:43uh 100,000 Nvidia h100s training
05:48coherently so that's you know a half
05:51order of magnitude basically more
05:53training um and then you really start to
05:56have running into this data problem
05:57where you you have to either create
05:59synthetic data or use real world video
06:02those the the two sources of kind of
06:04like unlimited data are synthetic data
06:08and real world video which I should say
06:10Tesla has a pretty big advantage in real
06:12world video um Tesla has by far the most
06:15real world video of anyone yeah you've
06:17got a huge Library there so when do you
06:19think so when do you think we'll see
06:24AGI well it depends on how you define
06:26AGI if you define AGI as smarter than
06:28the smartest human I think it's probably
06:31end of next year like like within two
06:35years um but but that's that there's
06:38still there's still a pretty big leap
06:40beyond that to say smarter than the the
06:42machine augmented human Collective so
06:46like is it smarter than all humans
06:48working together uh who are also using
06:51computers to augment their output and
06:55that that I think is probably five years
06:57away one one way to look at it is is is
07:00assess um like roughly what is the ratio
07:04of digital to biological compute last
07:07question on um on AI any new thoughts on
07:09regulation and um how it should be
07:12structured well I I think we probably do
07:14need some sort of regulatory authority
07:15to look at the safety of AI um just as
07:18we have regulatory authorities and other
07:21Arenas to um you know o oversee aircraft
07:26and the safety of aircraft and cars and
07:30and other things you know medication
07:35uh now the rate at which AI is
07:37progressing is is fast is faster than
07:40probably any regulatory agency can keep
07:44um but but I do have a comment on what I
07:48think is very important
07:50before achieving safe AI which is that
07:53uh it's very important to train the AI
07:55to be as truthful as possible um and not
08:01uh yeah just to be as truthful as
08:06um I think you can get some very
08:08dangerous things when you program an AI
08:10to be politically correct think that
08:14seem uh relatively innocuous now but
08:17will not be so in in the future if AI
08:20has immense power you can take the
08:23Google Gemini example where it it
08:26refused to publish to produce a picture
08:28of George Washington as a white man and
08:32and any in fact any historical figure
08:35would automatically be made
08:38diverse um because it's been programmed
08:41to insist on diversity which sounds you
08:43know perhaps okay at first but not if
08:46the AI has so much power that it can
08:48actually enforce diversity and decide
08:51there's too many of one kind of people
08:52or too many of one sex and kill
08:55off just just kill off enough until the
08:58the diversity number is
08:59is what it's programmed to believe is
09:01correct but don't you think this will be
09:03sorted out in the next
09:05version no no they'll make it more
09:08subtle okay and less obvious but it will
09:12there okay well we'll see but where
09:15where is China where do you where is
09:16China now in I relative to the
09:21US um I I don't know exactly what China
09:23is uh except there are a lot of very
09:27smart people in China um and they they
09:30won't be they won't be far behind the
09:33rest of the world or far behind the US
09:37um I mean the ai ai right now
09:40is very concentrated in San Francisco
09:45London um and then you know there's
09:49there's you know a lot happening in in
09:53China but I I'm I don't have insight
09:55into what they're doing uh except that
09:57they I'm confident they will not be f
09:59behind uh what is developed in the west
10:04so but but but mark my words the if if
10:09uh if we do not program an AI to be as
10:12possible that that is where it will go
10:16arai that is where the danger
10:20lies yeah mov moving T here moving to to
10:23Tesla um is is the EV conversion now
10:27going slower than you had expected just
10:29where is the speed of conversion now
10:34expectations I think it's going quite
10:36fast actually especially Norway um
10:38absolutely well it's pretty much all
10:40there is is your your
10:42Teslas yeah there's a lot of Teslas in
10:44Norway it's crazy thanks i' once again
10:46like to thank Norway for the support of
10:57it's we will the the that that electric
11:01that all vehicles will go fully electric
11:04uh it's only a matter of time um that
11:07includes aircraft ultimately and boats
11:10um obviously trains the only thing that
11:13is ironically difficult to well you
11:15can't really make it electric is Rockets
11:17because you need you can't get away from
11:20um having to expel Mass uh you sort of
11:23Newton's third law um but but all cars
11:28will be Electric it's only matter of
11:29time and we'll look back on combustion
11:32cause in the same way that we look at
11:37um that that I was it was inevitable
11:40that there would be internal combustion
11:42cars and and it's just as inevitable
11:44that o cost will go electric um and um
11:49there will be some e and you know so
11:51like it going to be a completely
11:53straight up line there will be some uh e
11:56and flow in how how far electric car is
11:59going but that but the ultimate um
12:03victory of electric cars is inevitable
12:05um and and I think the sooner we get
12:07there the better yeah how do you see the
12:10Chinese competition here
12:13now we generally find that the companies
12:16in China are the most competitive in the
12:17world and certainly in uh electric
12:21vehicles or cars in general the Chinese
12:23car companies are by far the most
12:25competitive um yeah that's where where
12:28we find the most toughest toughest
12:31challenges that they make great cars and
12:33they work very hard so when you ride in
12:36one of the Chinese cars what do you
12:37think I mean you're an engineer you know
12:38what about it what do you what do you
12:41think I haven't R I have not ridden in
12:46but uh because they're not all available
12:49here you know in the US or very few are
12:52available in the US um some are
12:54available in Europe um but from what my
12:57team tells me there are very
13:04good moving moving out out in space
13:09um what what would it take to be
13:15Mars to be self-sufficient in Mars it's
13:18really about the the total tonnage that
13:20is delivered to the surface of Mars um
13:27well um I I think it's probably on the
13:29order of a million tons maybe it maybe
13:31more but somewhere between probably a
13:32million tons and 10 million tons are
13:35needed to make Mars self-sufficient and
13:39that well I gave a presentation on this
13:41recently if people look at my my recent
13:46but if you if you have
13:51U really uh if you have 100 tons per
13:56flight you need 10,000 flights to get to
14:01tons um and that's 100 tons landed to
14:03the surface of Mars so in order to get
14:06100 tons land to the surface of Mars you
14:08need 500 five times that number in Earth
14:11orbit um so we do a lot of orbital
14:16refilling um so launching sort of uh
14:20Rockets uh tanker ships over and over
14:22again that that would replenish the
14:24propellant of the ships that would go to
14:27Mars um and then youd need a on roughly
14:29on order of 10,000 of them to get to a
14:36uh but we we plan to do that that that's
14:40uh that's we think we can get that done
14:43within 20 years really so and when do
14:46you think so when do you think we'll be
14:51time first first uh well the
14:54first Starship that will land on Mars
14:58which obviously would not not have
14:59people at first I think it's probably
15:01within about 5 years um and then it
15:04would probably launch several ships and
15:07just confirm that they can land okay on
15:10Mars um we'll also be doing the moon
15:13simultaneously with that so uh go taking
15:16well I think I think we'll get people
15:18back to the Moon I should say within 5
15:19years and we'll get uh uncrewed ships
15:23landed on Ms within 5 years and and then
15:26we be building up the production rate um
15:29and improving the design of the booster
15:34um so in the first people on Mars I
15:37think within seven years or so seven to
15:41years um and from from there we need to
15:45rapidly increase we need massive numbers
15:47of shifts going and Earth and Mars Only
15:51are in the same quadrant of the solar
15:54system roughly for six months every two
16:00it's only possible to really transfer
16:03efficiently um from Earth to Mars I say
16:07every six months but really there's
16:08about there's a couple months where
16:10where it's ideal every 26 months um so
16:14every two years that you would see a
16:16basically a fleet depart Mars and I
16:17think it be quite a spectacular thing to
16:19see a thousand ships depart from Mars
16:23all at once like Battle Star
16:24Galactica what kind of new technology do
16:26we need before we'll be self-sufficient
16:30there actually I think we have all the
16:33tech we already know all the technology
16:34that's necessary for that it just needs
16:36we just need to build so no new physics
16:38is needed for this why is it so
16:42you I think it's important for
16:46general um so if if we wish to maximize
16:53Consciousness then being a multipled
16:56species will result in a much longer
16:59uh existence of Consciousness
17:01Consciousness than if we're on one
17:02planet if we're on one planet we're
17:04simply biting our time until there's
17:07eventually a Calamity it could be soon
17:09it could be a long time but eventually
17:11something will happen it could be you a
17:13global Therman nuclear war it could be
17:15simply That civilization merely
17:18subsides our civilization may not die
17:21with a bang it may die with a whimper
17:25gradually falling into obsolescence but
17:28if we're multiplet species then we've
17:30got two planets and and they can support
17:33each other um and we can go beyond two
17:36planets ultimately to the moons of
17:38Jupiter to the to the
17:41uh um Beyond to the the outer parts of
17:45the solar system and ultimately to other
17:48systems so this tiny this tiny candle of
17:51Consciousness that we have in this vast
17:56extended um and Amplified and we're just
18:01far more likely to uh survive as for for
18:04Consciousness to survive if we are
18:08species you don't think it' be better to
18:10use all these resources and try to sort
18:14Earth well just to put this into
18:16perspective the amount of resources I'm
18:18talking about for making life
18:19multiplanetary would be less than 1% of
18:23Earth so really you can think of it as
18:25resource allocation do you think it's
18:26worth spending half a percent of Earth
18:29Resources to ensure uh that we have
18:33redundancy in Consciousness and that we
18:35extend Consciousness Beyond Mars to
18:38other planets to to Mars and other
18:40planets and ultimately other star
18:42systems um and then also take into
18:45account the fact that there are certain
18:46inevit there are certain things we
18:47simply cannot avoid on Earth um like is
18:51it within your power of mind to stop
18:53World War I I don't think so no if it
18:56happens um and if we have theral Warfare
19:00our technology level will drop to the
19:02age um and we may never survive and then
19:06there are we maybe get may get hit like
19:08by a comet like the dinosaurs and um you
19:13know if the dinosaurs had spaceships
19:15they they'll probably still be around um
19:18so and then if if you wait long enough
19:20the Earth the the sun will continue to
19:22expand and eventually engulf Earth and
19:25destroy it and destroy all life so just
19:28to give it amount a certain amount of
19:30time no matter what you do on Earth no
19:33matter how careful you are um Earth will
19:36life all life on Earth will die that it
19:44certainty on a slightly less gloomy note
19:47uh X Twitter yeah um what is your vision
19:53now what do you how do you see the the
19:56x i goal of X is to be the best source
20:01internet um and I think we're making a
20:04good you know good progress there and I
20:06mean this it's going to be like I call
20:08the everything app like if anything you
20:10want to do you can do on the xplatform
20:13um whether it's text audio video uh
20:17payments Financial stuff
20:20um Communications of all kinds um and
20:23then but but then also where there is
20:25publicly disseminated information is to
20:27be the best source of Truth um and I
20:29think it I think it already is that um
20:31now people may say oh there's some piece
20:33of misinformation disinformation I say
20:34yes but look look at the replies the
20:36reply is correct that misinformation and
20:38look at Community notes and the and how
20:40good the batting average of community
20:42notes is it's extremely good it's by far
20:44the best factchecking system on the
20:46internet um so and and a lot of people
20:49still labor under the illusion that the
20:51the the Legacy newspapers that they read
20:53are actually true there's so much
20:56nonsense in them I mean ni
20:59how many times when do you read an
21:00article in a newspaper where you know
21:02the circumstances of what that article
21:04is and how often is it spot on no of
21:08course it's uh normally no no of course
21:10we all know it's normally wrong but but
21:12how do you look not sure but how how do
21:14you look at the situation now for
21:15instance with with Russia uh you know
21:18the work Russia does in Germany with
21:20fake accounts on it's pretty pretty huge
21:25right I mean we don't see a lot of
21:28Russian it to be frank um on the system
21:35so we we see very little
21:38um we do we do see a lot of lot of
21:41attempts to influence things but they
21:43seem to be coming from from the West not
21:46Russia right what about um what about
21:50things like the latest developments in
21:53Brazil and so on yeah sure yeah so the
22:00uh we we kept getting these demands from
22:04um uh this uh Judge Alexander um that's
22:10his that's his name on Twitter Alexander
22:14um and there would be to suspend
22:17accounts um immediately we're given
22:20typically two hours to suspend an
22:22account or face massive fines um and the
22:26the final sto we were were being given
22:28given demands to suspend sitting sitting
22:31members of the parliament and major
22:34journalists and moreover we could not
22:36tell them that it this was at the beest
22:39of uh Alexander Morales we had to
22:42pretend that it was due to our rules of
22:45service and that was the final straw and
22:47we said no now um when you when you
22:51bought Twitter um now renamed X did you
22:55expect that you would end up in these
22:56type of situations so it's is all
23:01well I knew it wouldn't be just a total
23:04B of roses um you know and it's fing I
23:13wouldn't um no I mean I thought it would
23:16be since we're just like rigorously
23:20trying to pursue the the the goal of
23:24being the most accurate and truthful
23:26place in the internet and that that
23:29what is said is always true or accurate
23:30but is it is perhaps another way to
23:32frame it is as the least inaccurate
23:34place on the internet do you do you see
23:36clear do you secretly think this is a
23:39fun it's fun yeah yeah it's fun at times
23:42it's stressful at times and it's fun at
23:44times um but overall we're trying to
23:48serve the people of Earth um and and and
23:51this is sort of an S sort of maybe an
23:52esic way of viewing it but um to try to
23:56be kind of like the the group
23:57consciousness of Earth so you can think
24:00of like if each person is like a neuron
24:03contributing to like the collective
24:04brain of Earth and you want to try to
24:06minimize the noise and maximize the
24:08signal of every neuron that's connected
24:10to the the X Network that that's
24:13basically what what is what is the
24:15collective will of of humanity and and
24:19how and and and how to yeah just serve
24:22the collective will of humanity and so
24:24serve the greater good that that's our
24:29there there's definitely going to be
24:30people who want to manipulate that
24:32information and so we have to fight that
24:34and try to have uh you know be it be the
24:38most accurate place as part to the best
24:40of our ability and have it be kind of a
24:42Marketplace of ideas where people can
24:44propose ideas and you know debate them
24:49and um I think so far it's working
24:53reasonably well in that regard um now
24:56people that don't like the truth will
24:58or if they want to manipulate things
25:00they will not like it but only but only
25:03a few years ago you were you were a guy
25:06um producing electric vehicles now you
25:10are you know through starlink you've had
25:13some you know I mean some big impact in
25:17Ukraine uh with Twitter you are kind
25:20of into some issues in uh you know
25:25Turkey um you know you're becoming like
25:29force and a really important one how do
25:32you how do you look at
25:36that well like I said I'm really I'm
25:40to take the set of actions that maximize
25:42the probability that the future is good
25:45um I mean we have to keep
25:48civilization going onward and upward as
25:51much as possible and um and try to
25:54minimize the civilizational threats that
25:56occur um you know we we we can't get to
26:00Mars if civilization collapses it's not
26:06um you know we've got to we've got to
26:10um keep civilization going um and I
26:14think we should view our civilization as
26:17being much more fragile than we think we
26:20kind of take for granted oh it's always
26:21going to be there but actually if you
26:23study history you realize that there
26:25Rise you know there's rise and fall
26:29um I mean I was I was reading in depth
26:33samarians um who were arguably the first
26:37civilization if you call civilization
26:39like writing and stuff you know they
26:40were the first to develop writing um and
26:44uh but eventually they died out and they
26:49gone so and then nobody could read the
26:51writing at all and and they they just
26:55civilization um but they're pretty
26:57impressive in their time
26:59and the ancient Egyptians the same thing
27:02um and uh you know one sort of one after
27:05another uh ancient Greek had it Greece
27:08had its day uh you know China and India
27:11had will have incredibly impressed
27:14populations but there's been EVs and
27:15flows in the uh CH China and Indian
27:18civilizations over the the the aons you
27:26so you know I I guess I'm just trying to
27:29take this this set of the steps that
27:34um increase uh the scope and scale of
27:38Consciousness that's that's what I'm
27:39trying to do it's not it's not that I'm
27:41trying to have a put a political thumb
27:44on the scale or anything like that um
27:46but I I think I'm trying to have the
27:48political will go where the people want
27:55go you you mentioned some um
27:58some uh uh really smart people here and
28:01um kind of just moving t a bit here to
28:03Copa culture now you manage a lot of
28:05geniuses in your in your companies what
28:08is the key to manage really smart people
28:14think I don't I don't think I manage
28:17smart people they manage themselves
28:23think well I guess with really smart
28:25people you know I don't really think of
28:27of it like managing them I think that if
28:30somebody's very smart and talented they
28:32they can go anywhere and do anything
28:33anytime like if they they they don't
28:36have to work with me they could go
28:37anywhere so I I really just say like
28:40look this is the the goal we're after
28:42and this is what we're trying to achieve
28:44and do you agree with this goal and if
28:45you do then let's try to get it done um
28:49and um you know provide my opinion along
28:53the way and once in a while I'll say
28:55look guys you just got to trust me on on
28:57this one we got to do this thing and if
29:01it turns out to be a bad
29:03decision you we can can all hold that
29:05against me in the future but you have an
29:07incredible eye for detail right I mean
29:09when we read the is book um it's pretty
29:13clear that you I mean you really are are
29:15deep into detail and know what you talk
29:17about so how do you how do you balance
29:18this um kind of micromanagement of some
29:21areas and then delegate
29:25other I wouldn't I wouldn't call it
29:30um it's just insisting on atttention to
29:34detail that um if you're trying to make
29:37a perfect product you must have
29:39attention to attent attention to details
29:44um and I haven't actually read the
29:46isacon book you should it's very good
29:50it well I I asked Walter isacon if I
29:54should read it and he said I shouldn't
30:00then he said I shouldn't read it so okay
30:03well I'll I'll ask you some questions
30:05from the book then they you he talks
30:06about you know you the kind of the
30:08hardcore and Ultra Hardcore
30:11culture what is an Ultra Hardcore
30:14culture I guess it's work I mean it's
30:15working culture right I mean how how I
30:18mean Ultra hard work how hard is
30:23that well when things get really intense
30:25you're basically just working every
30:29and how and how long can you do that
30:35for well continuously for sometimes like
30:39a few years what does it what does it du
30:42to you it really it's
30:46pain um and and every waking hour maybe
30:49it's an exaggeration because there are a
30:51few hours um obviously with friends and
30:54family and and critical other things um
30:58but 100 hour weeks would be I I've done
31:00many many stretches of 100 hour weeks
31:02like true 100 hour weeks
31:04um where roughly six hours per day is
31:07sleeping um I would not recommend that
31:10this is not that's for emergencies you
31:12know it's not uh all the
31:15time um you know during very difficult
31:19times at Tesla I've had to do that and
31:22sometimes at the beginning of my earlier
31:24start offs I did that where I just
31:26wouldn't leave the office I would just
31:27sleep under my desk and just work seven
31:32um sometimes it's necessary for success
31:36failure um but but do you you do you
31:39enjoy being in this crisis
31:49sucks okay no I I don't want to be
31:56pain but sometimes it's the between
32:04failure when you make decisions how
32:10speed he just gave me an idea which is
32:13um I'm GNA invite the uh Judge Al
32:17Alexander R uh to do a spaces and then
32:21he can explain why what I'm doing is bad
32:26and and and maybe he's right
32:28I challenge I challenge him to a
32:32spacers sounds good yeah but what about
32:35when you make when you when you make
32:37decisions how how important is speed and
32:39how do you how do you balance analysis
32:48the the best offense and defense is
32:52speed if you think of something like the
32:54SR71 Blackbird it really had almost no
32:58accelerate and it was never shut down
33:00even once like I think over 3,000
33:02missiles were shot at the SR71 blackb
33:07hit and and really what it did was just
33:11go faster so the the power of speed
33:15is uh underappreciated as a competitive
33:21Dimension um is that
33:23why um you know Space X expenence has
33:28because you've been mean and lean as an
33:31organization and fast think speed speed
33:35is uh definitely a factor now I should
33:37say you want to go in the case of a
33:39company you you need to be a vector not
33:41a scaler so it can't be you need to go
33:45at high speed in the right
33:54so and no company's going to be going in
33:56the right direction all the time you
33:59Corrections like a guided missile you
34:01qued course Corrections um and uh but in
34:06the case of SpaceX it's like okay our
34:08goal is to extend Humanity beyond Earth
34:12um and we didn't even know how to even
34:15frame the question correctly like what
34:18what which knew that that was the
34:21goal um we didn't know what pent we' use
34:25or what the raw materials would be or
34:27for the how would the rocket be built
34:28how would it be designed what's actually
34:32important um and uh you know so for
34:36example going from our Falcon
34:40architecture which is um uses refined
34:43jet fuel and liquid oxygen um in a um
34:49open cycle gas generator architecture
34:52engine to a to Starship which is a
34:58uh liquid methane liquid oxygen
35:01um uh propellant uh in a staged
35:05combustion very high pressure
35:09um that that that's that's a big
35:11architectural change um but we didn't
35:13know that we we needed to make that
35:15architectural change until we're pretty
35:18road like about halfway took us about 10
35:21years to figure out that was even the
35:22right architecture now we're confident
35:24it is um just um we were on um uh risk-
35:27taking and so on I think SpaceX is one
35:29of the best example I know about uh what
35:31we call failing well right learning from
35:33mistakes and moving on um what generally
35:36how do you how do you look at
35:39mistakes well I mean which which ones do
35:42you tolerate and which ones don't you
35:47think I don't really think of that way
35:51uh you know the first three flights of
35:53SpaceX failed um the fourth one
35:58and if if the fourth one had not
35:59succeeded we would have gone bankrupt we
36:01would have had no money left so it was a
36:05call um but since then space has done
36:08very well it's now the the Falcon 9 you
36:10know knock on wood is the most reliable
36:12rocket in the world um and launches
36:15about every um two to three days now um
36:19last question on risk what are the types
36:21of risk you would not want to
36:28well I I think in in terms of risks you
36:30don't you you don't want to take risks
36:31that where if if you only want to take
36:33bet the company risks if they're
36:35absolutely necessary so there have been
36:37a few times where saying the T with
36:39Tesla we we just had no choice but to V
36:42the company because if if we're in if
36:44we're doing a new vehicle program that
36:45is uh in order of magnitude larger than
36:49the past one then we're by we're just
36:51unequivocally betting the company
36:52because the new vehicle would be 90% of
36:55production so going from uh the original
36:58Roadster to the model S original
37:01Roadster was only you know about 600 6
37:04700 per year then Model S was 20,000 per
37:08year and um and then model 3 is sort of
37:13half sort of half a million per
37:16year um model y over a million per year
37:20so these are all bet the company
37:23vehicles but the the the reason we could
37:25do for example a cyber truck which was
37:28a a radical new design was because it
37:30wasn't a bet the company decision so I
37:32was like okay look let's try something I
37:34want to try something totally crazy uh
37:36it's like what what truck would Blade
37:38Runner Drive um except the when you're
37:41going to drive on was yeah I think it
37:43would be perfect for Ms um but like we
37:46could try something that where there's
37:48some chance that people might not like
37:50it um but it's it's radical and new and
37:54it's aesthetically aesthetically it's
37:56not derivative it doesn't look anything
37:57else on the road um whereas all the
38:00other sort of pickup trucks look like
38:02vague copies of one another um they we
38:05could afford to take a chance on failure
38:07and say like and talk it up to you know
38:10well we tried you know we try to do
38:12something interesting but but actually
38:13by the way cyber truck's doing great um
38:19uh but one of the things that I think is
38:22important for Innovation is that you do
38:24accept failure like like necessarily
38:28you have to always look at the incentive
38:29structure of an organization and say um
38:33you know is is is that is that
38:35organization properly incenting
38:37Innovation um and in with if you do
38:40Innovation you're necessarily going to
38:41Uncharted Territory so they going to be
38:44some mistakes they're going to be some
38:47failures um and you have you have to
38:49like like actually like for for SpaceX
38:53uh rocket engine development like I keep
38:55telling the team look if we're not
38:56occasionally blowing up an engine on the
38:58test stand we're not trying hard
39:00enough you know um absolutely absolutely
39:03how important are the P how important is
39:05research and phds and that kind of stuff
39:07think I've said seen somewhere you you
39:12useless well I think most PhD thesis are
39:16useless which I think is actually
39:19objectively true if you look at how many
39:23PhD you look at all how many phds are
39:25created every year and how many of those
39:27papers are actually used in anything
39:29yeah um then objectively most PhD TCS
39:33are have very low utility or maybe zero
39:37um because nobody uses them um or so
39:41once in a while you get something that
39:42is spectacular but it's pretty
39:49rare perhaps something more useful um is
39:51in the book that you haven't read uh
39:54talks about your love for uh gaming in
39:56particular like stret IC ability gaming
39:59and I've been thinking quite a lot about
40:00it um what have you learned from from
40:02those games and have have that learning
40:06and wisdom been helpful when you have
40:14I it's hard to say exactly what I've
40:16learned from video games except that I I
40:18do like playing video games as if I want
40:21to take my mind off work I'll typically
40:23play a very hard video game such such as
40:28well over the years it's been many many
40:29different video games um so you know
40:32when I was a little kid I was like you
40:35know pong and little tank games and
40:37things and um and but if you take a game
40:41like for example civilization it's
40:43actually quite a good um it tells you
40:46how how civilizations are formed like I
40:48remember I remember playing the original
40:49civilization with the technology tree
40:52and and how you invent different things
40:54you'd like invent literacy and uh you
40:57know invent democracy and invent gun
41:01gunpowder all all these things that like
41:03and you start to realize oh wow there's
41:05there are stages to technology like you
41:07can't um you know you can't actually get
41:11to democracy without literacy um and um
41:16you know so there's these these stages
41:18of of Technology development or stages
41:20of ideas that uh you know that's that's
41:25a helpful framework for a company um
41:28and I guess in in in like like I say in
41:31recent years there there's a game I
41:34played that was um actually developed in
41:36Sweden called polyopia which is actually
41:39quite a good game um like a lot of
41:41people like playing chess but I think
41:42chess is not a not a great um there's
41:46not a lot of transfer learning from
41:48chess to the real world because in chess
41:50you've got only 64 squares uh it's a
41:53setpiece battle same pieces every time
41:55there are no terrain differences
41:57uh there's no technology tree uh there's
41:59no fog of War um but say a game like
42:02polyopia has all of those things uh
42:05random terrain generation uh you know
42:09the differences in attack and defense
42:11bonuses depending on what type of
42:12terrain um you've got 16 tribes I think
42:15each with different abilities um you've
42:18got a a a technology tree that you can
42:21choose to develop in different ways uh
42:23and you've got of course fog of War um
42:26so that I think is much more much closer
42:28to reality yeah yeah um so I think
42:30politopia I mean I I was I was playing
42:34Diablo uh for a while pretty fun um
42:38Diablo as high levels gets very
42:40complicated they you could call it like
42:42a a spreadsheet with a game attached um
42:45so so that's that's and I briefly got
42:48the the for about a day the world record
42:51in this avire of zir on on a four-person
42:54team of of clearing the the hardest
42:56level um which was you know not bad for
43:0253 basically will be 53
43:05soon um there is still some uh twitch
43:08element to it and um it's hard to beat
43:11kids at games with a twitch
43:13element um but yeah I like uh I find
43:16these games interesting if you can be
43:17fully immersed in a game some last
43:19questions here um as you know we are big
43:21shareholders and uh made a a lot of
43:30o okay I can hear you
43:33here okay good sounds
43:41go sorry I can you I think everyone can
43:47me let's see thumbs up if you can hear
43:50me let's try again okay okay sounds good
43:53sounds good um now um what is the score
43:58now of in terms of the Union in Sweden
44:03bargaining actually I I think uh I think
44:06the storm has passed on that front I
44:07think things are in reasonably good
44:14so uh yeah I think things are
44:19good um yeah overall yeah I feel pretty
44:23good about the future I mean you know
44:25there's going to be bumpy quarters from
44:27you know here and there but I think the
44:29long-term future of Tesla is extremely
44:36um yeah I'm I'm back on just so um yeah
44:39we met with uh we met with your chair
44:41last month so we we have some update but
44:44any any of view on it why are you
44:46why why are you skeptical to
44:52colletive I was playing with a
45:06hello yeah last question for me um sorry
45:09I didn't hear the answer here because I
45:10was out but um we have covered this with
45:13with your chair but just a last question
45:14here what do you want your legacy to
45:20be I I don't I don't mind if uh my
45:25legacy is accurate or inaccurate uh
45:30provided that I I dive feeling that I've
45:34done the right thing for the future of
45:37Consciousness so just trying to trying
45:40to have this SL of Consciousness last as
45:43long as possible and maybe understand
45:45more about the nature of the universe or
45:47simulation or whatever this
45:52so um I have a philos philosophy of
46:01the understand the universe understand
46:03the nature of the universe um or even
46:07ask kind of like that I would say I
46:10would subscribe to the Douglas Adams hus
46:12guys of the Galaxy School of
46:14philosophy that we're trying to
46:16understand what questions to ask about
46:18the answer that is the
46:21universe okay I think that's a good
46:24place to end um for sure the life on
46:26this planet would have been been a lot
46:27more boring without you I'm I'm glad to
46:29[ __ ] it up a little totally all right
46:32well good talking bye take care now all