00:00Precision delivery of medicine
00:02entertainment franchise games absolutely
00:05exploding small modul reactors and the
00:07nuclear Renaissance plus AI moving into
00:11very complex workflows now these were
00:14just a few of the major Tech innovations
00:16that Partners at a16z predicted last
00:19year and our partners are back we just
00:21dropped our list of over 40 plus big
00:24ideas for 2024 a compilation of critical
00:27advancements across all our verticals
00:30from Smart energy grids to Crime
00:31detecting computer vision to
00:33democratizing Miracle drugs like gp1s or
00:36even AI moving from blackbox to clearbox
00:41you can find the full list of 40 plus
00:43Builder worthy Pursuits at a1.com Big
00:47Ideas 2024 or you can click the link in
00:50our description below but on Deck today
00:54you will hear directly from one of our
00:55partners as we dive even more deeply
00:58into their big idea what what's the why
01:00now what opportunities and what
01:02challenges are on the horizon and how
01:04can you get involved let's dive
01:07in as a reminder the content here is for
01:10informational purposes only should not
01:12be taken as legal business tax or
01:14investment advice or be used to evaluate
01:17any investment or security and is not
01:18directed at any investors or potential
01:20investors in any a6z fund please note
01:24that a16z and its Affiliates may also
01:26maintain investments in the companies
01:28discussed in this podcast for more
01:30details including a link to our
01:31investments please see
01:38disclosures hi I'm Alex immerman a
01:41partner on our growth fund here at a16z
01:4620124 is the consumer AI Battleground is
01:49going to move from the model to ux
01:542023 a minute not spent building the
01:57model is a minute wasted
02:00build the best model and the users will
02:03come the most popular consumer AI
02:07companies thus far have been the
02:09producers of their own models such as
02:11chat TBT character Bard and mid Journey
02:15differentiation has come from being the
02:17best model in their domain mid Journey
02:20with images character on entertainment
02:23chat gbt on overall text ux has largely
02:27been determined by the fastest way to
02:29get the models in the hands of
02:32users but thanks to a combination of
02:34factors the likelihood that Che
02:37shortages e is the availability of most
02:39Foundation models via API and
02:42increasingly powerful open-source models
02:45the groundwork is there for breakout
02:48consumer apps to be built on someone
02:51model in 2024 consumer AI apps will
02:55break out by delivering the best user
02:57experience around unique we use cases
03:01Less on model performance
03:04alone I'm particularly excited for
03:07Consumer AI apps that figure out how to
03:09include shared experiences in a
03:12multiplayer mode aggregate multiple
03:15models into a single interface or build
03:18more focused Solutions where workflow
03:22value llms can be a source of
03:26differentiation today they might provide
03:28a first mover Advantage but
03:31old-fashioned modes like Network effects
03:34High switching costs scale and brand are
03:37still will will likely win long term all
03:41right Alex since last year's AI consumer
03:44boom people are constantly asking this
03:47question of where Moes will arise I feel
03:49like every podcast we do on the topic I
03:51have to ask that question really where
03:53is the differentiator and a lot of
03:55people did talk about models but you're
03:57saying that here ux really might be the
03:59moat so why do you think that is and
04:02also what would you say to push back
04:04that maybe it might not be a moat given
04:06that design or ux feels like it can be
04:09copied easier than ever sure uh over the
04:12history of Technology paradigms we often
04:15see technology progress and features
04:18first and then a movement toward
04:21prioritizing design and
04:24experience I expect user experience to
04:27contribute toward the building of Moes
04:30an awesome user experience can raise
04:33switching costs strengthen Network
04:35effects and Empower a
04:37brand so maybe if you think about the
04:40last paradigm shift with mobile uh if we
04:43were to rewind the clocks to
04:4520110 people were taking photos with
04:48their new powerful phones there were
04:51many photo filtering apps I remember
04:54hips dematic there were social media
04:57apps like Facebook but there wasn't a a
05:00great experience that bridged the
05:02two Instagram came along it brought a
05:05slick interface it was simple to take
05:08photos edit share photos 10 years ago I
05:12probably would have called that
05:13experience magical my photos my friends
05:17they're all there and suddenly the
05:20likelihood of me switching to another
05:23application that doesn't have that
05:25information is lower why would I switch
05:29the the experience is so good that my
05:31friends and I we use the application
05:33more which in turn makes it better
05:36strengthening the network
05:38effect in AI we're starting to see
05:41consumers build really exciting
05:43applications on top of this new
05:46technology and I think they will
05:48leverage user experience ux to establish
05:52similar Moes to their
05:54predecessors yeah I think that makes
05:57sense but I think what some other people
05:59might might ask is if everyone has
06:00access to all of the same models as they
06:03do become more accessible how do new
06:05players really win in this
06:08environment it's important to point out
06:10that virtually everyone already has
06:13access to open AI or anthropic but there
06:17are real costs associated with using
06:20these models so unless you're very well
06:22capitalized it's difficult to run a free
06:25consumer app and afford the associated
06:29inference requirements for this reason
06:31many of the current AI consumer apps
06:34actually use a subscription or
06:36transaction model to cover the cost they
06:38are beholden to charging consumers
06:42what's changing now is we expect the
06:44cost to access quality models whether
06:47closed Source or open source to fall
06:51developers will be able to build more
06:53affordably for a broader set of
06:56consumers it widens the top of funnels
06:58for these Developers
07:00and presents a much broader
07:02opportunity one framework to consider is
07:05that models are just a new primitive to
07:08build applications on top of the same
07:11way storages or geolocation or payment
07:14rails are today so if you believe in the
07:17commoditization of models You could
07:20argue that models are similar to a rails
07:24and financial services a fintech company
07:28has to distinguish itself with user
07:30experience and workflows around a just
07:34like consumer apps will need to with
07:37these models yeah so in a way you're
07:39saying a lot of people think that as
07:41this becomes more democratized then the
07:43incumbents win because they have the
07:45financial resources and the existing the
07:48existing users but really you're saying
07:50that we expect those economics to change
07:52maybe another piece of push back from
07:55folks would be how do you avoid as as
07:57folks say just becoming a rapper right
08:00and being displaced by certain model
08:02providers like open AI I think a meme on
08:04the Internet is every time open AI has a
08:06demo day a bunch of startups get wiped
08:08out I don't know how factual that is but
08:11how should Builders really think about
08:13differentiating through model versus
08:15experience a classic investor question
08:18for many years has been won't Amazon do
08:21this now it's quickly becoming won't
08:24open AI do this entrepreneurs should
08:27focus on where they are uniquely best
08:30positioned that's not specific to AI but
08:33in general for company building a recent
08:36historical parallel would be with cloud
08:39computing should entrepreneurs build
08:41their own infrastructure or should they
08:43work with AWS gcp Azure for most
08:47Founders they should work with a third
08:49party if Founders believe that one of
08:52their teams core strengths is their
08:55ability to build a truly differentiated
08:58model that that is a worthy Endeavor but
09:02there are few companies with that
09:04caliber of a research team today it's
09:06true with character AI
09:08idiogram the list is
09:11short for most companies that's not
09:13going to be the case but fortunately
09:17building a defensible business does not
09:20have building a model as a
09:23prerequisite there are other ways to
09:25make it difficult for your users to
09:27never want to leave your application
09:29with regard to rapper companies most
09:33rappers today are just providing a user
09:35interface to call a model
09:38API you really need to add value on top
09:43model my colleague Jamie Sullivan would
09:46say it's the difference between being
09:48one of the hundreds of compass apps in
09:51the App Store and Uber using the
09:54geolocations of a smartphone to build an
09:57application that wasn't possible before
10:00and for which underlying infrastructure
10:02is not a threat in the AI context I'm a
10:05fan of perplexity which is an AI powered
10:08research assistant it does work with
10:11thirdparty models but it also offers a
10:14lot of seemingly small features that add
10:17up to a compelling user experience lower
10:21latency providing sources suggesting
10:24follow-up questions giving multimodal
10:27responses this results in a fun a
10:30different search experience that users
10:32can't find with just using a model yeah
10:36I love that Uber example I feel like it
10:38really cements the idea that a model is
10:41not the necess a model is not the
10:44starting point for many and you can
10:46actually add value on top of that if
10:48folks are convinced by that and they're
10:50saying okay I don't have to create my
10:52own model but I can create something of
10:54value with a differentiated user
10:56experience you mentioned a couple ideas
10:59there are like multiplayer modes
11:00interfaces that aggregate several models
11:03could you maybe elaborate on that like
11:04what are you seeing as an examples of
11:06really creative experiences like the
11:09equivalent of uber with geolocation that
11:12you've seen already and also maybe some
11:14breakout use cases that you'd like to
11:162024 I'm really excited about
11:19this a a first example cor's AI product
11:23po covers both multiplayer modes and
11:27aggregation so on the aggregation front
11:31po brings together leading models like
11:33gbt Claude Bard for text stable
11:37diffusion dolly for image and many
11:40others and so if you want a central
11:43place to access the many models that are
11:45popping up each and every day PO is that
11:48web browser for AI but then I can also
11:52use it with friends so when I ask a
11:54question of gbt 4 on PO I can then share
11:59with you Steph and you can continue the
12:02conversation in po looking ahead these
12:05shared experiences are going to go from
12:07just being text to being multimodal so
12:11imagine Steph you and I were having this
12:13conversation um but we're having it with
12:15an AI as well and the AI is not just
12:19answering and listening to the text but
12:22it's also listening to our speech it's
12:24responding with their own AI voice po
12:27actually recently added voice in
12:29we're going to have more realistic AI
12:33friends this should also translate to
12:36gaming so you and I we're playing
12:38fortnite we're playing Roblox and all of
12:41a sudden the non-player character the
12:44NPC who's listening in watching our gaml
12:48starts having a unique personalized
12:50conversation with with my friends and me
12:54I think that's just going to be a lot of
12:55fun it's G to be so fun and what's also
12:58going to to be fun is we can only dream
13:00up a fraction of these experiences right
13:03and as all these Builders go out and
13:05explore I think we're going to see all
13:07types of things that we never could have
13:09imagined today um one other thing you've
13:12mentioned in your big idea is the
13:14powerful nature of Open Source models
13:16and I just love to get your take on how
13:18key that will be to driving some of this
13:21change how or to what extent do you
13:24really feel like open source may become
13:25the industry Norm underpinning many of
13:29that are soon to come CLM CEO at hugging
13:33face uh a few days ago forecasted that
13:36in 2024 open- Source models are going to
13:39reach the level of the best closed
13:42models that may be aggressive for 2024
13:46but they are headed towards parody
13:49several of the leading open- Source
13:50models like llama and mol are already
13:54competitive with the closed Source
13:57equivalence it's early to call how the
13:59open closed Source will split but the
14:04better open the better the open source
14:06models get the more pressure it will put
14:10on the closed Source models to lower
14:12costs so what I am confident about is
14:15the cost to access models is going to go
14:18down and they're going to become more
14:21accessible and what that's going to
14:23translate to is more activity on top of
14:26Open Source we're already seeing a lot
14:28of really fun applications what I'm
14:31confident about is the cost to access
14:33models going down and they being broadly
14:37accessible there is so much early
14:39activity happening already on top of
14:41Open Source which I mentioned
14:44earlyer can be much more affordable and
14:47allow for free and Broad consumer
14:49adoption two fun image applications are
14:53ever art and can of soup that productize
14:56fine-tuning to a specific Style or
14:59people yeah like we said I think there's
15:01going to be a lot on the horizon and I
15:03feel like open source really
15:05democratizes that and the more entrance
15:08the better as we really think about how
15:10the product layer may be increasingly
15:13delineated from the infrastructure layer
15:16can you maybe think through some second
15:18order effects there and that could be
15:20applying to existing or new companies
15:23that we're referring
15:24to so as the application and
15:27infrastructure layers are
15:29split that'll make it a lot easier for
15:32new companies to be built Founders can
15:35just build for applications and not have
15:37to worry as much about the underlying
15:39model as an input to their own company
15:43creation this is this is akin to the
15:46mobile phone being introduced and making
15:48it so much easier for gaming companies
15:50to handle distribution via the App
15:53Store if there's one less major feat to
15:56building a big company it attracts that
16:01Builders beyond that much like uber and
16:04Snapchat could not have been predicted
16:06when the iPhone came out we don't know
16:09what gen AI will bring but we're excited
16:13it's going to be some fun sometimes
16:16weird hopefully always impactful new
16:21um AI will save the world all right I
16:25hope you enjoyed this big idea we do
16:27have a lot more on the way including a
16:29new age of Maritime exploration that
16:31takes advantage of AI and computer
16:33Vision Plus AI first games that never
16:36end and whether voice firste apps May
16:39finally be having their Moment by the
16:41way if you want to see our full list of
16:4340 plus Big Ideas today you can head on