You.com founder talks the future of AI: Opening Bid

[Music] all right welcome to a new episode of Yahoo finance's opening bid hosted here at the uh really absolutely amazing NASDAQ Market site in Times Square in New York City really where all of the cool Financial happenings go down uh as we say here on opening bid uh let's get smarter and make a lot more money here with us today is Richard soer uh founder of u. com Richard good to good to see you I just saw you at the uh the time AI dinner we'll drop some names and stuff you know that's what we do but good to see you in person thanks for doing this great to be here thanks for having me so look we have 24 minutes I mean tell us how AI will change civilization I mean no pressure right happy to I love Ai and all its facets I can talk about it different time frames and time what is what's capturing most of your time right now uh on the AI front I think the biggest thing is probably agents and how they will change work I think over the next few years and maybe decades almost every company will hit some kind of Innovator dilemma um in that the way they used to make revenue is going to change because a lot of their processes are going to get automated with AI and that just changes the nature of the work that you're doing you need to stop thinking about uh almost any repetitive Pro process every time you feel like oh I've done this before I'm just going through these steps you should be giving those as a prompt to an agent and then get an accurate AI agent there's also a lot of things that aren't very accurate it's easy to build prototypes these days it's hard to really get it into production but once you give it to an accurate AI agent it will just take go through those steps for you and I can give you a bunch of examples yeah so I full disclosure I never heard of Agents until our mutual friend uh Mark benof Salesforce uh co-founder and coo brought them up to me a couple weeks ago and then I heard more about it dreamforce um last week how do you see agents interacting with the workforce so we've had agents at u. com for quite some time now maybe uh 6 to 12 months uh and the idea is that it's not just a large language model that predicts the next token and creates text for you uh but you can think of these LMS as essentially neural sequence models they take any kind of sequence they train on a lot of examples of such a sequence and then they learn about that sequence type and then be are able to generate new kinds of sequences so what does that mean like a sequence could be a sequence of English words and then you can have it write a poem or an email or a marketing campaign for you or a brief as an account executive and sales uh the sequence can be sequences of programming uh languages and instructions so all of a sudden now this will be programming for you this kind of sequence model trained on code uh and that will change all of programming and of course software is eating the world AI is eating software so there's a lot there uh that sequin can also be a sequence of proteins and all of a sudden you have a foundational model for biology proteins govern everything in our body and that will change all of medicine in in the next decades everything is slower and wetwear which is even you know slower than Hardware which is slower than software but uh uh the sequence can be a sequence of uh tones and you have music generation can be a sequence of frames you have video generation a lot of things can be thought of as sequences that these large neural networks can train to understand then why do we need humans I'm I'm listening this I'm like I don't know I mean I'm guess I'm still needed as a host and to like edit are Stu am I really you are cuz people want a human to human interaction right no one likes to see a robot jump off a cliff and do a proximity Wings suit flight you know like you you want to interact with humans you want to have human connection I think there's still a lot there but also this will only automate things that are repetitive I think it's actually going to be exciting for Humanity to do less repetitive things and more Focus much more on interesting uh tasks that require creativity and require you to think outside the box and not just do the same thing over and over again but to come back to the agents now for agents it's essentially a sequence of actions interspersed with often programming or language and other things and so as an AI agent can essentially record what you usually do an email comes in then you click on these five things within either your CRM or your like content management system or whatever it might be and then it learns from those patterns those sequences and includes those actions and so what that means is eventually an AI agent can surf the web for you I think actually in next few years we're probably going to see more and more AI agents surfing the web for you than we have people surfing the web which will change the web as we know it too and how it often monetizes uh but it basically means that a lot of people will get access to their own personal assistant the way only really wealthy people have right now and I think that's exciting that's a good thing I I can dig that what happens then Mark made some pointed comments to me on on Microsoft and and co- piles that but I'm curious on what you think what what happens to the co-pilot model I think the co-pilot model is just a different sort of Microsoft terminology for agents uh I think they're going every software company is going to try to infuse those that AI into all their existing products I'm personally making a bet that with you with you.

com we also have a new kind of software which we call a productivity engine uh it's different than in word in word you can have ai say like okay like finish this paragraph for me or like change the style of this paragraph to be a little more formal or fix my grammar or translate this and whatever you might do that in word but you also want to have a an AI agent for knowledge that will go on the web for you program something for you and gives you give you very accurate an anwers and that is kind of a question like when is this when is AI going to just make an existing software product or general product better versus it's creates this whole new category what do you think the average person should expect in terms of changes to how they interact with the internet so I go to I'll start with Yahoo I go to yahoo. com I type something in it gives me a result I go to Google same thing I but how is that experience going to change over the next few years because of AI I think the publishing and news uh and media industry is going to change quite a bit uh we're actually partnering with a lot of folks we just uh announced uh yesterday uh a big partnership with what with wtak big German publisher in the medical space also a lot of doctors read it a lot of patients read it uh and Healthcare professionals and the reason we're working with them is to help make articles much more interactive imagine you're reading an article and there's a bunch of terminology and you're like I'm not familiar with that terminology just can you explain why are these two people like fighting against each other or like why did this company uh get like acquire this other company how is that related to their other business model or to this industry and so you can make everything much more interactive with these and right now a lot of folks go to you know chat gbt or u. com to get those kinds of answers u.

com they're more accurate and have citations and everything but in the future we want to enable the Publishers themselves to have this Tech technology accessible on their own side and have control sort of in what kinds of articles are going to get cited uh what kinds of knowledge it refers to when it answers your questions because you know you might on Yahoo finance have your own set of Articles right that that sounds like that sounds like AI for good I mean do you share some of the concerns on the other side or do you share some of the concerns that AI won't be used for good this is a really tough question because there's so much hype around it and AI is so complex and not only is sort of are the tide rising for a lot of things and all the boats are rising and there's a lot of goodness but on top of it there's sort of little bubbles that happen that burst right there are some startups that raise hundreds of millions of dollars and their seed round which as an investor means you're kind of combined seed stage risk with late stage returns which generally an expected value not ideal um and there's some hype also in it but it's also clear the technology is working better and better and so there are in some ways of course downsides to almost every technology right you can go as basic as a hammer which you can build to use a you know use to build a house or uh like use it as a weapon right it's really morbid here for the morning no but you're right you are right I mean it's like hammers are great right you can use them for a lot of things but like the internet the internet's amazing and the internet also has some terrible content on it like illegal content things we do not want to see in the world and now you could say well if there's illegal content on the internet let's just make it slower let's make the hard driv smaller so you can store less illegal content let's make the pipes smaller like you have less like transmission speeds for illegal content and so on or you can say let's make content illegal and make the internet overall faster and better and so on I think we see similar things with AI where some people say the number of parameters they want to block those numbers actually currently a bill uh underway in California which is terrible because it tries to regulate the abstract technology instead of the applications I think you know what you're essentially asking for is is how much we should regulate Ai and because you know we want to protect people from downsides and it makes sense to regulate AI in its applications I also don't want any random AI startup to drive by me uh on the highway at crazy speeds because they're just training their AI self-driving cars like Regulators are so far behind like I'm sitting here listening to you and i' like I said I met you before but like it's like you're already like operating 10 years forward if not further like The Regulators are are operating like 10 years behind like how big a problem is that for for society you know it's kind of interesting I'm originally from Germany uh and as a European I see how sort of some countries and and just regions fall in this litigation versus legislation spectrum and Europe tries to often regulate things before anyone gets hurt the US just lets people often not everywhere like FDA right like certain medications and so on you can just sell snake oil but in a lot of places uh in areas of law the US kind of lets thousand flowers bloom and then if someone sues you then they start the case law you litigate and then like law comes out of that and that often means that the US has an easier time with Innovation and actually let like Technologies start to do something useful uh and then if someone gets hurt you you get sued and so I think Europe has actually slowed down quite a bit because they're trying to pre emptive reregulate an AI industry that honestly is very small in Europe in comparison to the US uh and it's kind of getting stifled before it can even flourish and so I think we need to regulate AI in its applications you know medicine surgery again I don't want some AI robot practicing its like First Steps in my brain right in the middle of a surgery so where it makes sense you have to regulate it before it touches any human lives in a very major way but like I hope we don't regulate AI in terms of the number of parameters of the model that just makes no sense to me I would let you operate on my brain on an AI model if anybody if I'm going to let anybody I'm going let you do it Richard I we'll be right back here on opening [Music] bid just going to say we're having a mind-blowing conversation here at the NASDAQ Market site Time Square New York City with Richard soer uh founder of u. com it's very rare rare to hear or talk to someone who is innovating in search like because there's just these dominating forces I mean is someone doing battle with Google I mean how unfair are they being on the internet I mean I'm sure you're watching the case and and if they are broken up I mean but take us inside that that that race with the company yeah so we started as a search engine uh it is indeed a very hard space to compete in uh there's a lot of Monopoly power that Google has uh for instance we have a browser extension you can change your default URL navbar in the top of your browser to default to you. com but when you install that extension uh after a few minutes there's a popup and that popup has a okay button and a white button on white background in the bottom left and if you just say make it go away then boom Google is back to be the default uh in Chrome or Safari to be the like your default search engine uh and so we like it was very hard we lost a lot of people to that sort of dark pattern is what it's called and that's just one of many examples of course the fact that it's very hard on your iPhone at all to change uh your default uh in Safari um was hard now what we've also learned is that a lot of users ask fairly simple questions to a Google what's the score of the game like what's the weather tomorrow what's the stock price how old is Obama and the simpler the question and the shorter the answer the harder is actually to be 10x better and that's where you want to be as a star we want to be 10x better than uh the incumbents uh and so what we found is the people people that actually enjoy getting really accurate more complex answers are to a small degree students and to a very large degree knowledge workers like folks that need to take a big complex marketing campaign and then say all right I'm getting this new feature from my product and engineering team go now compare this to the competition who else has this feature describe it now for these two different industries that I usually have to Market to now write me three LinkedIn messages and an email marketing campaign for it and you just describe how what you do usually when such a new document comes in from the rest of your company and then it goes through all the steps it does all the searches and in many cases our agents will actually replace 10 Google searches for you and hence we're actually 10x better for knowledge workers so we're less and less describing ourselves as a search engine what has to happen I guess to to make the industry more or just make it fairer make that competitive battle fairer I mean because it's so simple to say in media land we break up Google what does that look like and then you know how would it help your business yeah Google is a very complex and large uh machine and Beast uh I think one area that could make sense is to separate search advertisement from having a search index uh the index is very powerful and it's quite hard to build we are building it at u.

com too for a lot of our use cases actually now have a better index for web news for instance and a bunch of other slices of the internet uh that we're also selling uh to customers and they're making millions of calls uh day uh in some cases which is really exciting uh so separating that uh then of course there's a bunch of other stuff like the their Maps um Gmail and so on I think it could make sense in some cases to potentially uh split that up too but I'm not reading expert uh in in massive antitrust things so what keeps you going in this battle like what made you get involved with search so I've Loved Ai and natural language processing since 2003 when I started studying it back in Germany and I think AI is the most interesting thing you could possibly do with uh your time right now you know there are certain times in humanity where certain technologies have their early days right and like 100 odd years ago that uh was in physics and you could like discover the foundational elements of quantum physics and relativity the like relativity Theory and so on uh and uh that time right now is it like you can change humanity and be part of that new Step function productivity you know over 150 years ago more than 90% of people worked in agriculture if you ask those people in that moment oh guys like imagine there's this big machine if you stand in its way you're going to get crushed but generally if you get corn man if you let that machine just go on the field right it will do all this work now only 5% of people work in agriculture that is a massive change to humanity we're not doing more interesting things I think we're seeing that again with AI so I've Loved AI for a long time uh and language is I think the most interesting manifestation of human intelligence written language seems like one of the big requirements to have long-term civilizations uh and so then you go double click into language like what's the most interesting thing is helping people with knowledge helping people learn things be more productive with their knowledge summarize things um one of our our official incorporation name is actually suie for summary search uh before we got a few weeks later you comml uh from from Mark there a fun story on the side there um but uh helping people learn deal with this massive influx of information that we currently have synthesize it make them more productive feels incredibly meaningful if people aren't picking up on this now uh Richard is very plugged into the AI industry and investing in the future of AI every day or it's not every day but it feels like every day there's a new upstart raising Capital at some really eye watering like value what do you make about make uh on some of these Ai valuations and the companies fetching these valuations does it feel frothy does it feel right to given where the technology is and and where it can go it's it's a really tough space to navigate um I've been investing since the exit of my first startup and a lot of companies been very fortunate to have uh various unicorns that invested in uh seed stage rounds uh and now kind of are scaling this up at a ventures my Venture fund and uh what we're seeing often are are early stage rounds that uh have just too high of evaluation for where the risk is and what that means is you often end up with seed stage risk and late stage returns which in expected value is not an ideal setup and so I do think a lot of AI investors are probably going to struggle also uh now it doesn't mean that none of those early stage highly valued companies will work out right some certainly will it's just the risk is not quite uh right uh and so at the same time we do think that pretty much every industry out there is going to change right and will get disrupted and there's some obvious ones like SAS for work you know everything kind of sales service marketing uh but also knowledge work uh analysts um uh biotech researchers we have I just said u. com to give you a sense like we have one of the largest hedge funds uh in the United States as a customer uh we have universities that are customers we have cyber security companies biotech uh Publishers and news media organizations all of them when when we do these workshops with them and we say hey turns out your sales team can also use the software your marketing can also use the software your developers can also use this they're like oh my God like there you see all the light bulbs going off so there is a huge impact on pretty much every industry out there I think physical Industries are going to take a little bit longer to iterate um so your plumber uh is not collecting data about how he changes the pipes right I don't want my plumber knowing how and when I went to the bathroom just be honest I me it's a podcast we can pretty much say whatever we want but like what do you what do you look for like how do you know you're not being sold a bunch of like smoke and mirror uh by an AI company that in some respects maybe you don't know the leaders in your meeting for the first time right so at AI ex Ventures we sort of have a different kind of model from like General Venture funds where we have ai experts like deep practitioners professors at Stanford Berkeley Founders in AI like myself who have a full-time job in AI staying current staying up to date and then we partner with sort of the traditional Venture full-time teams that basically do what you would expect from Venture fund they help you with go to market help you with marketing with hiring and things like that and so I think what that means and what we're looking for are is a combination of companies that are and Founders that are AI native uh as well as have deep industry insights so we're not trying to invest in a bunch of hammers looking for nails uh which is also an a thing in in AI right you'll see Nvidia and a lot of other companies in the tooling space that trying to sit on top of that uh trying like building tools and not all of those early stage startup tools actually are really needed at scale by a lot of people um and so we're looking yeah for deep industry insights uh and there's a lot of exciting stuff happening for instance biotech like all of medicine and Drg development is going to change in the next few years thanks to AI you mentioned uh Nvidia and I can tell you Nvidia the Yao Finance community is obsessed with Nvidia they love Nvidia but is someone like really in the trenches with AI how important is that technology for NVIDIA I mean because the way we look at it like this company is is the best thing since sliced bread but are we right I I think Nvidia is an amazing company they're also the second largest investor in u. com in our series B that we just announced a few weeks ago uh I was extremely impressed uh with with Jensen when I had one one conversation with him he's incredibly well read loves history too I gave him sort of this uh um answer about the Peloponnesian War and he's like oh you mentioning this uh theotis I think uh uh researcher or historian and indeed it was mentioned in the summary and I was like good good thing do you get you got nervous in front of Jensen like what is that like uh he he's he's so friendly it's uh like I I'm not I don't generally get that nervous uh with most people but um it it was very impressive to to hear him giving advice and so on so I do think think uh it's the right time the right place for this this company um it's also clear that they're you know thinking about not just the gpus but moving up the stack and um I'm yeah very excited about Nvidia what type of advice would a Jensen give you I mean is someone who I mean you've been in this industry seemingly forever like what does he what does he tell you I he doesn't tell you just work hard cuz you're working hard like what does he say um there a bunch of things there was a fairly long conversation almost two hours in the end but uh um one of that was really interesting is that you know he said in the beginning he focused a lot on speed of Nvidia but then at some point realized that the best way is to focus on gaming at first and and like really uh like dominate that Niche it's not his exact words but sort of how I interpreted which generally good startup advice and then expand from there and still remember in 2011 12 uh my research group um at uh Stanford Andrew in's uh AI group like uh we were trying to buy more and more gpus and uh at the time Nvidia was like who are you why why are you trying to buy our gpus you're not even the graphics group right which is sort of traditionally where that sold most of the gpus and now of course we have gpus that don't even have a graphics out right they're just for AI workloads I we're out of time Richard I could do this for another like five hours this is fascinating stuff but I can't I have to go uh Richard seers uh founder of u.