A conversation with Sam Altman, CEO of @OpenAI

foreign good afternoon everyone and of course we've got 2 000 people registered online as well for this event so thank you so much for all turning up today I've got a pretty exciting session ahead with Sam and before we bring begin the program as I say I'm Sarah Cocker I'm your MC today and I'm just going to run through some basic housekeeping just to keep those questions going during this one-hour session we've got loads of people on the ground with microphones because it's such a well-attended event so if you have any question please raise your hand if you could just briefly give your name an organization and try and keep those questions concise and of course that will help Sam orientate his questions towards what you'd like to ask today we want to give everyone the maximum opportunity to ask a question so please only ask one question at a time and then return the microphone to the team members on the ground we promise those to allow you to ask us many questions as possible which is what is the point of today and we've got one whole hour to do so finally if you could please abide by the rules and guidelines of the team on the ground of course they're for your safety and for the safety of everyone else as well well it now gives me great pleasure to invite to the stage Deputy CEO of Hub 71 Ahmed Ali alwan for his welcome address thank you [Music] good afternoon everyone my name is the ptco of Hub 71 your excellencies as team partners ladies and gentlemen thank you so much for taking time to attend this session today within 36 hours of Us opening up this event got completely full from a physical space perspective the amount of attention we got and the build up that has been phenomenal so I am ever grateful for you being close to obviously open AI to HUB 71 and making time to attend today and it gives me great pleasure to welcome you here at this session and I would like to specially thank open Ai adgm g42 and Mohamed bin Zaid University of artificial intelligence for making this event happen today at HUB 71 we have created an enabling in our environment that supports the growth journey of startups from ideation up to scale-up and maturity and through that Journey we've been able to build a network of partners that have supported us and carried our mission throughout not only the UAE but beyond positioning us to capture an opportunity like the one we have today so before we welcome on stage our special guest Mr Sam Altman the CEO of openai I'd like to welcome Dr Andrew Jackson who leads g42's AI research program in case you didn't know [Applause] thank you good afternoon everybody I'm Andrew Jackson I'm leading the Inception Institute for artificial intelligence and I'm here today to represent the Abu Dhabi and UAE AI ecosystem let me start off by characterizing the ecosystem as unified emerging and also very ambitious we have a leadership mandate from a leadership who profoundly believes that AI will transform the world we have unified our different people who are working on AI from mbzo Ai tii and IAI it's part of group 42. we're currently working to unify all the data sets of UAE and leverage them in new artificial intelligence models we're also trying to digitize the world of Arabic content we are contributing to open source this is seen in vicuna and falcon and soon more models will follow we are a political Powerhouse and will be Central to AI regulation globally we've already seen our participation in cop and moving towards an energy transformation and we believe the same will happen for AI and finally we're collaborators we work with everybody and especially in in this new race for AI we hope to collaborate with as many people as we can and contribute back to the world in terms of achieving AI for Humanity with this thank you very much guys and I really want to welcome Sam to the stage [Applause] [Music] it's great to be here uh I have not done one of these without uh like an interviewer before so this will be fun the the energy in this place for AI is uh unlike anywhere else I've quite seen in the world and it's been like that for a while which has been special to see um so we're excited to figure out how we can collaborate and I would love to answer questions I don't know I don't know how we're going to do this uh here we have one question in the middle so if you could bring the microphone here please and if you could say your name and organization briefly one question only um hello uh my name is Majid Alma Chaney I work for do Telecom I'm an account manager so I deal with multiple clients here in the UAE I have a use case from a client from today around two hours ago where they want to deploy chat gbt within their um within within their communication process between users internally and externally so through WhatsApp or through a chatbot but the what had come out was the focus on security yeah and how can we or how can the client secure the information that's being sent across so we're trying to figure out exactly how to get this right um we're super open to feedback on what people want for it we we don't train on any data that's ever submitted to the API but it does go in transit and people are still worried about that and then there's like data residency laws in some countries so we're taking this very seriously the Enterprise adoption has just been way faster than we expected but we're open to specific suggestions of what you'd like if you want to email me after and we want to make something that enterprises feel totally comfortable with even when the data is going in transit so I think we have another question here on the front row please uh Hi Sam happy to have you here in Abu Dhabi it's a great honor for all of us my name is Elias eliquiti I work for mobile Investment Company I've been looking into your um debates in the US Congress last month you've been questioned alongside another three experts in AI on so many ideas and future and challenges and so on and you stressed on the fact that opportunities of AI by far larger than risks yeah and that's why the Congress and I I would imagine most of the Regulators will be pushing to regulate this industry because it's new in your opinion with regulations or regulating this industry will close the gap between opportunities and risks or in view on your view how can we capture those opportunities in much larger scale it's a great question I think the reason the world has such a hard time thinking about this technology on one hand we have what most people would agree is the greatest Economic Opportunity the world has ever seen coming at us in real time and it's going to come fast and on the other as with any really powerful technology we Face serious risk we Face existential risk and the challenge that the world has is how we're going to manage those risks and make sure we still get to enjoy the tremendous benefits so everybody wants you know to solve all diseases everybody wants to provide fantastic Health Care everybody wants great education productivity gains discovery of new science all of the stuff that's going to happen and no one wants to destroy the world and no one wants to do things not even that bad but still bad so that's what we've got to focus on I totally believe it is possible to not stifle Innovation and to address the big risks I think it would be a mistake to go regulate the current models of today I think we need the thriving open source Community I think we need companies like ours and others pushing on models but it's important to remember that we're on an exponential curve the number one thing about this technology that I think people don't understand is that in a few years gpt4 is going to look like you know a little toy that was not that impressive we have I think we have two Miracles we have an algorithm that learns really truly learns and it gets predictably better with scale which means we're on an exponential and it means we're not about to s-curve it's not about to top off maybe somehow we're wrong but I think it's just going to keep going and that's hard for any of us to wrap our heads around humans evolutionary biology did not make us good at Intuit and exponential curves for whatever reason and now we're on one and so we have to simultaneously say on the stuff that's not very dangerous let's make sure to get the benefits but as we approach the stuff that can be very dangerous let's make sure we come together as a globe and I hope this place can play a real role in this let's make sure we come together as a globe and put some guard rails on this we talk about the iaea as a model um where the world has said okay very dangerous technology let's all put some guard rails and I think we can do both I think in this case it's a nuanced message because it's saying it's not that dangerous today but it can get dangerous fast um but we can thread that needle next question hi my name Isis I'm Syrian founder of air Hub and we organize hackathons for artificial intelligence and I really wanted to ask a question and I even asked child GPC which question should I ask you so the question was uh what is the future of artificial intelligence how do you see it but me personally I'm super interested on how child Deputy will just trap different Industries and where do you see the biggest progress thank you so in terms of what the future is going to look like there's all the obvious things we can talk about you know gpt5 will be smarter than gpt4 gpt6 will be smarter than that we'll go get every modality working so there will be images audio video text computer programming all together and then biology and stuff like that and I think we can all assume that we can all say all right you know it's just gonna get smarter and smarter every year the thing that I think is hardest to wrap our heads around or for me at least is the phase after that and then what we call like the next Paradigm when the AI can't just learn more and be smarter but when it can develop new knowledge so if you think about Newton for an example there was a time where what he was doing was reading every math textbook he could find every paper talking to every friend every smart Professor I think and then at some point it was time to go beyond that and reading more books was not going to help him and right now the way we go from GPT two to three to four to five to six whatever is to do the equivalent of read more books at some point you have to start generating new ideas you know if calculus has not been discovered yet you're not going to figure out calculus by reading more books you got to just start thinking and we need the models to learn how to start thinking uh to figure out how to come up with those new ideas and to contribute you know one model here one model here just like humans do contribute new knowledge to society and build on each other and that's what's going to come at some point and that I think is when the world can just get dramatically better faster not only do we make cognitive labor available cheaply to the world which will raise them if which we're doing now but we begin to invent new knowledge and I think the only the only sustainable way to make the world better to increase quality of life for everybody is with new science new technology so that's what's coming next and then in terms of disrupting Industries like most of them uh I think if ai's going to touch a lot of things there's some that it won't there's a lot of things where we we really want humans to do things we really care about what humans do um but if you just think about automating most knowledge work that's powerful and it's not powerful because it just takes all of our jobs that would be boring um it's powerful because the floor do we're just going to get your hands up like sorry it's cussing out there uh it raises the floor and what all of us can do and also the expectations so as you know time goes on when we have this powerful technology we can all accomplish more we can be 10 times than 100 times more productive which means we can do fundamentally new things it's not it's not just a quantitative shift as Sam I just wanted to go online because we've got 2 000 people registered for this event and Muhammad sinan from let's passner is asking me which I think is really relevant because of course a lot of this comes down to cash he's saying how will large language models impact small-scale researchers in universities who don't have that much funding and then to your previous point will all the software software Engineers lose our jobs so you know our goal is to make the smartest best most capable model and make it available as cheaply as possible um like that is how we think we can contribute to the world researchers do all sorts of amazing work on top of that and it contributes into that but if you're trying to just like compete directly with us to scale up llms like probably we will win the race um but there's so much else to do uh don't do the thing that everybody else is doing go pick out everything else I think this is a time where researchers in Academia contribute more than they have been able to in a long time now that we've produced these models there's so much we don't understand about them yet there's so much more we need to do we have great Partnerships with a lot of academics doing work uh on top of the models or work that feeds into the models we desperately need new efficiency gains better architectures algorithms so work on that we'd love to work with you but trying to just like I don't think Academia is like best position to just run the scale up race but that's okay because that's not the most exciting thing anymore there was a time when kaina only open AI believed that and we had like a big Edge there but now everybody believes that and there's I think less Alpha there so go to go do all of the other work that's important to get to the next Paradigm and the one after that uh work on great data sets work on great evals Academia absolutely leads the charge there and then on computer programmers no I mean I think there was a period of time where computer programmers were very worried that copilot and then Chad gbt was going to take their jobs and no one thinks that anymore now everyone just it's a great example of when the floor is lifted up the expectations lift up too and it turned out the world wanted way more code and way more sophisticated code than existed and so now people can output maybe you know two or three times the productivity level and it's still not enough we have one at the back here hi I'm Sarah I'm a college student and my question is how far off do you see the singularity can you define at this point I think we're close enough we have to be more precise on the definition what does the singularity mean to you to me I Define it as a point in where technology reaches a point where there's it's out of control and there's I'm sorry it's out of control and it's at a point where it's irreversible I would say I hope never then thank you okay I think let me add one more sentence that um a one more paragraph I I think it's important that we realize these are tools and not creatures that we're building and that we are we get to make the decisions about how they work I think it'd be a mistake to just say all right you know human out of the loop hand this over do whatever you want change your own architecture go do all these things I think it's very important that the future of humanity is determined by humanity and that is like an active choice we can make please stand up and ask your question Hi Sam Vic Wells from outlaw um my question relates to ethics and I'd like to know how can we design AI systems that reflect our shared human values and should we certainly we should um the question of how I'm very glad you asked this is something we've been thinking about a lot recently we're going to try to learn it I think it's very hard to write down the ethics of all of humanity but maybe you can learn it so maybe you can and we we recently published a new program of grants for people that have ideas about how to do this how you get Democratic input on values of the world how you define what the broad limits of what these systems will do will never do what the defaults are how much an individual country or user can customize it so if you're interested in working on this please apply for our grant program we really want new ideas but this idea that the system can talk to each person on Earth understand their value system maybe help them make some moral progress and say have you thought about it this way here's what other people say here's the perspective you may not have heard it's still up to you if you want to stick with your answer that's fine but let's have a dialogue and then from all of that the system can learn like the collective moral preferences of humanity I think that's pretty exciting hey Sam I'm ridden a 16 year old building Quantum batteries in the backyard super excited uh to see by AI is going to go into the future and I'm part of the organization called TKS here's my question to you what is an AI specific skill that young people should have that will compound over time as we're young ambitious people trying to make a dent in the universe what do you think is an AI specific skill that we should have I'll give you like a meta answer instead of one specific skill a thing that we've been hearing a lot from companies is that you know it used to be that the people that joined right out of school were not the most productive and it was the people that you know had been there for five or seven years were of course more productive like you'd expect and they've seen an inversion of that in the past few months because young people have embraced as always new technology faster uh and so people who have figured out how to use chat CPT in their workflow before the older people have had a huge Edge and are now in many cases more productive and so my meta answer and I think this happens anyway and it's why young people sort of lead technology revolutions is just keep using the tools as right when they come out and figure out how to use them to do all of your work better it's not any one specific skill but it's like the mindset of like there is this amazing new thing in the world it can impact almost anything and I'm going to figure out how to leverage it better than anybody else hi I'm ifato from the Israeli Embassy and Ben Gurion University and my question is about you being a successful entrepreneur so drawing on that what are the main qualities that young entrepreneurs should cultivate in order to be successful and to navigate successfully the uncertainties in the entrepreneurial ecosystem so before I ran openai I ran this thing called y combinator for many years and we invest in lots of startups and so over you know the course of a decade I probably met tens of thousands of entrepreneurs uh funded I guess thousands of them uh at least many hundreds and I spent a lot of time trying to observe what those skills are there's all of the obvious ones like you need to be really smart and you need a good idea and a growing market I'll leave all of those off because they're all over the internet the things I observed that were less obvious number one you need to be very resilient um like the the whole experience of the beginning of a startup is quite miserable it has like moments of fun but on the whole like stuff's going wrong No One Believes In you um you kind of like can't get anything to quite work and it's like a long period of that and most people just give up um one of the things that we learned at YC is you actually can teach people to be resilient at least you can cultivate it and that is like super valuable uh another one is the spirit of like and it's closely related but whatever problem comes my way I'm gonna figure it out even if I don't know how to figure it out right now um and that that sort of just like It's a combination of self-belief and adaptability and resilience that's that's very important another one is you have to be at least a very good and ideally a great communicator your job especially the beginning of a startup ends up being like kind of like Chief evangelist uh you know when we started openai and we were talking about AGI most people thought we were like crazy at best many of them probably thought we were just like full on like running a scam um and to be able to convince people that like hey you know this is unlikely but let's try it's so valuable if it works it's not prevented by the laws of physics there's these promising things getting the energy together to do that um requires a lot of good communication to people you're trying to hire investors uh you know Partners customers whatever and that was a skill that I think just doesn't get talked about but is is quite important um and then another one and then I'll stop this is I could give like an hour answer on this I thought about it for a long time um a very long-term orientation so I think most people are you know like what can I get done in a few years and if you're willing to say what can I get done in a few decades even if I have to be totally misunderstood for the next few years uh that is a very powerful skill I can't resist on one more uh just like the ability to get a lot of stuff done quickly which I think is also a teachable skill some combination of like decisiveness willfulness prioritization whatever you want to call it but like a lot of productivity per unit time a startup really Rewards someday I'm going to write like a very long blog post about the other 300 things but there's a quick one Hi Sam I'm Bjorn I'm the founder of care syntax and we're bringing AI to surgery and I have a question I have two daughters that are very young and sometimes I get worried about the future and if AI is gonna turn malignant on humans and then I watch The Internet's videos and it says it's a 50 50 whether or not that happens and my question to you is um do you think are you worried about that and do you think that actually um with more intelligence eyewitness at least in humans more Humanity unless there's a personality disorder or something like this do you think that is or is that you believe as well so I am worried about this uh but I'm way way way more optimistic than 50 50. I think people say 50 50 when they're just like oh I can't think rigorously about this but more than that I think it's a number we can affect the work we all collectively do changes that number a lot I think there are very unsafe paths and there's like much safer paths I think we're actually heading to a pretty safe path when when it looked like we were going to maybe get to AGI by having these reinforcement learning agents learn to play games they were just trying to like beat humans and deceive humans while they were beating them and then play more and more sophisticated things and eventually release be released into the real world and be totally inscrutable I was like that doesn't sound great uh but I think the path that we're on now of a language model that takes in some text and gives you back some text and can express what it's doing step by step in natural language it can still be horribly misused it can still go totally wrong but I think we at least have a path to the optimistic outcome I think the degree to which the world is now discussing AGI and the risk associated with it and what the world needs to do to come together to have a regulatory framework to mitigate this risk is a tremendously positive sign that if you asked me a year ago would this be happening right now I would say very unlikely and so I you know I think that's a great step as well um so I'd say we have a lot of work to do but playing the guessing game of what the probability of Doom is is not particularly helpful the answer is to go do the hard work to like get it to Trend closer and closer to zero hi uh from the financial times hi um I just want to give you if you could talk a bit more about your Ambitions in this region uh what kind of conversations you're having with the authorities here um and in terms of Regulation what those conversations are and where you think that might be going here especially given your concerns about regulation in the in Europe thank you very much I first of all I think regulation in Europe is good I I think I was like quoted completely out of context and actually just there were like the articles that I said something I just didn't say um what I'm meant to say and what I believe and what I think I said is that if we can't comply with the law in Europe of course we won't break it but we expect a productive conversation and we think that regulation makes sense and there'll be a path here that's really good in terms of the region there's uh there's been like a huge amount of excitement we've been here the last couple of days uh and I I am hopeful that the region can play a central role in This Global conversation again I think there's been discussion about AI here in particular Abu Dhabi in particular before it was cool um you know now like everybody's on the AI bandwagon which we're excited about but we have like special appreciation for the people that were like you know talking about this when everyone thought hey I was not going to happen uh so I think there's desire commitment willingness here to participate and this idea of a global organization that leaves models of the current scale and certainly anything smaller probably bigger two alone but really thinks about how we're going to address that existential risk we were just talking about what a licensing framework looks like how we establish this sort of series of tests and evaluations a model has to pass while it's training above a certain scale how we decide we get to deploy it I think there's excitement in the region to play a lead enroll there thank you Sam once again my name is I'm Deputy CEO of Hub 71. perhaps my question relates more to your prior experience with the YC so I have the pleasure of looking after Hub 71 which is a tech Hub based out of Abu Dhabi I'm actually here next to my colleague who looks after the tech Hub out of charger another Emirates in the UAE so we'd love to get your advice opinion on what's the best environment or initiatives for us to support and scale startups out of the UAE and this region you should discount me heavily what I'm about to say but I would just go all in on AI as like the thing that is gonna matter in the next decade I mean I think like we are about to have the greatest startup boom since the internet and anyone everyone's going to win like it's it and I would just stay focused on that hi uh my name is O'Hare I'm uh one of the co-founders of a company called jalebi we're building a restaurant technology company but that's unimportant right now um thank you very much I I just wanted to bring a little bit of context and it's kind of a generational question so I'm not sure if you're familiar with this concept called Future Shock it's something that I'm very fascinated by it was a book written in the 60s very dystopian and I think about this this problem that I'm having where I've got a three and a half year old son and I have a 91 year old dad and I want to find a way for them to be able to communicate with each other in their geographically distant and I'm wondering if there's a way that AI or technology today could actually help that because my worry is that that generational Gap is so large that my son won't have a chance to have a relationship with his grandfather it's so this is I don't know if this is in the realm you know I I'm very interested in this as a general topic one of the things that I think is interesting about let me back up even a little bit more I I think the history of Technology of computers at least has been towards simpler and more natural interfaces so we used to have like punch cards only very sophisticated people could do that probably much older people never learned that and then we kind of like got to the command line and then by the time we got to the graphical user interface and a mouse and a keyboard you know definitely young people were more fluent and it per it gave some Future Shock to older people but like older people learned to use those that was a mass-market thing and then the iPhone came along and I think multi-touch was even more natural because we got rid of that mouse and keyboard now we're just using our hands and little kids could pick those up old people who never were able to figure out how to use computers or they could have but they didn't want to figure that out and natural language the natural language interface the sort of chat gbt idea is even a step further everybody knows how to talk almost everybody knows how to text and we're now at a kind of where I call it like the Star Trek computer but it's kind of like the computer that sci-fi always predicted or close you know we got to get the whole Holodeck thing working still um but probably your three and a half year old and your 90 and a 91 year old can both use it and that's an unusual thing about a new technology and I I am hopeful that although Future Shock with a like the pace is going to feel breathtaking the rate of change in the world is going to be intense but it's wrapped in an interface that kind of everybody you know were very naturally suited to use and and I think it's going to bring people along in a way that we haven't yet seen with any technology before do they both use it my my dad is like 70 he can't hear so when they have a conversation it's mostly just hello hello hello so I'm I'm not sure exactly what the use is but there's a sort of a a gap that's forming and it's it's harder and harder to fill that and I'm wondering if you know I I am hopeful that somehow this is going to help good afternoon uh my name is I run the Film Production house right and I'm referring to one of the latest blog posts you had in on your company's website where you talk about that we might expect this super intelligent ATI within the next 10 years at some point do you think that the global Workforce is moving towards a universal basic income in order to sustain the massive amount of efficiency that AI provides and probably a Workforce being laid off thank you so I think we go through technological revolutions we've gone through many in history and it appears to me from trying to study them that in about two generations we can adapt to any amount of Labor Market change the hard part about this one is how fast it may happen now we may be totally wrong we're not saying we're sure this happens in 10 years we're just saying it'd be good for the world to uh you know put some effort into planning for it on a 10-year Horizon like I think that would be that would be prudent um but it's the speed that I think makes it challenging I think a Ubi is a good idea something like that is a good idea but it's totally insufficient people work for a lot of reasons even if they don't need to make more money the the Fulfillment that comes from it the sense of purpose the desire to create the desire to be useful that's not going to go anywhere and I'm also not a believer that we just run out of work to do I think with more powerful tools we will do a lot more stuff it may almost certainly will look super different than many of the jobs of today but you know I think I could like name with confidence many jobs of the future and there will be many more that none of us can imagine uh or at least we'd only imagine if by chance we're just going to do different stuff and it'll be I think way better we will look back at you know we will look back at this period of 2023 and say we didn't have very good lives then relative to what what's possible now hello hi Hi Sam yep hi hello I don't know what to do um my name is I'm a student at credit University I really a research intensive University here we focus on AI computer vision you know a lot of engineering disciplines we focus on I work also out out of a early stage Fund in Silicon Valley my question is simply we know that open Ai and llms are huge on in terms of you know making productivity exponential and a 3dp is the technical depth that has been there or have been there for the past 20 years since the Inception of coding and programming and computers I really want to know and ask you about you know we see a lot of companies like zap here where they've scaled their company to you know a bunch of billion dollars per year in Revenue uh and and you know in a team that is under 20 or under 30 people can we see that um sort of companies where we see companies that build billion dollar companies um with three or four or five people I I have in my WhatsApp group with my friends uh we have a little betting pool for the first year that a single person one billion dollar company is founded so I think so yes so I'm just going to quickly dive in because we've got another question online is this is falling on from simeon's question and it's Donna from CNN business uh Arabia and she's saying that you threatened to leave the EU over the AI law and then reconsider your decision what is your action plan once the EU and other governments start implementing AI regulation particularly copyright I think she's focusing on yeah we did not threaten to leave the EU we said that we are going to work with the EU we expect to be able to comply you know there's still more clarity we're waiting for on the EU AI act but we're very excited to operate in Europe um unlike some other AI companies we currently operate in Europe uh what we said is if we can't comply with the law then we won't break it which I thought was like one of the most innocuous and least offensive statements one could possibly make but apparently not um so that's our plan uh in terms of copyright look legally I think from a fair use perspective we're in you know that's very clear but morally we want to find this is like a new technology and we want to find a way that people that are creating content people that are making the AI systems better um people who are you know helping benefit users of the AI get rewarded for it like that seems like a great thing to do we're engaged now with content owners and content creators about what they want people want very different things so we're going to have to like figure out a system that works reasonably well for everybody um but that's like that's totally where we all should get Hi Sam hello yeah my name is Kashif I'm a founder of a company called Peko so PECO is an all-in-one platform for smbs to manage all their payments like a superf or smes so what we have done is as as the entrepreneurship ecosystem is growing in the region so we are building an AI platform called Phantom which Phantom is a platform which will help these company startups you know smbs uh in you know in in setting up their business from scratch to helping them on every level and we are also building a lot of content which is which is really uh you know uh localized content in terms of you know helping them in in certain processes but now we want to also once we are live here we want to also build this it's like a new y combinator startup School kind of a thing but more in terms of AI uh you know platform so how can we how are we able to scale it in different uh different you know uh different uh geographies with the same concept that it will help entrepreneurs build their businesses faster the question is what will help you scale it faster yeah like the same because because we feel that we have to build a lot of content in order to make sure that the entrepreneur has you know all the processes streamlined whatever they need in in their uh entrepreneurship Journey yeah I mean I think llm should be really good at that uh like if a GP if you can't figure out how to make gpt4 pretty good at that I'd be surprised and quite disappointed let me know if it doesn't work and it'll be interesting research question for us thank you Hi Sam sorry executive office can you give us a sneak peek of what gpt10 looks like I mean I am I promise I'm at least as curious as you uh I think by the time of like something we call gbt 10 we are we're out of just the pure like I actually can't even predict I was gonna make a bunch of statements and I was like maybe these will all be wrong I don't want to say something stupid I think it will be like an extraordinary thing uh I think it'll be something that like is a giant multiplier on human ability and productivity but the exact shape of it that's a little hard predicting that for how it is pretty I mean that's like you know we're talk we're talking something that's like more than a decade out that's pretty hard Hi Sam hi my name's Lee I have been working with the atgm fintech team for the last year using llms like open Ai and in a regulated space hallucinations are quite a problem um is your thought on critical industries that we have to accept hallucinations in LMS or as we get to five six seven eight they will slowly disappear we are working hard to make them slowly disappear uh it's definitely one of the biggest challenges in the way of using these models we have a lot of ideas for what we can do I'd say our team is like optimistic that we're going to solve this problem but it's still research so I can't like Promise You by exactly when but we are very focused on it hello Hi Sam good afternoon my name is Louis I'm from Venom blockchain Foundation and my question is about giving the power of decision to llms so they're fantastic knowledge base and we're working with a particular design and an architecture and one of the things that we I'd love to hear your thoughts on are inserting memory and inserting time into the llm so it can make a decision and predict going forward in the future we have a particular architecture which is connecting different type of AI to neural networks and I I'd look we'd love to hear your thoughts on what you predict going forward to give AGI llms this decision-making power it's definitely important to do um I don't know exactly like when we will get to it as a research problem uh but we will at some point I think you know uh Power like a really useful system needs both of those things hello Sam my name is Jose people from Hub 71 they call me crazy Jose Jose sometimes because I'm always running here and there and we create technology for people with disabilities to use computers and everything if you were Bill Gates I would say that we created this to use Microsoft but as you are some Altman I can say that we created this for these kids to use chat DPT and it's working because kids with six seven years olds they are writing books uh using our technology and uh to to support them to write books and to do everything and I would like to ask you what have you been doing in this field for people with disabilities and if you accept collaborations I have already written something here if you want to check later yeah we'd be happy to talk uh we we we're working a lot on new modalities to increase accessibility we also hear that even with our current systems people with disabilities for example are able to use our speech recognition system that may not be able to use others but we want the system to be as usable for people with disabilities as possible and I think we've made some progress a lot more hi I'm darika I'm a student at annoy Abu Dhabi and I want to get your opinion on what is your perspective on how to best integrate artificial intelligence in college education and what do you see in the future of college students across across different majors and how do they use AI to improve the future thank you I mean I think it's already transformed in education I know we can look in the future but I think we already see what's happening which is people will use this for you know something between one-on-one tutoring and interactive textbooks but you can have a system that'll really get to know you your learning style what you know what you're interested in what you're bad at what you don't know and provide you like a very personalized experience even even just using chat GPT with gpt4 out of the box you can do pretty well and if you use some of the partners that have built services on top you can do very well so I think we'll just see more of that hello Sam I wanted to ask if their AI aict measures that you have been using in open AI to combat specifically the alignment problem great question um in the short term I think we have pretty good techniques so our lhf which we developed we used heavily for gpt4 along with some other ideas like refusals where we train the model to not do certain things and for the current scale that's worked pretty well in fact this works so well people look at gpt4 and say the alignment problem is solved and that's very scary it's definitely not and I think people can get overconfident looking at what the current models can do we will need to develop new techniques as the technology continues on we've put out a couple of blog posts about the ideas that we're currently most excited about where we're sort of using models that we understand to help us evaluate the output of models that are more powerful and then eventually we can use the models themselves to help us come up with new alignment research ideas and then we have many other ideas we just started a new team focused on more advanced ideas and more of that will come if anyone is interested in working on this you should definitely get in touch with us I think this is like one of the most important challenges of the next decade hey Sam hi I'm founder of purchase payloads the space Tech startup where we're like a Marketplace between the rockets and satellites and I would say the third of our operation already been run by chargeable including legal contracts and everything uh so I have a question rather personal so um I'm looking at you you know I look at you face to face look at you in Congress hearings and everything and you've been asked where whether you have equity in open AI right and your answer is no I have enough money but you know it's the thing that I cannot reconcile in my head being founder myself and especially you know I've done the founder in Silicon Valley like I question myself if today your company worth as we say mashallah in you know thousands of billions of dollars in valuation now and inshallah in you know a couple of years it'll be hundreds of billion dollars right how do you we all you know we all live in a capitalism mode right how do you reconcile with that you know how you can tell that if your company would grow to like a trillion dollar club and you don't have the equity in that you know what what about the loss of opportunity about all the power that you've gotten with all this Equity how how does this work I I cannot reconcile that I don't know I mean I never think about it except when like someone asks in Congress and then I'm like oh yeah it's just not like I'm already like I'm the more money you have to like figure out how to give it away effectively and I just want to like think about Ai and I I mean I have like lots of selfish reasons for doing this and as you said I get like all of the Power of running open AI but I can't think of like anything more fulfilling to work on I don't think it's like particularly altruistic uh because it would be if I like didn't already have a bunch of money but like I get quite a lot of value that is worth much more to me than the money from working on open AI so in that sense like I don't think I'm like giving up that much and I don't know it's just like I don't think about it I have like a lot of investments in other startups like you know yeah the money's gonna like pile up faster than I can spend it anyway um I like being I like being non-conflicted on openai because I think the chance that we have to make a very strange decision someday um is non-trivial and you know I like seeing numbers and scores go up so I guess there's like some value there but honestly the way it started was like there was a technicality where it was better for me at the time not to have Equity because of the way our board was structured I thought I would like take it someday and I kind of never got around to it and now it's just like I'm happy not to have it but I don't again like the value I get personally from doing open AI so much eclipse is what more money would do for me that I don't think it's like a particularly notable thing thanks Hi Sam Hi Sam my name is here is open AI planning an IPO soon no thank you no we have a very strange structure we have this cap to profit thing we also like when we develop super intelligence are likely to make some decisions that most investors would look at very strangely I don't really want to be like sued by a bunch of like public market Wall Street whatevers so no not that interested Hi Sam great to meet you it's really a great discussion that we're having with you here Abdullah from Emirates Sports Group Sam we want to see your opinion how would uh AI web 3.0 blockchain iot would really be a core system that would really cause the disruption that we're waiting for and today I think the biggest disrupter in our hands is our mobile phones one would AI be a backbone of all our smartphones I think there we'll see a bunch of ways where some of these Technologies come together uh one is that verification of content may become quite important if if there can be generated content you kind of want to know if somebody's you know if it was made by a person if by which person and it was authenticated and I think the blockchain might be a very interesting way to do that I think we will also need ways to distribute the wealth generated by these systems the governance of them access to them and blockchain could be an interesting way to do that so you know like we can't predict the future exactly I'm not sure what else will happen but that's the kind of stuff I can imagine happening and then in terms of this becoming the backbone of the backbone of smartphones I would bet that in a few years it'll take some time for this to propagate but I would bet within like a few years a big part of the way we use our smartphones feels like the way we use Chachi PT hi Hi Sam I'm over here in the middle right in front of you hi uh thanks for coming to Abu Dhabi and giving us this opportunity to ask you some questions um I'm Vera from lamsa in recent news we hear various governments and people who have worked with the technology closely expressing concern on how AI could lead to the extinction of humankind without proper regulation we know that AI is here to stay and you say that there's so much more work to do so as someone who is aware of this impending danger being the man of your position influence and knowing what you know what can you advise us as users of AI as change makers in government business and just in our personalized personal lives even on how to ethically make use of the intelligence to ensure a symbiotic relationship moving forward for Humanity thanks for the question I I think some of the most important development has already happened which is we're having this conversation a few years ago the world was not very focused on AI we sort of saw this freight train coming uh I totally think we can manage through this transition but the more we can work together the more we can have this conversation now while we still have a lot of relatively a lot of time to decide what to do the better we'll be and I think the fact that we're getting this Global conversation going the fact that policymakers are paid attention the fact that there's real momentum towards some sort of international organization I think that's not no small thing you know we have a lot of work in front of us but it's worth celebrating where we are now and the reason this is happening is because it has become something so many people are talking about seriously and and weighing the importance of so I think continuing to push on that more is great advocating for support of organizations that can help continue to make sure AI is used responsibly and that we keep an eye on long-term safety in addition to short-term it's great too but I think the right things are really happening Hi Sam my name is hello this side my name is Neymar fellasi and I work for a model Investment Company so my question is very much related to regulation so we've seen a lot of effort from different regulators and policy makers across the world recently so what do you think would be the best model to regulate Ai and is it a mix of some Universal concepts with adaptation at local levels so what are your thoughts on this thank you yeah I think I think something that is global for the parts that affect us globally um you know in the same way we did for nuclear material makes a lot of sense and then I hope that we don't have too much regulation on the smaller systems that have some risks but risks that we can afford to be reactive to and we really want to let develop and Thrive and then I think countries can make a huge amount of determination Within These very broad Global bonds so I think it'll have to happen in multiple tiers Hi Sam my name is Dima I'm product manager at bezat when you opened your apis uh to other companies we started using it almost next week but here but so many companies did the same and um now you can see AI power it's powered by charging PT powered by open AI is like all these tables are everywhere on internet so my question is is the product questions how do you think companies should build their AI powered products so they create competitive Advantage for them and it's hard to replicate when the iPhone app store first launched there were a couple of years yeah maybe not a couple years maybe it's like 18 months where people would say I'm a Mobile powered company I'm a mobile company and that was like a differentiator and it was like a big deal and whatever and there was a little boom of new companies that only made sense with these new super computers we had in our pocket like Uber or whatever but no one would call themselves a mobile company now because every company is expected to have a mobile app right now people talk about being AI powered or an AI company in a few years I think no one will say that because everybody will have ai built in it'll it would be ridiculous not to now you could say that's bad because you can't differentiate that way but I think it's fine because you can't really differentiate that way now either AI does not get you off the hook for having to be a good company you still have to like figure out a way to differentiate in the market you still have to find some sort of network effect competitive advantage and this is just like one more tool that lets you do more things better but it doesn't it doesn't it doesn't excuse you from the hard work of building the company that's fine hello thank you very much Charles my name is Willie Gomez ambassador from Guatemala um I have a question what do you think about biases and uh do you think that it's something that we can uh you can handle or manage or avoid especially for those countries that are regions that are producing a lot of data but there are other regions that are producing small amount of data what's going to happen in the future with the biases of AI so bias in the traditional sense uh I think it will turn out that these systems can be much less biased than humans I think these systems can actually be a tool for reducing bias in the world you know if we'd like think of like sexism or something like that a classic bias um because they don't have the psychological flaws that we all do and we can teach them to be unbiased and I think it's easier to do that than a human with all of their built up programming and emotion and sort of societal upbringing that as hard as we work to erase we're imperfect at so I think these systems can help the world be less biased and I'm used to be quite worried about this but I think with some of the new techniques that are developed and you can read recent papers on this issue I'm optimistic we can make progress there but on the second part of your question a world where the value system for the AIS come in from a small number of countries or a small number of not even let's say countries but viewpoints that are producing most of the alignment data for an AI or the values that would be really bad I'm very worried about that part of the reason we're doing those grants that I mentioned earlier is to explore ways to get actual diverse Global representation in there because I think whatever AGI we eventually build really should reflect the like sum total and the expansiveness of humanity so thank you so much Sam I know that's the last question maybe we can squeeze one quick one in while you're standing but make it brief please thank you for come here my name is Emirates post I was just wondering if Open the Eyes going to get into the Robotics and and if you are planning to integrate with Tesla bot or the Tesla Android or Boston Dynamics or Empower those with AI that you have we started in robotics um and we were very excited about it we made this little robotic hand that could do a Rubik's Cube one-handed and then we stopped because we're a small company and we have to be focused and Robotics were hard for us for the wrong reason they weren't hard because the ml was hard they were hard because the robot always broke and was inaccurate and that's a bummer because I think Robotics are awesome but we don't have and still don't have the luxury to pursue every bit we'd like to pursue so when we saw the language model approach and we said this can get to AGI and that's really what our mission's about we put the robotics thing on hold I'm still sad about that personally and I hope someday we get back to it Sam thank you so much big round of applause please for Sam Altman as CEO of open AI I'm so sorry I know there's not enough time hey yeah thank you I know you've got a plane to catch and that you've been very generous with your time today I wish we had more time thank you all very much thank you [Music] [Music] [Applause]

Channel: Hub71

Views: 46,805

Keywords: livestream, Sam Altman, samuel altman, chat gpt, openai, ai, artificial intelligence, Abu Dhabi, Hub71, ChatGPT, ADGM, Abu Dhabi Global Market, sam altman interview, In Abu Dhabi, united arab emirates, technology, technology video

Id: RZd870NCukg

Length: 60min 49sec (3649 seconds)

Published: Thu Jun 08 2023