No Priors Ep. 14 | With Sarah Guo and Elad Gil

Key Takeaways

[Music] foreign welcome to no priors today we're going to switch things up a bit and just hang out and answer listener questions about tech and AI the topics people want us to talk about include everything from the evolution of Open Source models to the balkanization of AI Elon AI which I think will be super interesting to cover regulating Ai and AI hype in the investing environment let's start with the march of progress for open source models um I guess Sarah what have you been paying attention to what are some of the more interesting things that you view happening right now yeah so there's nothing out there today in open source that is like GPT 435 or anthropic clog quality right so there is a there's one player out in front and that's open AI but I think the landscape has changed a lot over the last couple months like Facebook llama is quite good um many starts are just using it despite its licensing issues assuming Mark won't come after them and then you have a number of other releases that have happened right uh tomorrow just released a pre-training data set which seems quite good stability just refused um released diffusion excel in the imagen space um and so I think the like larger Dynamic is that there's been an increasing number of people and teams that now know how to train large models the cost of a flop is only going to go down there's a lot of investment in like distilling models and a lot of researchers would claim that you and I know that it's going to be 5x cheaper to train the same size model the second time around like once you've made your mistakes and know what you're doing and then you have these other accelerants like you can use these models to annotate your data sets and increasingly do Advanced like self-supervision so if VCS are going to continue to fund Foundation model efforts including open source um Foundation model efforts like if I were a betting woman and I am I'd bet there's a three five level model in the open source ecosystem within a year um and I I didn't personally believe that would be true like a few months ago I guess that puts it about two to three years behind when GPT 3.5 came out though and so do you think that's going to be the ongoing Trend that there'll be a handful of companies that are ahead of Open Source by you know one or two generations yeah I think that's like the status quo so if we just straight line project I imagine that will continue to happen and the real question is like can you stay in the lead if you are open Ai and um and like get uh get paid for that or is that the is that the objective of the organization anyway like I I think you know if you feel a um a great leader and a lot of resources and a lot of really talented people that's not something I want to bet against is there anything you think um is coming in terms of other big shifts in the model World either on the open source side or more generally yeah I mean we we should also talk about just like stuff that you are um interested in investing in and generally paying attention to but I think the the big idea that's been very popular over the last few weeks are autonomous agents right and I don't think that that's like a I want to hear what you think about this too I don't think that's necessarily an architectural change but for our listeners the basic idea is to orchestrate LMS in this like iterative Loop towards some high level goal where they're doing planning and if memory and prioritization and reflection and so you're you're not necessarily changing the architecture of the LM itself but this orchestration allows you to like do many new things possibly the classic example being like make money on the internet for me and there's a good number of hackers trying to figure out how to make agents that for example um like analyze demand find a supplier set up a drop ship um Shopify store generate ads then promote that store on social right the whole Loop being like one call to an agent with this high level goal of make money on the internet for me like do you think this stuff is interesting around autonomous agents I think it's I think it's um super interesting and I think for there's an old saying that the future is here is just not equally distributed and I feel like that's one of those things that people in the AI Community have been talking about for a while and there have been very clear ways to do it and then I think there's one or two people that went and implemented interesting things there in terms of Auto GPT or other things and then everybody's like oh my gosh this can this can happen and I think a lot of people in the community are like this is really cool but at the same time of course it can happen you know because effectively you have um some form of of context as a AI agent is acting and then you use that context to inform the next motion and sort of update you know uh The Prompt or what the model is going to do I think there's other forms of memory that people have been talking about that are super interesting like how do you make that a bit more of a cohesive part of how an llm or AI agent functions because right now effectively every time you start a new instance of chat GPT a new chat you've lost the context on all the other sessions you've had and so a lot of what people are thinking about is how do I create ongoing context so that whatever chatbot or whatever API I'm using remembers everything else I've done with it over time or perhaps everything it's done with every other user over time and then that becomes really powerful because you're effectively crowdsourcing and understanding of the world and then integrating it into an AI system and agent and so suddenly you have Global context like imagine if you as a person understood the life of every other person who's lived and then you had all the context around what that means in terms of just how you operate in the world right and so I think I wouldn't operate anymore a lot we'd be hive mind yeah exactly it's just the hive mindset I think I think that's we're all heading so you heard it here first okay okay fine while we're on this topic of directionally AGI there's been a lot of call for regulation of AI from you know Sam Altman to Satya to Elon Musk do you think AI should be regulated you know I think the first question is why do people want to regulate to begin with and I think there's you know two or three reasons um one is if you're an incumbent it actually really benefits you to get lock-in and one of the best ways you can get lock-in is in Industries to have Regulators get involved because they start blocking Innovation and creativity and new efforts and if you basically there's that famous chart of um where price prices have gone up by industry and where they've gone down and they've largely gone down in areas that have been unregulated traditionally that's things like software or certain types of food products or other things and then there's areas where prices have gone up dramatically and that's education it's Healthcare it's housing it's the most regulated Industries so regulation tends to lock in incumbents it means you have fewer people making drugs and you have fewer people doing all sorts of things that could actually be quite useful so that's one thing the second is um I think some people are just scared and in some cases you could say well there's there's reasons to be scared right like what if the AI is used to unleash a virus or what if the AI is used to cause War and if you look at the history of the 20th century humans have done that pretty well on their own already right it's not it's not a New Concept that bad things will happen and often they're driven by other people versus technology right and of course technology can have accidents or can be misused but fundamentally usually people have driven a lot of the really bad things that have happened over time and there's a really long history of doomers who are wrong right and I should say by the way an AI I'm a short-term Optimist long-term Doomer right I actually think eventually there may be an existential threat from AI but I think in the next 10 years um you know everything will be okay uh there may be accidents or maybe terrible things that happen but fundamentally being different from any other period uh but if you look at the dumerism in the past it's things like you know public intellectuals uh worried about swine flu and nothing happened or pop uh you know a lot of people were reading the 70s about population collapse we're gonna have too many people in the world will starve and we're gonna have Global famine and that didn't happen and so uh we have a lot of examples of people in the past kind of predicting Doom when nothing happened and we also had that during covid where a lot of people said kovid's the worst thing that ever happened to the world and then they would be hosting dinner parties unmasked inside with large groups you know later that evening and so I just think you kind of have to look at people's actions versus their words um and fundamentally you know my view would be let's not regulate right now at least most things I think the things that maybe should be regulated are things related to export controls so there may be Advanced chip technology that we don't want to get out of the country and we already have those sort of export controls on other capabilities uh we may want to limit the use of AI for certain defense applications you know do we really want a really smart hyper-intelligent AI agent driving swarms of offensive drones or weaponry um and so there may be some need to do some sort of global regulation for things like that or at least you know something like what we've done for chemical weapons so like and then I think in the long run we may want to think twice about Advanced Robotics and their implications as AI becomes more of an existential threat to humanity but overall like if I had to choose it right now I'd say don't regulate in the short run except for those areas that I mentioned um and then I I think that um the big pivot point for regulation may actually come during the 2024 election because I think that's the moment that um people will show examples of AI being used to influence the election or influence voting behavior just like ads influence voting behavior right but hey I could write better at copy or do other things I worry a bit about that becoming the reason that people claim that they should not regulate things just like they got really aggressive about social networking so I don't know what do you think uh I I largely agree with that I feel like it's worth like describing what I think are the two more rational cases by the way I I don't think I I think it's too early to regulate I just want to make sure that's very clear but I think the two rational cases I've heard because I keep asking smart people um that I don't think are taking cynical actions why they're afraid or um you know short-term afraid or why they think this makes sense and the two things I've heard are one you know this is unlike the past because of the speed of progression like this hard takeoff idea of like especially if you and you know I I see you nodding and smiling but when when very very smart people who are working at the state of the art tell me that they're um concerned within a 10-year band for Humanity because of the ability of this current generation of models to be used to train the next generation of models and we're all very bad at thinking about compounding I'm like okay it's not a like completely unreasonable point of view I think the other is more of a um it's more of a tactical thing for the industry which is as you said for um you know whether it be the election or some other trigger like there's a version of the reaction to this from um you know people who are afraid or from uh you know political opportunists to um go in two directions one is like Mass surveillance right or one is like complete lockdown right so I think the Tactical thing is to like try to create a democratic process that gets ahead of it with something um something that's a reasonable path forward but largely I'm just uh I I feel as far as is very early to be um figuring that out and then you also have the problem of like uh if you're talking about the more existential risks or sort of the AGI risk like alignment research is very tied to capability research right and so it's sort of impossible to be like we're going to stop making any progress on research but figure out how to control this stuff yeah absolutely and I think related to that um you know I think it's really important to your point to separate out almost what I consider technology risk from species risk right technology risk is there are some bad things that can happen due to technology being abused right and that could be a nuclear disaster or that could be an AI being used to shut down a pipeline or to crash a flight or to do something really bad and those sorts of things already happen but you know you could imagine it could accelerate it in that case you could literally turn off a bunch of servers right you could turn off every machine on the planet if you really needed to and Humanity would keep going and it'd be a reset but we'd reset fine separate from that there's species level risk like is there an existential threat to humanity and that's like an asteroid hits the planet and kills everybody and if I think a lot of the the people who talk about these things mix those two things and I think the true dumer view is while AGI eventually becomes a species and we compete with that and then it wipes out all humans and in order for an AI to rationally want to kill everybody um you'd need some replacement for the physical world because eventually all the hard drives would burn out and the AI would die right if it existed as a species or a life form so you need physical form for the AI in order for it to truly be an existential threat and that's why if I were to focus on an area I'd probably be robotics or something like that because that's where you suddenly give physical form to something and if you're like I was integrative AI can I build my house and I can now you know build a Data Center and now build a solar farm and you're eventually now build a factory you've basically created an external system that no longer needs people and that's when I think there's real risk and that's why on the 10-year time Horizon I'm not that worried because Robotics and atoms in the real world takes a lot of time so even if you have this hyper intelligent thing running the reality is if you really needed to you could turn off every server on the planet yeah I agree with the embodiment being like a key piece in this theory of the AIS are going to kill us and we're pretty far away from that okay so one question we got from listeners and then I'm sure you get all the time is there's a ton of hype in the AI investing in startup world right now what do you what do you think of it is it Justified is it appropriate yeah I think we've both lived through a couple different hype Cycles now right there was hype Cycles around social and mobile and then the cloud and then you know multiple crypto hype Cycles and the reality is out of all those hype waves interesting things emerged right and maybe in the standard hype cycle 95 or 99 of things fail but they're still like the one percent that work or maybe it's five percent work and one percent end up being spectacular and I think the hard part usually is to know what's actually going to work because so many things seem so overlapping and similar and so I remember when the mobile wave happened or mobile and social at the same time a bunch of different people I know started mobile photo apps and each one of those things took off and so you'd suddenly see something go from zero to a million users in a week just literally it just spread virally and none of them stuck right they all burnt out at cycles and the only one that really stuck was Instagram and in part that's because Instagram emphasized filters which things like Camera Plus already had and then in PARTA emphasized the network it's like let's have a follow model like Twitter and that's the thing that really worked and so it feels to me like if you'd um gotten excited about the overall cycle you were right but if you got involved with the wrong set of photo apps so you built the wrong thing then you were wrong in some sense I guess you were right about the trend but wrong about the specific substantiation of it and it seems like the same thing here and so I think often it's that question of you know Peter Thiel has a good saying which is you wanna you don't want to be the first to Market you want to be the last standing and so I think it's a similar thing here how do you end up being the last person standing in that or last company and it may be the same thing as being the first mover right it's Amazon and books or things like that um but sometimes it means you actually do something a bit smarter and you come later in the cycle and it's fine so I mean Are there specific areas you're most excited about in this wave or cycle or opportunities that you think are these things that are obviously going to happen are important to happen absolutely I mean um well and I want your ideas too some of them are shared ideas to be fair um but I would agree it actually added data point I was just over at open AI yesterday and they're biased perhaps in a way that I'm also definitely biased but a friend was saying they actually think that investors are being somewhat wary at the application Level right now because they can't figure out what's going to be standing right it's a very different competitive Dynamic but the market is Extreme for researcher-led foundation model companies because everybody is pretty sure open AI is going to be around right um and I agree the applications are going to be non-obvious but as one example like any investor that claims they knew image generation from text was a killer use case like a year or two ago besides you is just empirically wrong given David's completely investor-free cap table an amazing business right so David in case you're listening I still love my journey and want to invest um but that's why this podcast exists but um you know in terms of like specific things that I'm interested in now I'd say like I think there are a lot of things on the application side that are exciting so to start with some of those I think you know voiced into this and dubbing are going to be just a huge unlock for like content providers and Publishers like I'd like to back something in that space um I was just talking to some people at um a very large financial and they said the like biggest potential cost savings you know on order of tens of millions of dollars a year for us is in turning every line of code we have into explanations for a regulator and that's at once like pretty specific to them but also not right I think the areas of audit tax compliance accounting reconciliation like there's a lot of um natural language understanding that could be um better served by semantic understanding um and and so I think that's an obvious area um I think annotation is changing again right and we can use this is like a very specific idea but we can use llms too much more here um we talked about agents and then um this is like this isn't a necessarily a specific company idea but I think architecturally um retrieval is a field of active research but the idea of personalizing LMS with Enterprise data is an important but like very tricky one you have to do data management you have issues in scalability sync Access Control you likely want to apply traditional IR if you own both retrieval and the model you can do very magical things and so I think like the the chat GPT retrieval plugin is super cool but it doesn't just serve a whole host of use cases and I think this entire like half of the stack is still missing so those are like a couple of the things that we're um sort of explicitly like hunting around but um what are you paying attention to yeah I mean I think we have a lot of overlap as you know so I'm super interested in sort of voice synthesis stubbing and related uh both in terms of infrastructure but then also in terms of application areas and so I think that's going to be a really big sea change that perhaps people aren't paying enough attention to um I'm actually quite uh long on compliance in general like I've done a bunch of things like agent sync and Medallion and other compliance related companies in sort of the old world and so I think that's just an area that there's always going to be you know converting spreadsheets and offline processes and you know random checks and docs into code is you know really powerful um I think there's a lot to do on the upside actually maybe on the other side of people who think that it's impossible to tell what's good and you know nothing's defensible and everything's just a wrapper on GPT or whatever and I actually think there's tons and tons to do there I mean Harvey AI which we're both involved with I think it's a great example on the legal side but I think there's a there's you know two dozen things like that to build over time and it probably takes five years for all those things to get discovered and built and substantiated so I don't think it's like this year there'll be 12 of them but I think like every year there'll be a couple of really interesting ones and then there's probably a lot to do on the tooling side right obviously link chain is sort of a hot one in the in the area but there's everything from you know people exploring Vector DBS like chroma On Through To other forms of um infrastructure uh and so I you know llama index and other things so I just think there's a lot that um to be done at every level of the stock it would be interesting to ask like what happens on the foundation model side because to some extent the question is if we locked in a few of the leaders or is more to come and I think the Elon must start up this rumor to exist as sort of an interesting um example of a new entrant uh and back to regulation you know musk was asking for a six-month moratorium on progress which seems to be very self-serving if you're simultaneously starting an llm company you know up until I catch up right yeah and if I was in that position I'd do the same thing don't get me wrong you know so it's not it's not meant as a dis it's just meant as a you know remember people's incentives um but I do think there may be some interesting things to do on the foundation side and I do think some people are doing that in a vertical specific way they're saying hey we're going to build a healthcare specific model and we're going to build a you know uh Bloomberg did like their their Bloomberg GPT or whatever it was called on the financial side and so I think um you can clearly see these verticals uh um emerge and a lot of people obviously are debating what will general purpose models discover all those use cases are you going to have bespoke sort of vertical models and what what parts of the actual logic and synthesis and sort of magic of these AI models comes from the fact that you've trained on a massive amount of data and language and then you're applying it to a specific area with potentially unique data sets overlaid or is it something that's just you know that can be dealt with vertically specific and you don't need that broad-based understanding of the world so I think that's a really interesting area of like exploration and it's I have no idea what to predict there I don't know if you have any thoughts on that uh well I I would agree I think it's um I think there is real opportunity for vertical specific models where you can imagine that control for either a compliance or a safety or a um performance like reliability of input data makes sense right as well as like if there are architectural differences because um for example you have multimodal data in healthcare and Pharma right if you are looking at protein structures and radiology and Healthcare records it's not clear that um you would want to do that train that in exactly the same way as a general web text model right so I think that makes sense on the um the broader Foundation model question I I'm you know we were talking about open source at the beginning I think that open AI will continue to be a leader anthropic it's very dangerous here like really talented team um but the number of people who know how to train large models and the cost of a flop goes down right and so I think there's like just a lot of incentive in the ecosystem for additional players to um to compete Pete um what do you what do you think is the opportunity for uh incumbents or how do you how should they how should they react to all this yeah I think um you know obviously with every Technology Way there's a differential split in terms of where market cap Revenue employees Innovation Etc goes in terms of incumbents versus startups and you know every wave is a little bit different right the internet wave was uh almost you know it's probably 80 startups in terms of value and 20 incumbents and then mobile was sort of the other way around it was eighty percent incumbents and 20 startups right the big platforms for mobile were the were Google and Apple um but then you had a lot of interesting apps like Instagram and Uber and others emerge um for crypto it was like 100 startup value right and it feels like in this Wave It's probably 80 20 again right it's probably Google will probably become a player right open AI is closely aligned with Microsoft and then you know uh Salesforce with AI is probably Salesforce right it probably is a new company it might be right I actually think certain companies are vulnerable for the first time because these capabilities and that includes everything from Erp providers where there's like a defensive mode through Integrations and obviously this could make integrating your data into multiple things really easy and fast instead of six months to roll out sap maybe you could have a next-gen approach where it takes you know a day or two on a new product right to do all the Integrations that you would have spent six months on Consulting fees for um and so there may be certain types of uh thing uh companies that are vulnerable but the reality is I think in most cases you know if an incumbent is already doing something and they're quick to integrate it then it works great the one area that may be really interesting is almost like there's there's probably room for a new private Equity approach where if you think about high private Equity companies but on things they basically look at cash flows and costs and all the rest of it and if you can radically decrease cost for people heavy businesses by using llms as like a replacement for certain types of work or at least in augmentation then you can differentially bid on companies as a private Equity shop and so I think like people who do buyouts could have this as a strategy I don't know that any of them will because most of them tend not to be very technology Savvy but I think there's really interesting alternative things to do at scale there that that tend to be kind of under discussed the healthcare side that you mentioned earlier I think is kind of fascinating because if you look at the cost of developing a drug for example say it's a billion or two billion dollars to develop a drug whatever it is most of the early stage of development is on the tens of millions of dollars at most and so I think a lot of the default focus of people who don't understand Healthcare very well is to say I want to use this for drug development and it may help with certain aspects of drug development later but usually I think the places in healthcare where this will really get applied fast is on the more operational or Services intensive related side it's Healthcare delivery it's lowering the cost of a doctor visit or telemedicine it's making payments easier and more streamlined if you're dealing with insurance reimbursement and so I think there's really exciting things to be done there like color a company I co-founded is for example thinking about different application areas and I just think that that's like a real wealth of fruitful areas for people to explore if if their Healthcare is heavy and of course with Healthcare the technology you're using the issue usually have to go to market is a hard thing right so I think Market access is really hard there yeah um I pushed back on that a little bit I started with saying I agree on just the um operational uh friction in healthcare that we can take down right there's so many processes like if you look at prior authorization it's a battle on two sides to fill forms and like uh compare like EHR data and clinical recommendations against a policy right and so there's a piece of that you can't get rid of because you know insurance company has incentive not to pay and like hopefully providers trying to provide the best care but there is a piece you can get rid of right like we have models that can you know read data try to understand it fill out form and so I think that there are lots of interesting applications there the the minor pushback and you know much more about Healthcare and Pharma than I ever will but you know VC is the job of having opinions anyway and um I think if this wave of AI can change the cost curve in drug development it's because it you know you're not actually impacting the 10 or 20 million dollars up front on um what's traditionally considered like research you're increasing the probability that you're right right and so like all of the cost of you know expensive recruiting and clinical trials it is more efficient because you're right more often you'll just understand yeah I think I think the hard part is that a lot of um a lot of drug development ends up being hey this works great and mice and let's try it in people now and to your point there may be things that you can learn heuristically in terms of when do things translate versus not but I think one piece of it is just basic biological differences and then the second piece of it is this is back to the point on regulatory capture to some extent the incumbents have an incentive to drive up the cost of drug development so no new startups can actually ever enter in terms of actually making oh that's very cynical yeah oh yeah but you know it's interesting like it's it really is this weird regulatory capture and so if you look at um the last time a biotech company outside of moderna which I think is you know an exception because of covid the last time a biotech company hit I don't know it is 30 40 50 billion dollars in market cap something like that the last year such a thing was founded was in the late 80s so it's been uh at this point what is that 35 40 years without a new major biotech company started in terms of biopharma actually developing Drugs That's shocking right in Tech during that time that same time period there's dozens of companies and if you actually look at the aggregate market cap of the entire biopharma industry and as a reminder Healthcare is 20 of GDP and farm has about 20 of that right or ten percent of that um if you add up the top four or five tech companies their market cap equals the entire industry for biopharma and that includes Pfizer and Eli Lilly and Genentech and Amgen and all these companies as well as all the small startups and all the mid cap companies and everything else and so then you ask why is that and these are very profitable companies right there they have software like margins in some cases and so as you start digging into the district you realize wow like there's there's um strong reasons for incumbency to remain as incumbents and there is this um regulatory process that really delays things quite a bit in some cases rightfully in some cases wrongfully and if you look for example at the covet era we were able to develop multiple vaccines and do clinical trials on multiple drugs really really fast part of that was we had a lot of patients but part of that was we removed all the regulatory constraints and we didn't have mass scale Adverse Events and bad things happening to people we just moved really fast this actually also happened during World War II Winston Churchill wanted a way to treat soldiers in the field for gonorrhea and so they rediscovered and developed penicillin in nine months they again removed all the regulatory constraints and boom nine months later they had a drug that worked really well that was safe and so I think it's something to really think about deeply in terms of what are the incentives that we're driving against and how are we thinking about cost benefits societally but also the second you start adding a lot of Regulation things slow way down and Innovation goes way down and Costco way up and that's the reason that you know for the earlier conversation I think regulation of AI for most things you know export controls make sense if other things make sense but for most things it's probably a really bad idea right now I would agree with that um I do think that there is that was my rant by the way no no uh stay on the soapbox um uh learn something about gonorrhea today but um but I I think like you know if you think about the power of government and I'm strongly on the like reduced regulation encourage Innovation side um you also have these wartime examples of production of airplanes in World War II going from a few hundred planes to 6 000 in also less than a year right and that you're here we're fighting like atoms not bits right you have to build plants and to figure out all these engineering processes and so you know I think that there are ways in which like from a um industrial policy National Security perspective like if we wanted to be winning in AI in a really durable way like I think the paths are pretty clear actually like people need compute and like we have to make it a um a priority in the United States but I I would also say in the field of Pharma I remember like asking you like I don't know seven eight years ago like hey a lot like I know you're interested in aging and like weight loss and the intersection of areas where like um where the demand is very consumer driven right you might break out of um and demand and also like uh the ability to access um different solutions that are on the edge of like consumer purchase right especially as we have more like web um diagnosed doctor Network diagnosed prescriptions right do you think this is interesting and I'd send you a company or two and you uh gave me these same extremely consistent view which was like Hey despite the phds like you know the data-driven person investor inside me says like don't do this just do tech companies so no change um you know I think that the healthcare services and operations side is super interesting right now due to llms and so you know that's an area where I think there's lots and lots of room to do interesting things and I have invested in some software related companies in the past like benjling or medallion um in these areas but I think it's really about what's the healthcare infrastructure that can be served through software and then how can llms accelerate it I think drug development can be extremely useful societally and really important and impactful and obviously there can be really great outcomes for people as well as financially it could be a really great thing but it just comes back to like why hasn't anybody built a generational company in a really long time in the area and there's all sorts of reasons behind that I mean we tried that when I co-founded color right the whole Focus was trying to make Healthcare more accessible to people and I still really believe in that mission so it's more just you know what are the obstacles to getting there for different types of companies and do you want to take on those opticals and and if nobody takes them on the society really suffers and so it's almost like how can you make sure that you remove as many obstacles as possible while still safeguarding the public right so that people don't get hurt by this stuff but at the same time perhaps these things have gotten too extreme and that really you know strangles the ability for the industry to innovate in ways that it could otherwise so it's a really interesting area are there any other topics that we should cover from the audience uh I'm good what do you think a lot I think we got it all thanks to everyone who submitted their questions foreign