The Remarkable Opportunities & Challenges of AI

[Music] good afternoon and asalamu alaykum um we have exactly 13 minutes to tear through uh some of the most important and interesting topics today uh so a big boo to the program director for giving us so little time but we'll try so um everybody knows uh David seagull for your work in the finan cial uh industry you run two Sigma one of the most successful um and biggest hedge funds in the world but less known is the fact that you studied uh you got a PhD at MIT you studied AI before AI was known as AI so that makes you a clever dude in addition to being a wealthy dude so you know what you're talking about so I want to start off by asking you um with all this AI the some have said some people have said that the progress of AI has stall as as such do you agree and is there too much myth in AI uh well I don't agree with that I think we're in perhaps the most exciting moment uh for the field ever and that the progress that we've seen over the last several years is likely to continue but not necessarily in the ways that people expect uh there was a really uh a very typical technological cycle that has occurred it some people call it an scurve where you have a a relatively slow buildup and then as a technology gets perfected you have a rapid increase of in its capabilities and then people start to uh run out of tricks with the initial set of ideas and it slows down people refer to that as an s-curve um the uh initial successes for example in large language models have been going on for a long time it's not some people think it just came out of thin air but it's been a build up for more than a decade probably two decades and uh today uh we've been seeing this rapid increase now without new innovation we're going to end up plateauing but so much money is being invested in the field today that it's very likely that there will be more uh innovation uh perhaps not in expected ways so I believe that this uh increase in capability will continue for quite a long time right Maan you do your work studying about AI uh is there too much myth in the in AI how much of it is just advances in Computing and how much of it is actual artificial intelligence yeah I I I think as if today the biggest progress has been in the scaling of it and and like David was saying the amount of compute that was just just thrown away on that made that all the algorithms that we had built already in the '90s because a lot of the algorithms that we're using today are mostly built in the' 9s there's been very few progress in terms of innovation but it's the scaling and the fact that it's massively deployed around the world so in terms of might I wouldn't say no because of course llms are used on a daily manner people are seeing it and how it works and so of course the progress is there what's more interesting is what's going to come next and what's going to come next is is of course all these AIA Regents that we're seeing and a lot of potential also in coming into more sophisticated algorithms at this stage where progress is needed just because of the energy consumption that we're seeing which with these with these big I would say data centers that we're building so in terms of Might no it's continuing of course the scaling made all the algorithms and techniques that we have built and we were discussing just previously with David around basically the fact that people were doing optimization had built already a lot of the algorithms that we're doing today and and it's going on and and of course the fact that that it's the common use has made it of course so successful right but r r sorry no I was going to add one point which is there uh to your question there's uh the the raw advances in AI may not be what we care about most in this room what we really care about is the applications of the technology that we currently have and the technology has advanced so quickly that even without uh any great breakthroughs in the field the impact that what we have on business will continue to grow as people figure out how best to deploy it right so regardless of whether it's myth or reality uh the fact is that it is a buzzword everybody is embracing it everybody's jumping on it so the question now is for governments that are pursuing AI strategies what should they be watching out for should every government with the resources uh be pursuing their own llm well one thing that the governments should watch out for is uh overregulation uh I think that there's been a lot of talk about AI being extremely dangerous and uh uh you know needed you know this is related to concern that AGI is coming and super intelligence is coming and then this technology could somehow have a very dark side and be very harmful to society which is leading to the thing thinking that lots of Regulation uh will be needed to to protect ourselves from it uh I am fearful that this uh line of thinking could actually end up holding the field back and uh so I think really um my number one worry would be people being too worried about the technology H yeah I I I don't I I mean I think of course each country should have a national strategy on AI and I think think this is mandatary just that like we had a few years ago National strategies on digitalization because before coming to AI you need already to have a digital infrastructure now what happens is not every country should have a foundational model built and that would not make sense I think there's going to be consolidation of the market because of the amount of money that is required to build these super great foundational models which are now above 200 300 million dollar each uh the the one of Elon Musk xai we still don't know Croc 3 may happen around $1 billion dollars to build that gives you a b amount so not all the countries have this capability however at the national level there should be a good understanding of how you equip with yourself with this open source Clos Source models and here in the uee we've been playing a very important role in being an open source Hub and helping and supporting these different countries because on the national strategies front of a lot of countries in Africa you will start building your llm based on some open source one and trying to tweak them in the right Manner and understanding also as David was saying the verticals where you using it the verticals are not the same in each country some are more energy oriented others are Agriculture and that understanding is very important right so the the the first principle is to get a strategy get your head right about what you want to do with AI before you start deciding whether you want to invest in it what about Investors now if investors in in this room are all presented with uh business plan business model one way or the other what should they watch out for I think the the most important thing to watch out for is exaggeration of capabilities um the advances that we've seen today are are quite remarkable but um they also have limitations everyone is aware of for a large language models hallucinations imperfect results but I think even more concerning to me is that people are in a way I divide AI into two categories the generative AI which is like chat GPT and the predictive AI which is like AI being used to forecast the weather uh predictive AI is something that people need to uh recognize uh can be uh confusing many things that we think we can predict are actually not really predictable to begin with and if it's not predictable AI won't help you predict it and so what I see occurring uh in many uh Industries and verticals are frankly people trying to predict things that are inherently unpredictable AI isn't going to help you with that H what about startups yeah um there are a lot of startups right now involved in various forms of AI and you mentioned earlier on that a consolidation is probably going to come in the industry when that happens what should they do yeah so coming back already to your question where you should invest I think you should invest in liquid cooling companies because the amount of of of of data centers which are being built uh I think the stock market will raise quite rapidly for liquid cooling companies because that's what you need but I mean that's a side effect if you want to invest now in terms of startups I think exactly when you're start investing you need to be sure that the teams that you're having are not just Tech oriented but they're also have a good understanding of the verticals where they're applying it because a lot of of these companies are coming with Solutions and then they want to change basic Bally the way doctors work other places work and they end up there is a misalignment in terms of their technology and the changes in the startup front you're not going to wait like 10 years before your technology gets in you need to have a very good understanding of how it applies let me just give you an example all these tools in AI are not 100% accurate there's hallucination so if you start saying I'm going to replace all procurement in companies well when you go talking to the procurement guys they tell you look well uh it takes me a bit of time to check the documents generated by your gen and checking your documents with just this 200 Pages takes me more time than doing than writing it myself because we're so efficient and so there's a lack of understanding of application because you start building a business plan saying I'm going to replace I'm going to replace at the end it doesn't work like that so make sure that you know the people you have have a very good understanding of the verticals where they're working in the startup right okay we don't have very much time left so I'm going to uh play a quick game with you um I'll give you some few questions give me yes or no true or false bullet points okay first question uh some internet has splintered some people have argued will AI Splinter yes not yeah yes or no the answer is yes would you like more just yes yeah yes okay um next question uh can AI achieve reasoning Chain of Thought and self-a awareness uh to some extent it already has yes yes definitely there's a uh quen deep seek have shown already some capabilities but then you need to define a bit what you mean by reasoning but at least at the level that the kid could see it I yes okay um what is the biggest misperception today about AI that AI is actually thinking it is not it is just Computing it's no one actually knows exactly what it it's doing that is very interesting little uh problem I I think the blackbox approach meaning the fact that we don't understand the kind of results it's giving and the fact that it's very probabilistic meaning that if you interact you get one result another person which interacts with the same question get another result so I think this is a bit of a risk for a lot of people second risk that I see also is the fear related to losing your job that's is AI thinking is AI AI thinking no what is your go-to AI tool every day I use the various uh large language models uh instead of using search engines I use Falcon because I'm very proud of it here in the UAE but also perplexity as a tool right yeah um AI is making its presence felt in education uh what should uh universities and education institutions do to guard against this education should embrace AI it should allow students to use it and should incorporate uh AI tools into all aspects of teaching I think that uh pushing back against it uh will will will just uh create uh lost educational opportunities I think they had to they have to build and that's what we've been doing at Khalifa already trying to build AI tutor platforms because tutorship is very important and AI is very linked to that second what they have to watch is rethink how they do their curriculum by making people who are more what I call problem makers than problem solvers because if you use these tools you realize that all the results you get are based on the quality of your question the quality of your question gives you the quality of your answer when you use an nlm well thank you gentlemen for playing we are right out of time so thank you for being a great panel [Music] h