A conversation with OpenAI’s CEO, Sam Altman | Hosted by UCL

nice and quiet fantastic so I now have a chance to rest my Noggin uh for a few minutes um we have a fantastic panel and we are actually taking questions from the floor really almost from the outset so please be ready there are a few roaming mics uh kicking around um and I will I will choose you as um as I spot you and when it comes to questions obviously it's a really um opportune and exciting moment please keep the questions to questions um rather than statements and I will up front say that I will interrupt uh any long statements or things that don't look like they're going to be uh questions um and if we can just keep phones uh turned off again at least on silent uh for this that would be fantastic so let's start with the introduction to our really really fantastic panelists uh you of course everyone knows Sam next to Sam we have Professor Yvonne Rogers who is a professor here at UCL in human computer interaction uh next to her David Barber Professor David Barber who is the head of the AI Artificial Intelligence Center here and next to me is Margaret McCabe who is an entrepreneur and a philanthropist who's focusing on education so it's a wonderful panel thank you for the three of you for joining us and there is so much from the previous session I wanted to start perhaps with um a question for for the the panel which is as you three educationalists in a sense you look at what you have seen over the last six months um what is the one thing that you would like to offer to open AI as we look at this question and perhaps we'll start with you thank you very much thanks Sam for coming but I run debate mate I set it up I found that I'm embarrassed I'm not like I'm not a scientist whatsoever I'm here representing humanity and yeah yeah I know that's why I'm excited so what have I seen when I first heard about the chatbot AI fabulous this is our chance my aim my ambition is to make excellent education available across the world to everyone everywhere and I used to think a couple years ago how am I gonna get there how am I going to get there and uh is it AI turns out I said this is exactly how we're going to start across Africa and and we're going to take it from there so basically what we do we teach basic skills drawn from British debating which sounds a bit like obtuse and far out there but it works we've been doing it for 14 15 years and being a human you have to have the skills we teach so artificial intelligence is not real it's not that intelligent human intelligence dictates as you've seen a very intelligent person here and here human intelligence dictates that goes all right I'm used to it yeah that goes without saying a human so human intelligence basically uh dictates what's going on in Ai and what we've noticed is that you need these human skills so what we do at debate mate we make better humans we have to have better humans intelligence is one of those evocative words that's a little bit intimidating because people either see themselves as intelligent or not intelligent but intelligence really just reflects what education you've had so and that's not a perfect system and what we're doing at debate mode so with with respect to the GPT sorry what was your your sense of what it would mean for Education well what we're basically doing is we can help those who are writing the chat we've helped those who are programming writing it have the skills that are needed and they're the skills of empathy creativity and critical thinking and resilience we can teach those skills to the people who are writing the programs they don't have those now but also what we can do is I want to be able to offer our curriculums to people who are using it and that's where I see the opportunity to widen widen access well Bill Gates said that children now will be able to learn to read and write in a year and a half we're going to be part of that how about you David from an education's perspective what was the thing the one thing briefly that uh came out when you saw these these new technologies uh well I don't know about from an education perspective from a researcher's perspective it's maybe a little bit easier for me to think about I think it's a great example I think of how you can put resources focus them on a very interesting problem like Ai and make great progress and I think it's it's very interesting that these kinds of developments have happened actually outside of the University environment I think it's a it's something that we need to reflect on a little bit to think about you know well why didn't that happen I mean maybe some of the core ideas actually were there in coming out of University but the scale obviously wasn't available and I think it's very it's very interesting and I think we need to think very carefully about how we organize ourselves universities I ask you to consider are we doing it right I mean why is it that these developments are and the outside universities are not actually inside them so I would like to see actually more of um poorest relationship between universities and you know and businesses and companies who are actually trying to push these boundaries I think it was obviously it's a golden age for computer science but I don't know I don't know if we feel it so much within the universities we feel very much into pressure and funded and I'd very much like to somehow change that mindset a little bit that attitude and I think there's something we're not quite doing right I think it's uh it's amazing to me that if these things happen outside the universities and how do we address that balance how do we get that that better fascinating and and everyone when I first saw chat GPT I thought oh wow it looks a bit like Google search you just type in your words and then when it came back with the answer one word by one word I thought it was very clever and I think it's a really good interface to start with but I think what we can offer open AI from a human computer interaction perspective is different ways by which to design that interface to radically change how we do things and carry out our activities so we might like to combine multimodal interfaces but we can also think about AR and other technologies that are out there and how we can combine these and also good old-fashioned graphical user interfaces so that it's not just a chat box but there are different ways in which people can interact whether it's teams of people or individuals whether it's young children or very old people I think that's what we can offer you is that we do lots of research in that area that's fascinating Sam I I want to take questions from the floor so please warm your brains up but can I ask you to comment on what David said uh about the the fact that it seems very exciting but perhaps not so exciting within uh traditional Academia I mean how important is maintaining that the kind of research that the academy is uniquely placed to do how important is it yeah I think it's super important uh we obviously built on a lot of that hopefully we contributed some back there there will be things that the academy does that we're just not well suited for I think we're seeing that now with data sets and evals and some of the interpretability work but I think it's great if the academy is good at some stuff and companies like us are good at some stuff we're a very different shape we can concentrate a ton of like we get to run as more of like a focused effort than I think an academic Department ever could and so we can say hey we have this one idea we're going to put hundreds of people and a ton of capital into it and we have a good shape for that and I don't think the academy always does but I think it's great that we have multiple ways important intellectual progress can happen thank you right let's get to the audience if we can have hands up uh there's a fair number I don't even know where we're going to start um there's a there's a gentleman in that full spray waving he's got nice great hoodie on oh there's a mic actually if you can pass it over and just a brief question thanks afternoon thank you very much for your time [Music] uh yeah um what role does regulation have in AI adoption um thank you very much in AI adoption um I think it's important to get the balance right here and the right answer is probably something between the traditional European UK approach and the traditional us approach I hope we can all get it right together this time in the middle I think you really don't want to over regulate before you know what the shape of the technology is going to be I think you really don't want to allow um you really don't want a lot of great games where you capture for companies like us but you know like more than that if they're these calls to like stop the open source movement um that I think would be a real shame um or or even smaller model development however on the other hand I think most people would agree I would assume that most of this audience would agree that like if someone does crack the code and builds a super intelligence however you want to Define that like probably some Global rules on that are appropriate I think there is a moment right now where people are realizing that's possible to that this something like this could be built and where we still have a potential of getting to some degree some Global agreement and some sort of global set of norms and enforcement and I'm very excited that conversation's been happening um it's been it's been a recurring topic on this trip I've been doing around the world um and I'd like to make sure that we have you know we treat this at least as seriously as we treat say nuclear material for the Mega scale systems that could give birth to Super intelligence um I think it's easy to caricaturize and lose the Nuance on both sides so you have people saying well you're trying to shut down small projects or well you know you're you're just like you're not calling for enough regulation but at some level it's like let's figure out the capability threshold let's move it as we learn more and let's have a global framework for Stuff above that and not screw it up on the downside I see a world coin orb in the first row I had to sign up while I was in Europe because you can't do that in the US and I think like that's an example of where the US has made a mistake in regulation that traditionally I think other countries make um so we just got to get the balance right on that points about a global yearly compact the argument is is always that the global system moves really really slowly the UN apart from this the incredible speed with which it was set up is not renowned for moving moving quickly so what is the kind of feedback you've had as you've been um traveling about the ambition and the ability to bring a coalition together quite quickly yeah it like it's like always initial like ah we can't do that and then we talk more about it and kind of why it's important and also what the degrees of freedom are and then people are like all right maybe we can let's try I think we're going to try which I think is great but observations from the from the panelists on that question think locally we are very concerned in education the universities and schools about students using chat gbt to to do their assignments to do their exams and and the worry is there what should we do about it do we need new regulation and I don't think we you know we're not there yet but I think we definitely need to have those discussions about how much do we regulate or how much do we Embrace these new technologies and change the way in which we teach and assess and I think there needs to be important debates about the specific types of governance well you know funnily enough maybe I'm a contrarian I'm more concerned about students who are not using chat GPT to enhance their education [Music] it's really good for learning difficulties dyspraxia dyslexia we know that it's amazing it can help kids to read and write I'm really uh impressed with Sam admitting that America's made a mistake about this because once you can admit in a mistake that's a an adult skill right and that's where the Innovation happens so I think and I can't go into what we're doing but at the moment we're very keen on making sure that children in school get the best out of this but they're not known they'll they'll actually be educated in my plagiarism and cheating and that sort of thing so don't be worried don't be afraid for example no well some you look like you want oh I was gonna say you can find these articles uh of teachers calling for the Banning of calculators in math class when they were first invented saying it was the end of mathematical education and it turns out that we just when we get new tools the the level of ambition and expectation and capability just goes up so you have chat CPT um people try to ban it for a little while they've already said all right actually this is the most important educational tool in a while and we're just going to figure out how to teach at a higher level and that's great and I agree if you're not using it that's a probably a real mistake David yeah I feel like you know it's a super important question right but I'm not sure that we really know these technologies that well yet but it's really I think very early days and I think there are existing Frameworks already to regulate you know you shouldn't do bad stuff anyway right there are laws you know against deformation etc etc right so you know you shouldn't feel that you can take these Technologies abuse them and get away with it anyway even right now so I've been cautious about over regulating right now I'd like to see you know people get very aware of the capabilities research that but I think how we regulate them I think it needs to be needs to be dealt with some caution can I just add something in there as a barrister there's a massive difference between regulation and legislation I mean there are laws against all this stuff now uh but regulation is a very uh it's a very lawful sort of field we've got good and bad regulation in this country to see what's happened with those sewage problems uh and the uh EU is passing you know widespread legislation that everyone's sort of relying on that's going to solve everything so uh it's a question of a good and bad regulation how it affects the laws that already enforced right we'll take some more questions if I get hands coming up here there's a gentleman oh there's a lady there in the whites uh basically yeah thank you everyone for coming today Sam um so question to you Sam when you're thinking about the future and you're thinking about something that you guys touched on earlier which is job creation and we're talking about how disruptive um you know generative AI can be to education when you think about the role of Education as education for employment historically humans used to be good processors and computers got to be better at logic and compute and we were no longer I'm going to push you for the question the question is if you had kids today what would you push them to learn to be employable in 15 or 20 years if you think about the speed at which the changing I don't have an answer on the hard skills I mean I have like some guesses personally but I not thinks I'm confident when I'm confident in is the rate of change in the world will be high and so adaptability resilience creativity and drive and passion and the desire to make new things and to sort of run a tight feedback loop I think these are all very teachable skills I don't think we teach them enough I actually think teaching well I'll say that in the interest of I'll give a quick answer too um and and you know real fluency with these tools uh like that's what I would push on Margaret this was in your initial brief with me as well you wanted to talk about this what's your take on well basically we we teach these human skills these transferable skills resilience critical thinking communication which includes listening and speaking how does work strategically in a team uh has to be empathetic and compassionate these will be these skills that we teach will are now moving very fast to the center of mainstream education and so many learned people say that these are the skills we UPS everyone absolutely needs to learn if you're asking about hard skills about what careers you know we're still going to need doctors and we're going to need dentists and lawyers but I think as Sam said the science is going to go ahead so fast and we're going to progress so fast that the actual content of these existing courses will change but also there'll be new ones like there's going to be masses of new new jobs around so it's just a question of keeping your own what's going on right we should get some more questions going and what I'm going to try and do is take a couple at a time now and bundle them so maybe from one of the upper rows there's a there's a lady and a Gentleman right yes right by you right by the mic person with a red tie and then there's the woman next to we've got a technical problem hi thank you so much Assam for coming in today College I want to ask you a couple of questions but first you get one question and one more brief question what is the what is the uh what's the strategy and the vision of open AI for the apis and use it with like third parties and how do you envision that a relationship is going to develop over the future great thank you for that one question the next question is uh yes thank you so much I would like to talk about the subject that you have brought previously in your speech about the societies that we build so the question is uh what do you think would be the role of artificial intelligence in playing uh in creating this sustainable sustainable Life on Mars okay uh great I can I can put those two questions together um so the first question really is about how the business model right and how you imagine the um the the business of open AI to uh to to expand through the APR we'll take the second question in a second um yeah we want to make the platform that people like the most for intelligence we want to have like intelligence as a service that gets smarter and smarter and better and better and cheaper and cheaper and that people plug it in once uh we'll probably change the format of the API a little bit over time we can make it more stable we can do a bunch of other things um but what we'd like is that a lot of people integrate our API and then as we make the underlying model better it kind of like lifts the whole world of products and services up it's a very simple strategy and I think all we have to do to make it work is make better and better models and get better and better at like listening to what users of the API want that's been a big part of this trip for me too which has been super fun um Mars question which is um can we rely on you to help us create sustainable Life on Mars look I have no desire to go live on Mars it sounds horrible um like but I'm happy other people do uh and I think if it's so inhospitable as a place to live that if we can send robots first and we can like kind of spruce it up a little bit before people have to go live there uh that seems like much better than subjecting humans to Mars in its current form but yeah I think Earth is like really quite wonderful agreement the Earth is wonderful um I would totally agree and you could take a lesson out of uh Virgin uh yesterday I think it was all this week Richard Branson has pulled out of the space competition so yeah yes space is great for robots like we we need to go we need to send out robots to space we need to send other probes but uh I'm not it's not so great for humans I think right thank you let's get a couple more questions going so um just at the front there was a gentlemania I said to you then I pulled you away in the dark t-shirt and then there's a someone with a check t-shirt behind him yeah the skating laws are pointing in the pretty exciting scary Direction at times um do you believe that humans can control vastly super intelligence and also second question sorry okay I didn't hear that one Etc it's a similar question actually do you think the world can be trusted with the technology that you're unleashing okay good so we can bundle those together the first question was um you know will we be able to um control superintelligence and what why might we be able to or not be able to and and I think the second question is are we ready for this technology or unleashing I would as a not ready to say I'm not sure you are necessarily unleashing it uh on us as a different sense but but this is a question of control yeah um I think there are some versions of super intelligence that absolutely are alignable and controllable and can build Safety Systems around I mentioned like One Vision for it earlier that I think is workable uh but there's ways in which people control systems smarter than them all of the time like I mean most people at open air are much smarter than me but certainly open AI collectively is much smarter than me and yet I kind of get open-ended to do what I think is the right thing and there there's a lot of other versions of that at much greater scales um we have a lot of ideas that we've published about how we think alignment of super intelligence systems works but I believe it is a technically solvable problem and I I feel more confident that answer now than I did a few years ago um there are paths that I think would be not very good and I hope we avoid those but honestly I'm pretty happy about the trajectory things are currently on I feel much better like these are systems that can express themselves in language and we can have them do it in small steps that we can understand they can help us understand our things so I feel very good about kind of our part to nudge the AI revolution in One Direction but it could get nudged back into one that I don't feel as good about for sure um I'm again pretty pretty happy though with where things are uh is the world kind of ready for this can be trusted with it I'm a very optimistic person so like please discount this and I think it's important there's a lot of pessimistic people in the world to counterbalance it I think the world is going to surprise us on the upside I think we are for all of the potential of these systems for good and for bad I believe they'll be orders magnitude good and bad we're seeing it right now and I don't know I think like can we trust the world to do the right thing without this like if we stay in a low growth world are we going to keep having Wars are we going to keep having like you know democracies fracture a little bit um I I think like more growth and continual better lives uh and new technology is is important and we're going to get this one right I I really do I feel more confident about that every year some that's a really future positive uh note to bring this discussion to a close we are um out of time now so if you could all thank our fantastic panelists and the people always is gonna say a few words foreign let me just say what a fascinating afternoon it has been full of fascinating insights and some diverse thoughts and some things that I think we will all want to reflect upon not least the desirability of going to Mars or not um it's been fantastic for me I think just to pick out one thing that I would reflect on when Sam was chatting with azim and talking about the future of work um I think he emphasized the hope that in future um Not only would AI Empower people in many areas of the world but it would also improve and enhance the status of human to human interactions and the work that involve that for me that really resonated because as a university as a community that's of course one of the things we do every day whether it's Educators inspiring the Next Generation innovators creating the next Innovation researchers thinking about everything from what it means to be human to how to create the next technology and of course it's what we've been doing here this afternoon coming together as a community interacting talking with each other and discussing these really important issues and I'm really grateful to Sam for the leadership he's showing in that space by coming here today we're going to close now and you're going to leave by that door over there please where the sunlight and no no Mars Attacks is going to happen out there the sunlight is waiting for you um but in closing may I just say um how proud UCL is to have hosted this open AI event with Sam Altman azimazar may I thank all of you for your attention your questions our wonderful panelists azim but of course at Samuel thank you very much thank you [Music] [Applause]

Channel: UCL Minds Lunch Hour Lectures

Views: 3,733

Keywords:

Id: FhivpGsMJTw

Length: 25min 51sec (1551 seconds)

Published: Thu Aug 03 2023