Sam Altman New statement On GPT-5 Is Surprising!

the the most important thing and this sounds like the most boring obvious Tri thing I can say but I think it's actually much deeper than it sounds the most important thing is that the models are just going to get smarter generally across the board there will be a lot of other things too which we can talk about but if you think about what happened from GPT 3 to 3.5 to 4 it just got smarter and you could use it for all these things it got a little more robust so Sam mman recently gave a rather fascinating talk and it wasn't that long but it was really interesting and in this video I think there are two key statements that you guys need to pay attention to because there are a few slides that really do show how crazy things are so one of the slides you do get to see is this clip and I'm going to show you guys this clip right here so basically what this guy is talking about is he's talking about the compute in terms of the future models and the image shows a graph which looks like it's a giant giant increase in terms of the capabilities and from what we've seen pre obviously this looks like it's going to be pretty insane with regards to the kinds of you know improvements that we're going to see so what I'm going to be showing you guys now is of course the clip of the guy basically stating that you know we are pretty much nowhere near a plateau which is definitely definitely not surprising due to the major things that I've discussed previously but after that I'm going to show you guys a clip where Sam Alman actually speaks about what's coming next with GPT 5 cuz this was at the Microsoft conference software but we are riding like a fundamental wave in in um in the development of this AI platform where if you just sort of look at compute over time like how much uh how much uh GPU Cycles or accelerator cycles that we're using to train the very biggest models in the world since about 2012 uh like that rate of increase in compute when applied to training has been increasing exponentially and we are nowhere near the point of diminishing marginal Returns on how powerful we can make AI models as we increase the scale of compute so yeah you can see that this is where you know he literally said that there is you know nowhere near any kind of you know slowdown in terms of what we're going to be able to achieve and of course one of the crazy things that we can see um is we can also see this image here and he literally states that the next sample is going to be pretty incredible so we can see that presumably although there aren't any numbers on the chart we can see that gbt 5 according to this chart in fact I probably should ask like an AI how much big this looks the only problem with this image is that we don't know what kind of chart this is like we don't know if this is like an exponentially increasing chart we don't know if it's a log chart but uh yeah so I can't say if it's going to be three times two times like we really don't know but what we can see is that the chart does look like it's going up in that you know exponential way so this is definitely pretty pretty crazy that this was you know even something that was talked about so I think that this is H something that's pretty pretty cool and I think it's uh you know a key indicator on where we're going to go next so now I'm going to show you guys the clip where Sam M actually talks about how the future models are going to be and what they're going to be like and obviously great to see you but developers have been such a core part of what's been happening this last year and a half um there's millions of people building on the platform what people are doing is totally amazing and the speed of adoption and talent and figuring out what to build with all of this over what has really not been very long like when we put gpt3 out in the API uh some people thought it was cool but it was narrow where the and people have done with GPT and seeing now what's happing with GPT even though it's new and hasn't been out that long uh is quite remarkable I've never seen a technology get adopted so quickly in such a meaningful way uh the what people are building how people are finding out how to do things that we never even thought of possible which is why it's always great to have an API uh that's been very cool to see yeah and I I think you know what you just said is like one of the most important points to me like there there's a version of AI that could have existed that is uh you know like a bunch of smart people like building uh you know things at extraordinary scale and then just building it into a bunch of products where everybody gets to passively use them like the the really brilliant thing that you all have done is like taken the exact same set of things and like decided to make it available to like any developer who's able to sign up for an API key yeah we we try to be really thoughtful about what makes a good API for this there's going to be all kinds of ways people can use this but the more this can just be a layer that gets built into every product every Service uh the better and we've tried to make it such that if you want to add intelligence to whatever you doing uh any product any service we make that very easy yeah and like again I think the progress has been stunning so you know I I think you know the the setup for uh like introducing you onto the stage here was uh I saw that big blue whale yeah like you know you're you're making good use of the whale siiz computer right now and so like I without like getting too specific which we can't be obviously like what are the category of things that people should be expecting over the next you know K months so the the most important thing and this sounds like the most boring obvious Tri thing I can say but I think it's actually much deeper than it sounds the most important thing is that the models are just going to get smarter generally across the board there will be a lot of other things too which we can talk about but if you think about what happened from GPT 3 to 3.5 to 4 it just got smarter and you could use it for all these things it got a little more robust it got much safer uh both because the model got smarter and we put much more work into building the safety tools around it um it got more useful but the underlying capability this amazing emergent property of like we actually are seeming to increase the general capability of the model across the board that's going to keep happening and the the jump that we have seen in the utility that a model can deliver with each of those half step jumps and smartness it's quite significant each time so as we think about the next model and the next one and the incredible things that developers are going to build with that I think that's the most important thing to keep in mind uh also speed and cost really matter to us so with GPT 40 we were able to bring the price down by half and double the the speed um new modalities really matter uh voice mode has been actually a genuine surprise for me in how much I like the new voice mode and I when people will start integrating that I think that'll matter but but it's the overall intelligence that'll be coming that I think matters the most so you for a while now have been one of the most successful startup investors uh in the world um and like now you are one of the most successful uh CEOs of one of the most important companies in the world so you've got a room full of developers here you know like I think there are 5,000 people in the room and there about 200,000 people online right now um what's your advice to them is like they think about how to spend their precious time given what's happening in the world like what what's your advice two two things number one uh this is probably the most exciting time to be doing building a product doing a startup whatever it is uh that we have seen at least since the mobile boom um and probably I would say since the internet and maybe even bigger than that we don't know yet um but the the big opportunities the big you know the ability to sort of build something new and really kind of like change the landscape that comes at the platform shift times and we haven't had a platform shift in a while and this looks like it's really truly a platform shift uh and so my biggest piece of advice is like this is a special time and take advantage of it this is like not the time to delay what you were planning to do or wait for the next thing like this is a special moment uh in a few years where a lot of stuff is going to happen and a lot of like great new things are going to get going um the second thing also about platform shifts is when the mobile phone Revolution started or really got going like 2008 2009 you would see people say um we're a mobile company you know we're have a mobile app and then only a few years later no one said they were a mobile company because it was like table Stakes an amazing new technology which I would bias but we put AI in that category uh it doesn't get you out of the hard work of building a great product or a great company or a great service um you still have to do it AI alone is a new enabler but it does not automatically break the rules of business and so you can use this as like a new thing to do but you still have to figure out how you're going to build enduring value and whatever you're doing um and it's easy to lose sight of that in the excitement of the Gold Rush yeah so one last thing before we let you go so you know you and I and like members of your team and members of the Microsoft team have been doing really an extraordinary volume of work over the past uh year and a half two years thinking about safe deployment of an awful lot of AI capability like everything from you know apis and developer tools to end products uh and you know I think we you know have accumulated a really interesting volume of experience like experience is sort of hard to get if you're not doing deployments at this scale um so I I you know and I think you just mentioned something that's like really really interesting like part of uh you know part of the interesting and surprising progression capabilities of these models means that they're more useful in like helping to like make AI systems safer so I I don't know whether you had some thoughts you wanted to share there as well you know when we first developed this technology we spent a lot of time talking about all right we've made this thing it's cool are we ever going to be able to get it to an acceptable level of robustness and safety and now we kind of take that for granted with gp4 um you know if you use it it's far from perfect we have more work to do but it is generally considered robust enough and safe enough for a wide variety of uses and that took an enormous amount of work across both teams and fundamental research like when we started this we're like we've got this thing we've got this language model it looks like kind of impressive and kind of not and even then how are we going to like get it aligned and um what it what does it me you know what is it going to take to be able to deploy it the number of different teams we've had to build up uh to go from research and creation of the model to Safety Systems to figuring out policy to how we do the monitoring um that's a huge amount of work but it's it's necessary uh to be able to deploy these and use them like you know when you take a medicine you want to know it's going to be safe when you use an AI model you want to know it's going to be robust and behave the way you want and have been super proud of the work that teams have done together and I think it's amazing how fast this much work has happened and that we can all now use this and say oh yeah it basically it basically works as the models get more powerful there will be many new things we have to figure out as we move towards AGI um the level of complexity and I think the new research that it'll take will increase I'm sure we'll do that together but uh we view this as a gate on being able to put these things out into the world which we really want to do yeah it's definitely table sakes so thank you so much for uh for being with us here today like I really appreciate your time uh it's awesome to hear from you awesome so with that being said what do you guys think about the future of these models one of the clear thing that Sam Alman has constantly reiterated is that these models are going to consistently get smarter and you know with the recent statement that you know we're nowhere near the kind of you know cutoff point to where these models are going to be getting their Improvement I mean right now I think the hype is definitely real and I am really really excited to see what some of the most notable AI critics do state if these models do have a massive jumping capabilities because some people are stating that this isn't going to happen at all for example one notable AI critic Gary Marcus has said dear Microsoft CTO Kevin Scott want to put money on what what you just said $100,000 and he's clearly betting that these models just won't get any better at all so I think this is going to be super interesting to see in the future if he was someone that was right or was someone that was just literally you know I guess you could say being a AI Doomer in terms of you know trying to state that these AI abilities just won't get any better I'm someone that believes that they will get better just based on the you know the things that we've fundamentally seen and just because openi may have not published any research okay about what they're getting up to I think it's clear that you know the fact that these other companies that have had such a long time to catch up in fact not even that long and they've managed to catch up just so us that in the next Cycles open a are clearly clearly clearly going to come up with something incredible but I do think that this Microsoft build has released some very very interesting products and some very very new interesting things that we all going to be using in the next future especially with the new computers if you didn't see that don't forget to click the link down in the description

Channel: TheAIGRID

Views: 64,583

Keywords:

Id: VbF8t66SjOo

Length: 12min 55sec (775 seconds)

Published: Tue May 21 2024