AI NEWS: Even MORE People Leave OpenAI, ChatGpts New Feature, Future Model Sizes, AI To Cure Cancer?

So let's take a look at some of the new stories in AI that you of course did miss. Now, one of the biggest stories that I just immediately want to dive into is the story from Tim Brooks. And I think this story is arguably one of the most important when it comes to discussing AI because it shows us exactly what is going on at OpenAI in terms of wider aspects of the company with regards to certain individuals leaving. For example, one of the things that we recently just saw here is the fact that he said Tim Brooks. And if you don't know who Tim Brooks is, Tim Brooks is the SORA research lead at OpenAI and he said that he's going to be leaving OpenAI to join Google DeepMind to work on video generation and world simulators. Can't wait to collaborate with such a talented team. I had an amazing two years at OpenAI making Sora, thanks to all the passionate and kind people I worked with. Excited for the next chapter. I think this is one of the things that is one of the most surprising considering the fact that when we actually look at how many people have left OpenAI within the last year or so, it's a notable amount of individuals. And one thing that I keep reiterating is that it isn't just employees that are leaving the company. Unfortunately for Sam Altman, it seems to be the fact that the people that are leaving OpenAI are actually key individuals. We've seen that Ilya Sutskever, the lead of the Super Alignment team has left, Jan like as well and Leopold Aschbrenner, key individuals of key key teams have left. We've also seen Greg Brockman go on leave and many other individuals such as VPs of product have also left the company. I think honestly this is starting to get really concerning. Before I would have initially said that yes, OpenAI can maintain its lead, but I do think that a company is nothing without its talent. And when you do have talented individuals continuing to leave your company in favor for other companies that are also competing in the same exact space, I think it's going to be very difficult in order to catch up. You have to understand that SORA was the first in its kind technology. It was a huge leap over what we previously had. And if the trend continues of key individuals leaving OpenAI to go to Google to go to Anthropic, I'm not sure how much longer OpenAI can maintain their lead. I think we've certainly seen a early indications of this when we saw Claude 3.5 Sonic be released. And of course when we saw Runway release their video model, which far surpassed what was publicly available from Sora. Now, some people have thought that this might have been an issue to where Sora hasn't been released. And the problem is that if you are working on a product and it's just being used internally, there isn't much motivation to continue working there, which is understandable. But I think that it's because OpenAI is in a really tumultuous phase. On one hand, you have individuals that are pushing fast ahead, individuals like Sam Altman that are trying to get products like Sora out. But of course we do have other individuals in the company that says that things require a lot more safety testing. I do think now though, that whatever's going on at OpenAI, we are likely to get a lot more products quicker because. Because previously individuals who were working at the company have now left and it seems like it's only mainly the accelerationists left at the company. I mean, if you fire your Super Alignment team and if the CTO leaves, who was worried about the speed of things, I think this is going to change things. So with this being said, I'm not entirely sure what's going to happen with Sora. Now some people are stating that there wasn't enough compute for it to be used, but I do think it is rather interesting. And that brings me onto the second story. I know I did a video on this just yesterday, but if you do remember Meta's moviegen, one of the things that I didn't talk about was how good this system is, in the sense that this system is remarkably expensive to run. One of the things that we're consistently seeing is that the very best AI systems are currently really expensive and really time consuming to to run. Which means that, yes, it seems that the scaling laws for things like video simulators and images is certainly true. You're going to need more data and more compute to make these models more effective. Now, if you don't know what Meta's moviegen is, if you didn't watch the video, this is basically Meta's video generator. It is currently the best in the world. And there aren't any plans to open source this, but we have a variety of features. You can, you can see, you can edit different videos, you can edit them with a simple text prompt and the lighting. The effectiveness of the video does look certainly well. The thing that I do want to talk about though is that Meta actually recently made a statement regarding this video model, giving us more details on how it works. And I think it gives us some information to why Sora wasn't yet released to the public. That could potentially clear the air.

We can see.

Chris Cox says that the feedback we've heard from filmmakers and video creators was to prioritize ease of editing, but even more the ability to generate videos with a specified character slash image, which these models now faithfully achieved. Now here's the kicker. They state that we aren't ready to release this as a product anytime soon. It's still expensive and generation time is too long. But we wanted to share where we are since the results are getting quite impressive. So I think this gives us an insight to what these companies are currently dealing with, which is why I do believe that many companies are still increasing their computing capabilities. One of the things we know here is that it's still expensive, which means that it's probably not as efficient as it could be. One of the things that continues to happen with AI technologies is that when it is released, what companies just focus on is they just focus on making the technology work. Then once the technology works and is reliable and is effective, then they can focus on making it cheaper. It seems that the same method is used in Meta's moviegen, which is why they're stating that right now they've managed to make it work, but it still is currently expensive. Another issue that they are facing is that the generation time is too long. If you do want to release some kind of products to consumers, the generation time is something that needs to be fast enough to the point where it is cost effective for an individual to use that over someone else. I'm not entirely sure how expensive and how long the generation time is, but a article which did disclose that how long Sora's time generation was said that it was about 10 to 15 minutes for a one minute clip. Now, in the grand scheme of things, I don't think that it is that long. But the problem with these clips is that since it's generative AI, sometimes you're going to have to regenerate clips time and time again. And if you have a generation tool that you have to regenerate every time and you have like a minute, the time being too long can quickly add up to hours at the office trying to regenerate clips, which isn't necessarily a time saver. One of the things that does make AI effective and usable is the fact that it saves individuals time and it saves them money. Considering that right now it's probably expensive and this generation is too long, they're probably going to continue to delay this software until it becomes cost effective to run at scale. Maybe this is the reason that they were facing atsorat. This is the issue that led them to not release it just yet, because whilst it probably is higher consistency than what's available now, it just isn't commercially viable. And that's okay. But we do know that that will change likely in the future once we get massive data centers. So for those of you who are wondering if Meta is going to be releasing this anytime soon, the results are in and that is not going to be the case whatsoever. And if you remember how I spoke about how only the acceleration list are left at OpenAI, this article from the Hollywood Reporter where basically they state what the heck is going on at OpenAI, they dive into some key things which are rather fascinating and one of the things that I think is most important is how OpenAI is set to change after Mira Marathi's departure. One of the things that was really fascinating was that this person said after the November drama, Marathi decided to stay at the company, in part to try and slow down Altman and Greg Brockman's acceleration efforts from within. So it seems here that when miramoratti decided to leave a couple of days ago, of course, which was quite the surprise to Sam Altman and other individuals in the leadership roles at OpenAI, it seems that it was due to OpenAI releasing products too early. Like I said, it says here that it's unclear what tipped Moratti over the edge, but the release of 01 last month may have contributed to her decision. So I think what we have here is a situation where Marathi wanted these products to be more safe than they currently are. Maybe there are certain safety issues that we may not know, because I do think the O1 model is vastly underestimated and vastly underrated in terms of its capabilities. But I think what we're now seeing as well is, is that the average person doesn't need a PhD level AI in their pocket, and I don't think that will change anytime in the future. So overall it might just be the early release of 01 that seemed to tip Marathi over the edge and cause her to release. So this is something that is rather fascinating because I'm not sure if OpenAI can stomach any more people leaving. And I think you have to understand that these AI companies, they're not like traditional companies in the sense that you can just replace people easily. There's a huge reason why there are wars being fought over individuals. If you don't remember, when OpenAI was founded, there was literally A war between Google and OpenAI of who was going to get Eliaskova. These individuals that work at these companies are really integral to the success. Of course you can hire another engineer, but certain people are just irreplaceable. Which is why when I'm seeing the lead of Sora, leave the cto, leave Ilya Sutka, believe it leads me to believe that OpenAI might not maintain its lead past 2026. Now, of course, you do have the fact that AI is exponential, which means that it could surpass what we know to be true if we do manage to get systems that can work within OpenAI and they manage to scale that effectively and get to AGI first. But I think that we're not that close to it yet and we still need remarkable talent to achieve certain breakthroughs. Now, one of the things that I wasn't able to cover was introducing Canvas, which is your coding surface in ChatGPT. This is something that is remarkably effective at helping you to code and of course helping you to work within ChatGPT on a day to day basis. If you're doing long pieces of text or just have a project that requires continuous edits in a nice user interface. Take a look.

I'm Katya from OpenAI and today we're going to see how I can use Canvas to write code. So I have this 3D model of the ISS and I want to display it in a 3D scene, but I have the latitude and longitude and I want to translate that into XYZ coordinates. So let's see how we can use ChatGPT to help us with that. Okay, so here it is. It's writing code, but it's in Python and my app is in JavaScript. So I could just go ahead and hit this port to a language button here and select JavaScript. And here we go. It's just translating the code into JavaScript. And this looks great. Actually. The only thing is that I'd like the position to be relative to Earth. So let me just add this, the earth position, just adding coordinates. And then I can just select this line and say I want this to be relative to Earth.

Okay.

So it's scanning the code and then changing anything that needs to be updated. And here we are, we have the relative position. And this looks great. Just going to hit the code review button here to get some suggestions on what could be improved. Okay, so it's giving me advice. Okay, cool. And I can actually accept the suggestions or not. Great, that's perfect. Okay, so that was Canvas. We're going to be adding some more features over time. I hope you liked it.

Now let's take a look at today's sponsor, Notion, who just launched their incredible new AI. The New Notion AI is a real time saver, letting you quickly find information and organize your work effortlessly without needing to jump between apps. For example, I was recently working on a video about people leaving OpenAI and I had all this information scattered in my notes. Instead of manually sorting through everything, I simply asked the New Notion AI to create a table of everyone who had left and we, and within seconds it gave me a neatly organized table. From there, I could easily drop it into my document or even screenshot it to include in the video. The New Notion AI makes it that easy to streamline your work and boost your productivity. Now, another example where we can see that Notion's AI is really helpful is the fact that it has access to this entire page. So if I need to brainstorm the introduction to the video about OpenAI's future, I can simply ask it to write the intro for me based on all the data in this one page, and then it can easily go ahead and create that. Overall, I think this is the kind of AI implementation that we are striving for, one that is really easy and efficient to use and actually saves you time. Now, if you want to try it out, don't forget to click the pinned comment or the link in the description.

Now, do you remember a few days ago when we had this news article from Reuters? Now, I want you to pay attention here because there's a bit of a debate going on about whether or not this statement is true. So the article reads that OpenAI has asked investors to avoid five AI startups, including Sutskivers Superintelligence Inc. So if you aren't familiar, OpenAI recently had a round where they raised $6 billion at, I think an $156 billion valuation, which is pretty hefty considering that they were a nonprofit not a long time ago. Now, the craziest thing here is that they also wanted investors to refrain from funding five companies they perceive as close competitors. Sources toward Reuters now, the craziest thing about this is that Reuters is a really reliable source of information. Any information that I've gotten from Reuters, I've never had it be wrong. So we know that this source is quite true. Now, I want you to remember what I just said, because the list of companies, of course include Elon Musk's X AI. That makes sense, since Sam Altman and Elon Musk are currently feuding if you aren't familiar, Elon Musk actually filed a lawsuit a few months ago stating that OpenAI have gone away from their original mission and Elon Musk, I'm not sure if he's seeking damages, but there is some kind of issues there. Of course Ilya Sutskeva's company which is pursuing superintelligence, of course two AI application firms including Perplexity and enterprise search firm Clean and of course the last one being Anthropic. Now the craziest thing about this is that whilst Reuters is stating that this is true, recently someone who works at OpenAI has come out and said that this is false. I'm not sure who is in the right tier because Reuters is usually a reliable source, but maybe OpenAI are just doing damage control. You can see that OpenAI CFO Sarah Fryer says that the total market cap of AI or the AI industry is only 1% of what it will become and denies that OPERAI is asking investors not to invest in other AI companies. Take a look.

Want to make sure the ecosystem continues to develop from here. Look, we are in such the early innings. I was listening to someone talk about, you know, how much of what we now think of as the market cap of the Internet existed about two years after Netscape launched and it was about 1%. So if we think about two years after chat CBT launches, are we just at 1% of the market caps that will be created it in terms of this round, look, we want to make sure our investors are really focused on open air, but of course they're going to go out there and invest in the ecosystem and this round is no different from any other round and we want to make sure the ecosystem continues to develop from here. Look, we are in such the early innings now.

The CFO also said some things that were quite interesting. I don't think this was anything that we didn't already know, but the CFO Sarah Fryer said that their next AI model will be an order of magnitude bigger than GPT4 and future models will grow at a similar rate, requiring capital investment meet their really big aspirations. I think this just shows that whilst many individuals think that the scaling laws won't continue, the AI industry is going to continue to expand in order to accommodate all of those products and services such as image generation, video generation, text generation and of course for the future, agent generation. In terms of having those agents actually work and be effective, I think they're going to need a lot more scale than they currently do have.

Now normal operating expenses of a more traditional company. So to that end we do want to make sure we're being creative in where we can go to tap capital. To your point, sometimes that's public markets, sometimes that's debt markets, sometimes it's private project finance, structured finance. There's a lot of things that I need to get my kind of fingers into as I look forward over the next several quarters. But I think there is no denying that you are. We're on a scaling law right now where orders of magnitude matter. The next model is going to be an order of magnitude bigger and the next one on and on. And so that does make it very capital intensive. It's a, it's a really different take technology cycle than if you think about the last cycle which was much more bits and bytes, a lot cheaper. This is much more like the telephone being brought, dropping cables, electricity going up the railways. I mean I think you're in much more of that sort of capital intensive cycle. And that means for us too, we're going to have to really be careful and smart about how we raise money.

Now if we are talking about agents and the future.

In a conversation with Mark Cuban he.

Discussed how agents are a feature that and not a product. Because as AI gets smarter, it will be able to create its own AI agents.

We're seeing the gold rush right now where everybody calls everything AI, particularly with agents. And I think you can put all these vertical agents together to do all these different things, but agents are just going to be a feature, not a product. Because inherently in AI as it advances and gets smarter, then it's going to be able to create its own agents for its users and go forward from there. So I've been really hesitant now because you know, you're not going to invest in the foundational models. I mean through a fund I have part of OpenAI but and some others. But it's. That's just so expensive you don't know who the winners are going to be. But yet everything that, everything that happens is going to be a derivative of what's your.

What's your.

Now one of the things that was actually breaking news was that Elon Musk actually is sticking to his plan of open sourcing AI models. This is actually rather surprising because I think open source is something that isn't that favorable to a company that's trying to compete with other companies foundation models. I mean it kind of does make sense considering that they're not releasing the current model that they have, but rather releasing the model that is going to be essentially discontinued, he states. It's worth noting that XAI has been and will open source its models, including the weights and everything. As we create the next version, we will open source the prior version as we did with Grok 1 when Grok 2 was released. This information is considerably surprising, but I think it makes sense considering the fact that Elon Musk was a huge proponent of open source AI in the beginning. Now I'm actually really excited for Grok 3 because Elon Musk previously has said that that this is going to be largely one of the biggest AI models in the future and probably the best AI model in the world. And I think if a small lab is able to actually pull this off and surpass OpenAI, Google, Claude and Meta, I think that's going to largely be one of the most surprising things to happen. So it will be interesting and I will be paying attention to very intently to any release from Elon Musk in terms of his company's releases. And of course, if you haven't been paying attention, we are only days away from when Elon Musk will introduce Robo Taxi and things that operate in the real world. This is from an interview on cnnbc. And if you haven't been paying attention to it, there is a Tesla Robo Taxi event that is going to be on the 10th of October. It's probably going to be a very large day for AI, considering the fact that autonomous vehicles are going to be a key part of that future. Many people like myself are quite excited for this day as we do know that it is going to be a pivotal moment.

Now, recently Sam Altman in an interview.

Said something that I personally think is fascinating because I think it gives an insight to where he views AI is heading and. And it always seems crazy until it's done. And I used to think that statements like these were rather insane, incredible. But now that I see the trajectory that AI is on, now that I've recently seen the scaling laws change once again, leaving us to just apply more compute to generate more results. I think this statement here that he's suggesting that AI will someday be able to cure cancer and in collaborations with human collaboration, seems that it might actually be not that far away, which is.

Of course going to be one of.

The best byproducts of this AI stuff.

Yes, there is the job loss, the.

Widening disparity between the rich and the poor, but better healthcare overall seems like a net positive gain that's going to happen regardless.

I had a conversation last night, it's not quite about Creativity bombing. Just say it anyway, because it's just stuck in my mind so much with a scientist. And he was like, you know, started using these tools, been studying the field. I am convinced that someday I will be able to ask an AI to cure cancer. And yeah, you know, maybe it will email me a few times and say, I need you to run this experiment in the wet lab and tell me the results and then go off and think for a few more weeks and.

Then.

You know, ask me one more experiment and come back and whatever. And at the end, like, cancer will be cured or this particular cancer, some specific thing, this cancer will be cured. And I was like, yeah, I think that'll happen too. You know, that's actually like one of the examples I often give. And he's like, it makes me so sad. I was like. He was like, look, of course, as like, you know, member of Team Human, I'm thrilled, thrilled for that. But like, your thing is making AGI. My thing is curing cancer. And like, I wanted to be the one to do it. And he's like, I've really been wrestling with this because again, as a member of Team Human, the faster the better. Like, go at this from my own personal satisfaction. Like, what I want is a tool that tells me, go investigate this, but doesn't quite give me the answer. And then I still get to do it. And you know, notwithstanding everything we said earlier, which I, I also still believe in that. Like, that stuck with me. I think it'll be awesome when the AI can do my job better, but I've been doing it for a while and, you know, I got to have the fun part.

Now I'm gonna leave you guys with.

This, which is rather thought provoking. And this was what I spoke about.

When I said that AI images are.

Going to fundamentally change everything. And not because people won't realize that they are AI, but even if an image is AI generated, it can still change your perception of someone. As previously spoke about how the AI generated image of Mark Zuckerberg actually boosted his public image. But this was something that most people didn't acknowledge because they think, okay, it's AI, it's fake, people don't care. But this right here is a clear example, is that even if an image is AI generated, people are still going to be affected. For example, we can see that there is this image here. I'm not going to get into any politics at all.

I'm purely just focusing on AI generated.

Images and how it changes humans mind. You can see that it says this picture has been seared into my mind.

My head hurts. The community notes clearly says that this.

Is an AI generated image. It is not real. But the user says that I don't care if it's AI, it is still accurate. Someone also says CN says this is AI and in this case I don't care. We should all look out for our own Americans before the rest of the world, yada yada yada. This person clearly doesn't care. So that the image being portrayed is. Is AI generated. Which is why I say that AI.

Generated images are going to affect you.

Whether or not you think about them. Once again, another person stating that yes, it's AI generated but still an accurate description. And someone said that it doesn't matter. This is seared into my mind forever. I think this is probably the most accurate depiction of why this is such an issue. Many times we might see images that are striking or are thought provoking and those are the images that we remember. And unfortunately, the image, it doesn't have to be an accurate representation of what has happened in the real world. It just has to be enough to invoke some kind of human emotion. And in the case of these AI generated images, it seems that that is currently happening. I'm not paying attention to politics, so I don't want this to become a political video. But, but this is going to be something that. I genuinely don't know how we combat this. Because even if the AI generated images say that this is AI, people are still reinforcing their belief systems. Let me know what you think about that and if you enjoyed the video, I'll see you in the next one.