🚀 Add to Chrome – It’s Free - YouTube Summarizer
Category: N/A
No summary available.
00:00
the term AGI is everywhere now what are you excited about in 2025 AGI think if Sam were sitting on the seat he would tell you Aji is closer than most think all this geopolitical race to build AI before the other guy the hopium war the
00:15
idea of artificial general intelligence or AGI so the AGI it's AGI AGI AGI AGI how far out is Agi okay but seriously what the is Agi well imagine asking your phone not just to set a reminder of to plan an entire vacation for you to
00:32
figure out where to go and then to even haggle with the hotel to make sure that you get the best rate by telling it you're going to get a five-star review now imagine that kind of intelligence in the hands of world leaders Geniuses all sorts of other people who have way better things to do than plan their
00:48
vacation that's the kind of leap we're talking about when we talk about AGI artificial general intelligence open AI Sam Alman and anthropics Dario emodi both think that we're going to be seeing AGI maybe within as soon as the next couple years but defining what AGI is is
01:05
kind of a tricky concept that a lot of experts can't really agree upon it's part sci-fi dream part incoming reality and in this video we're going to talk about what AGI might actually mean some of the history behind it and then what it means for you let's get into it in the simplest sense AGI or
01:21
artificial general intelligence is an AI that can understand learn and process stuff over a wide range of subjects a lot like we can Wikipedia defines AGI as a type of artificial intelligence that matches or surpasses human cognitive capabilities across a wide range of
01:39
cognitive tasks but that sounds totally nebulous what does that mean any task a human can does that mean it's going to write a novel for me can it write a good novel for me can it invent a new technology can it tell a bad joke as bad as I can because I can tell some pretty
01:55
bad jokes but you get the idea here this broad definition really does make AGI a hard thing to pin down but essentially it's about a machine that doesn't follow scripts and can adapt and innovate on its own today's AI systems even the most advanced ones like open ai's 01 preview are still considered narrow AI they are
02:12
very good at specific tasks whether it's generating text from a question you ask it translating languages or providing really limited creative options but they're still operating within pretty strict boundaries of their training data and their programming and it's pushing the definition of what an nrow AI is I
02:28
mean we've all seen the surprising stuff that these modern AI tools can do but they ultimately lack the broader understanding and even kind of self-awareness of what we might think of as a true AGI system AGI or some people might refer to it as powerful AI or even
02:44
more confusingly as advanced AI should be able to or could be able to navigate entirely new problem sets transfer their learning across different contexts and truly think and innovate in ways that would dynamically adjust based on their goals like in the example I gave up top
02:59
it wouldn't just provide you possible answers from the internet it would actually be able to go book a trip for you do all that stuff negotiate and to have truly good AGI you would have to be impressed with the results when they came back because if it goes and does that and gives you something you're not happy with then it's not operating the
03:15
way that human might it's just following a system and AGI system is going to do the job as well or better than you and now the briefest of histories of AGI the idea of AGI really dates back to the 19 1950s if you can believe it with
03:31
Alan Turing and the Turing test Turing posed his most famous question can machines think in his Landmark paper Computing machinery and intelligence this was really the first real world exploration of computer intelligence and the Turing test which came out of this
03:46
paper was the idea that you would speak to a person and you would speak to a machine anonymously and the test was can you tell if that person is the machine or if that person is a human in the beginning stages it was super easy but this question has stayed with us and
04:01
since then AGI has had a roller coaster of a history if you ever want to do a deep dive on this it is super fascinating Big Dreams huge setbacks this is like a a story for the ages the Dartmouth summer research project which again was in 1956 was this kind of
04:17
beginning stage for optimism in this space and then there was the cold winters of the 70s and the 80s 1970s and ' 80s which really LED people down the path to think that this may not happen for hundreds of years and then in the late '90s and early 2000s IBM deep blue
04:32
was able to beat Gary Kasparov the human chess champion and even Watson when it won on Jeopardy both of those were big steps in the world of of AI but they were narrow AI they were really good at one specific thing whether it was chess or it was uh learning all the trivia
04:49
questions in the world and how to answer them quickly that's a very narrow thing and if you had asked Watson or deep blue to do some of the stuff that chat GPT can do today it would have no idea how to do that but in the tends the AGI dream got a new lease on life with deep learning and neural networks I'm not going to spend a lot of time going into
05:05
this here they are complicated ideas but basically it was a new way of looking at approaching intelligence the idea breaks down to essentially letting the machines learn on their own and get better at things over time and in 2015 deep Minds Alpha go beat the world's best go
05:20
champion and this was seen as a pretty big deal because go is not like chess necessarily where it's a strict rule set there's a lot more creativity involved and people started thinking some something interesting might be happening here and of course now we are sitting with open AI anthropic Google meta all
05:36
of these companies kind of rushing through the llm revolution which is a thing that we're all living through right now and we're on the cusp of what a lot of people think is Agi or at least coming very soon but the funniest thing to me about this is that nobody can really agree on what this means even
05:52
experts have totally different ideas of what AGI should look like some think it needs to focus on intellectual tasks and be able to do all the stuff that a that a human academic might be able to do whereas others think the creative side of the humans is what you have to be able to see that if it can't do that then it's not really AGI all these
06:09
different people having thoughts on this make it start to feel like a moving Target I do appreciate Sam Alman from opening eyes very straightforward definition take a listen to this the term AGI is thrown around a lot how would you define AGI and how do you think you'll know I think there's a lot of valid definitions to this but for me
06:24
AGI is basically the equivalent of a median human that you could hire as a coor worker and then they could say do anything that you'd be happy with a remote coworker doing just behind a computer I think I can get behind that I understand what that means and that's a pretty good Baseline to start from it's
06:40
a computer that can figure out on its own how to do the stuff that we do in our everyday lives this is the most important definition here if you really want to dig into the idea it's not about a a Super Genius which ultimately will be great if we get Super Genius AIS that's amazing but it's really about can
06:56
the AI do what you do in your everyday life the next question here really is why are all these companies rushing to make this happen do we really want this do we need something that can do all the stuff we do well I think the answer is probably yes and I think ultimately it's because we can which seems kind of crazy
07:13
but I think we do want this now do I want a version of myself who can go out and do the the normal stuff I mean it reminds me sometimes of the Calvin duplicator and the Calvin and Hobs Comics uh I hope that person does it well but if that person can do it well sure I would love a bunch of versions of me that I can send out into the world to
07:30
do the stuff so that I have more free time for myself but then I also think it's important is like I'm just an average person I am like you know my brain is relatively average what if you could duplicate the world's smartest people and have them be working on really difficult projects whether it's
07:46
preventing cancer or it's eradicating diseases all the sorts of things that we want to happen in this world AGI could help bring that to fruition Dario Modi was just on Lex freedman's podcast and talking about this specific idea th large organizations and systems there
08:02
end up being a few people or a few new ideas who kind of cause things to go in a different direction they would have before who who kind of disproportionately affect the the trajectory you think about the health world there's like you know trillions of dollars to pay out Medicare and you know
08:19
other health insurance and then the NIH is is 100 billion and then if I think of like the the few things that have really revolutionized anything it could be encapsulated in a small small fraction of that and so when I think of like where will AI have an impact I'm like can AI turn that small fraction into a
08:35
much larger fraction and raise its quality I made a whole video about Dario's vision for AGI which he calls powerful Ai and I'll link to that in the show notes down below but you can kind of see at scale how this could get super powerful and of course it is not just a tech race it is a geopolitical one we
08:53
always love to bring politics into this story but it's true right now more than ever the race to AI is not not just about getting a smart person into your life that can do things for you it's about how whoever controls AGI could dominate the entire economy or entire
09:08
Industries especially the idea that AI could start helping improve AI itself if you have essentially infinite amount of smart people to bring on to any task you can see how that gets to be a pretty big deal in terms of global power so the urgency of all this talk about Ai and
09:23
where it's coming from is This Global push for this competitive race to get to this thing that could be completely transformative for the world economy think about it as a leap from early computers to the internet but then multiplied exponentially that changed our entire world and that we're looking
09:38
at something that's significantly bigger and it's right on the cusp of happening and then of course you hear a lot of this also in the terms of AGI is there's a scary side here too right it's like what if this AGI does not align with who we are and it starts to get more powerful than us or who has control of
09:54
this AGI who's the company or the government that ends up winning this do we align with their values do they feel like something we care about are we in the same boat as them and there's always going to be somebody the other side of that fence and on top of all that there's a whole another idea of
10:10
something called an ASI or an artificial super intelligence in which the computer just gets massively smarter than even an AGI where we don't have any idea what that's going to behave like but that is an entirely different video and I will save it for next time so again in the
10:26
simplest form an AGI is simply an AI that can can do what we do a human on a regular basis and can do it well and for now that might be stuck in things like computers or documents or spreadsheets the sort of stuff that humans interact with computers but with the robotics Revolution happening that is going to
10:41
cross over to the real world very soon hopefully you know a little bit more about what AGI is now or do you um maybe we need an AGI to figure this out as well I'll see youall next time