Major AI News #28 - Sam Altman's Surprising Comment, Gemini Ultra Release Date, AGI "Achieved?"

From AGI being achieved to the Google Gemini Pro release date, there is a lot of things to discuss this week in artificial intelligence. So without wasting any more time, let's jump right into what makes this week so special. So one of the things that I did really want to talk about was apparently tech companies are slashing thousands of jobs as they pivot towards AI. Now there has been some news recently stating that technology companies are investing heavily in AI, which of course they are, and some workers are already paying the price. And it says that SAP, the latest big tech player to cut jobs as it pours money into AI. With the German software giant announcing this week that it's investing more than 2 billion to integrate artificial intelligence into its business as a part of what's called the transformation program. And at the same time, the company said Tuesday its plans to restructure 8,000 roles. Some of the workers will be laid off and while others are going to be trained to work with AI. Now, I do want to say that I'm going to read this architecture and then I'm going to give you guys my points on this because I actually don't agree and I'm going to explain to you guys why. So essentially it does say here that SAP is not an outlier. In the little more than a year since generative AI tools like ChatGPT, based on so called large language model technologies have been available to the public, a large number of tech companies have announced plans to plunge into AI and job cuts often follow. And it says that there is no doubt forthcoming work is going to show that coding and many engineering type occupations have very high exposure levels to AI. So we should take them at face value. So here's the thing that you do need to know. Whilst there are some layoffs right now, especially in the tech industry and you know, as a whole there are layoffs going on, this actually isn't due to AI. Now you might be thinking, what do you mean? AI is this new crazy technology that is just going to be replacing tons of people. That is true, but right now AI doesn't really have the ability to completely replace an entire worker. And before you say what are you talking about, AI can be used for this, it can be used for that. You have to understand in certain industries, in order to completely replace a human, you need to be able to verify and fact check certain things. And remember, one of the problems that we do have with AI is of course hallucination. So although this article states, and I'm going to show you guys with some more Information why this isn't as bad as it seems that tech companies are slashing thousands of jobs as they pivot towards AI. This is. While it is definitely a scary title to read, this isn't something as big as it is because of this image right here. So what we're currently looking at, and this is just a little bit of economics, but essentially, essentially what you're looking at is the, you know, interest rate. And essentially what this is, is this shows us the rate at which you can borrow money. So if you don't know how major tech companies works, is that usually what they'll do is, you know, depending on the interest rates. And this is just kind of how the economic environment, you know, was. Of course, after COVID 19, essentially what happened was the economy was kind of screwed, essentially. Remember the Federal Reserve printed loads and loads of money. And then essentially what happened after that was a lot of these giant tech companies, they were able to borrow money at essentially a 0% interest rate or really, really low interest rate. So they had a lot of money to go around. They're hiring people on giant salaries. And of course, because of that, now that the interest rate is hiking again and the Federal Reserves have hiked rates again, essentially that means that borrowing money is now harder and companies are now being forced to adopt a more profit focused approach, which essentially means that people are not focusing on growth anymore and they're focusing on actually prioritizing profits, which means laying off people in high salaried positions that they don't necessarily need. And if you guys don't believe what I'm saying, take a look at this. Okay, so this is an article where essentially it talks about meta's shares surging 20% and they absolutely just killed it in the stock market. And Meta is a huge tech company. Recently this company has been focusing on AI. And recently what they did is they decided that in 2023 that was going to be the year of efficiency. So it says. Some investors have questioned the company's hefty investment into the Metaverse, which is costing the company billions of dollars a quarter. And sales in Meta's Reality Labs passed 1 billion in the fourth quarter. But Virtual Reality unit recorded $4.6.5 billion in losses. Now here's the kicker, okay, it says that Meta has been in deep cost cutting mode, including cutting over 20,000 jobs over the last year or so in response to changes, guess what, in the economic Environment and Apple's iOS update and rising interest rates. So remember guys, these cuts that we're seeing now aren't directly related to AI. Whilst yes, some of these are going to be related to AI. That is of course if you're in the writing sector, for example, like Duolingo, which we did see those layoffs there were definitely due to that. But here you can see that of course cutting 20,000 jobs over the last year. These are changes due to the economy guys. And remember, these changes are completely normal. So whilst yes, in the future we're probably going to be seeing some tech companies layoff people and some other industries lay off people, right now we aren't seeing some huge, huge layoffs. Now I do want to see some of your comments down below in the comment section because I do want to see what you guys want to say about this. But this was definitely something that I did want to comment on because there has been so much chatter about companies laying off thousands and thousands of people when the reality is that right now it's just that we're in a different economic environment and that leads to these companies getting rid of jobs due to this bloating nature of how companies are formed during low interest rates environments. So you don't want to be worried too much. And if there is anything that is related to industry changes, I would say stay up to date for AI agents because that would likely be when we are likely to see more layoffs due to AI because AI agents will change the game. Now this is something that I did want to cover before and it is a very controversial clip. But since it was going viral again, I decided to take a look at this clip one more time and share it with you all. So essentially what we are taking a look at here is, is a AI software that can use visual data to track the efficiency of employees. Now I wouldn't say that I'm a fan of this because tracking people like this is very, you know, it lacks the context of what exactly is happening. Like yes, this person might be on three cups, but they might be actually performing different actions. And I think that this kind of technology, whilst it can be used for good in some instances, I think when we steer or veer into the area where we start tracking people's individuals activities can become a little bit dangerous if this technology does get into the wrong hands. Now I don't need to tell you guys about all the crazy scenarios that this technology is being used in different parts of the world, but if we do have technology like this that is being rolled out in terms of, you know, certain offices and certain workplaces, I can't imagine that this is going to foster a healthy environment. Like, imagine you're at work and they're like, oh, I noticed that you've been at your desk for only 57 minutes. You're supposed to do an hour and 10 minutes or just, you know, something that's, like, so invasive. I'm not sure this would foster a good workplace environment. I mean, in terms of, you know, what I've seen this AI software usually for, which is like, tracking, you know, like animals and different kinds of, you know, things on, like a, you know, farm, or tracking multiple different things like cars and stuff like that. This kind of application, especially with some of the comments that I did see floating around, does go to show that this kind of technology isn't going to be something that seems to be widely accepted. And I can't imagine that any company that does employ this is. Is going to be one that's going to be met with some kind of backlash, because these kind of things do seem a bit over the top. It was also a rather interesting video of the Tesla Optimus Generation 2 bot walking, but I think that there was something that we did also. This was the fact that if you did slow down the video, you do see behind that there is actually a person controlling the robot with some kind of controller. So it does seem that whilst this did seem like it was autonomous at first, that that isn't the case. That nevertheless, this wasn't entirely a video that was a demo or presentation. It was simply a video of Elon Musk that just seemed to be passing by during the factory. So I do think, though, one thing that is impressive about this robot is that they've moved very, very quickly. If we do take a look at the images from 2023 all the way to where we are now, and I actually meant 2022, but to where we are now, this definitely does look like a very good walk. And I know that also that this bot does have some criticisms in terms of the walking speed.

But.

But, you know, across the board, if we do look at the entire industry, this is, I would argue, one of the faster ones, especially one of the ones that are walking, because a lot of the other ones that I've seen are actually a lot slower than this. Like they move really, really slowly. So I would argue that this is actually pretty fast in terms of walking pace for an actual robot that doesn't use wheels, if you've ever seen them. Besides agility robotics, this is definitely some nifty stuff. So I will be excited to see what. What Tesla Optimus Generation 3 does look like if we do get that sometime in 2024, because somehow the Tesla engineers seem to be moving at a rapid pace. Now this was AI news that actually kind of shook me a little bit because of the ramifications of this. Now essentially right here it says that the Mistral CEO confirms the leak of a new open source AI model nearing GPT4 performance. So it says the past few days have been a wild ride for the growing open source AI community and even by its fast moving, freewheeling standards. So, quick chronology. About January 28, a user with the handle mikudev posted a set of files and hugging face and essentially someone leaked this on 4chan. Now the reason I'm actually covering this is because, number one, this is pretty crazy that, you know, an open source LLM was leaked on 4chan. But number two, if you remember, some time ago we did a video on the very infamous Q Star and one of the claims, okay, was that there was a letter that was posted on 4chan that showed some secret details about a crazy breakthrough that potentially occurred in OpenAI's headquarters that led to some kind of super intelligence that could essentially end humanity. Now there were so many rumors floating around and I would say it is 5050 on whether or not that is true, but I would argue that the fact that this was actually posted on 4chan and it was actually real and true and, and the CEO actually did come out and confirm it and he did tweet that someone essentially leaked this and you can see right here that Arthur Mensch says an over enthusiastic employee of one of our early access customers leaked a quantized and watermarked version of an old model we trained and distributed quite openly. Quickly start working with a few selected customers. We retrained this model from llama2 and the minute we got access to our entire cluster, the pre training finished on the day of the Mistral 7 billion parameters release and we've made good praises since. Stay tuned. And I think this whole debacle of leaks does show that, you know, it gives even more credence to the fact that Q Star, the alleged letter, was leaked on 4chan. Of course it does have some absolutely insane claims if true, but I think it does go to show that, you know, leaks on 4chan are possible and llama 2 was initially leaked there. And if Q Star was leaked there, it leads me to believe that some of the stuff in Q Star is potentially even more ringing true. Especially when we did see some of the things coming out Reuters that told us that some internal OpenAI employees weren't too happy with whatever was being worked on there. Now here's something that didn't get the attention it deserves because there has been so much AI news every single day. But there was something coming out of Google Research that is pretty incredible, okay? And this was something that we knew was going to come, but I didn't think it was going to come this quickly. And this in terms of the quality that we see. So this is mobile diffusion, rapid text to image generation on device. So essentially here they're saying that text to diffusion model models, all these text to image models that we see are of course really, really good, like stable diffusion, Dall E and Imagen. But the only thing that makes them so good is that they, you know, have billions of parameters and are consequently expensive to run, requiring powerful desktops or even servers. But essentially they were like, look, we're going to introduce a novel approach and we're going to show you guys how you can do this on device. And you can see that this is rapid text to image and it's on your device. Now this is pretty groundbreaking stuff because if we are able to get this kind of technology on our devices without the need for a server, without the need for it to be computationally expensive, this is going to truly change the game. Because number one, these companies are going to be much more profitable, especially the smaller ones trying to start out in terms of their innovative research. And number two, it means we're going to be getting easier access to these models and we're not going to have to wait as long because everything is going to be more efficient and more fast. And not only that, it's also going to be on device. So I think this transition from having everything on a server to moving everything to being streamlined on device is going to be a complete game changer for how we use these AI models. Because what's crazy about this as well is that if we do actually also look at the quality of some of these images, although it is isn't groundbreaking in terms of the quality, I think when we take the fact that this is on device and it's generated in just a couple of seconds, I think that is something that we do need to be proud of. And of course, as always with research and as the technology is going to improve in the future, it's likely that we're going to have some really, really high quality text to image generators that are going to be on our phones. And honestly, that is going to be something that really does change the game because I mean, are people still going to be using Mid Journey if you have Mid Journey level quality and in an app that's literally just on device and on your phone? So that will be something that is quite interesting because of course they actually also talk about, you know, the latency here. And of course, I do think that, you know, 512 by 512 is a realistically small image, but I do think that this kind of thing being able to be done in half a second is absolutely incredible. And of course, the fact that this is simply done on device in half a second does show us that we are moving in the right direction. So it will be interesting to see if Midjourney does release some kind of app or if they do keep everything on their streamlined Discord interface. So then, of course, we do have some Gemini Ultra potential release dates. Now, of course, with the rumors and with leaks, we don't really know if these are going to be true, but it's always fascinating to cover to see if some of this stuff does come out to be true, because even recently with the chat GPT leaks, a lot of those did turn out to be true and they were actually discovered quite like with what we're seeing here. So you can see that it says Gemini Ultra is slated for a February 7th release, and it shows that there is a version called Gemini Advanced. And. And that seems to make sense because Google has said that the advanced version of Gemini Pro will be released later. And you can see here that the date there is of course, February 7th. And we can also see that the previous release date was of course, for the very infamous Imagen 2 and the entire update that Bard got to Google. So I wouldn't be surprised if Gemini Ultra was released on February 7, because that would mark a very new era in terms of Google's venture into the LLM space, because it now means that hopefully, based on the benchmarks that we saw, that they've dethroned OpenAI and this would lead to a cascade of events, including the release potentially of ChatGPT's infamous GPT 4.5 or perhaps even the announcement of GPT 5 and what we can expect from that, because there has been a lot of chatter from Zammel1 about GPT5, and we do know that that technically we're in a release, and we do know that technically we are in a race. So if Google does release Gemini Ultra, that would be a very, very big thing to compete with, because we already know that they did just release Google Gemini Pro, and it is currently sitting at number two on the leaderboards. So I honestly would hope that they do release this on February 7th because, like, some of the comments would say, you know, that this would of course push OpenAI to do their release. But let me know what you think about this post. Let me know if you think that, you know, what they are going to release this or if you think there's still going to be more delays. Also, this really cool feature, which is called diffuse to choose, and apparently this is going to be a feature that is coming to Amazon. And I do think that this is going to be the future of online shopping because whilst everyone, you know, does shopping in many different ways and many different styles, and although like clothing is of course something that is subjective based on the styles and based on how you kind of, you know, want it to fit you and how things kind of look, I think if we take a look at this right here, where you can literally just take a picture of your house and then of course you can simply input that image into the environment, I think that something like that is going to give you much more freedom and creativity to sort of envision how things look. Because a lot of the times what we do have is we do have a look for things, especially when shopping online, it can be pretty hard to see exactly how it would look on us and how it's going to fit us. And I think something like this is going to be really, really effective for foreseeing very, very quickly getting that quick feedback. And I think as long as, you know, these systems can run pretty quickly and efficiently, this is definitely going to be the future. Because a lot of the times if I could just upload a few pictures of myself, you know, or I was buying a sofa, or if I was buying like a wallpaper or if I was buying like, you know, like a painting or something like that, and then I could see how that poster, slash, painting slash, you know, table looks in the specified area that was going to be at. That would be pretty much a game changer because lot of times you take a picture, you try and envision it, you think maybe, I don't know, you like car an image sometimes. So, yeah, this is definitely going to be something that does really improve the online shopping experience. And of course Amazon is always trying to improve the shopping experience because they of course are the E commerce giant. So wouldn't be surprised to see them implementing something like this. And it actually does use stable diffusion to do this. So if this does roll out, I will be excited to see if this does happen and I'm pretty sure I did see a few AI app stores actually start to implement this. So it will be interesting to see if, if this does roll out to the masses and is widely adopted. Essentially we had an update to the Microsoft Bing version. Now something that's important to know is that whilst they do run on the GPT4 version, they actually do have different, slightly different, finer tuned versions. So essentially you can see here that this is the question why is Meta stock rising? And you can see Meta stock is rising because the company reported a strong financial performance in the fourth quarter, yada yada yada. And then the same question is asked and then we can see that there is a completely different answer which is really, really good. So this person says that was there any improvements that shipped to balanced mode? It seems that the answers are now bigger and better and ironically this one was even better than Bing's Creative mode. And he essentially responded saying that we are continually improving Microsoft Copilot testing a new version of the base model in a flight and fun fact, its internal name is Eucalion, the son of the Prometheus in Greek mythology. So I'm excited to see that if this is just like some kind of early test what the 4.5 is going to be or the next update is going to be. But either way I still think that it's always important to pay attention to the Bing updates because whilst they are different fine tuned versions of GPT4, I do think that, you know, these kind of versions are still pretty useful especially for browsing the Internet and we could potentially have some game changing updates here in the future. So yeah, I think that that comment there just goes to show how powerful these AI systems are going to be because Sam Altman of course knows a lot about arguably the most powerful AI systems in the world based on what we do know. And I think the fact that he's saying that does show us the fact that AI agents and potentially GPT 5 or GPT 4.5 will host some really, really tremendous capabilities and if they do come to market. So I'm thinking that this is really, really game changing because if we can get AI systems that allow us to get the amount of value out of a single system that you would get out of hiring 500,000 employees or even a thousand employees, this is going to be serious prosperity for the economy and also serious long term effects because if you no longer need to hire a thousand people to run a company, that then means that there is of course a thousand less jobs in demand. So that then means that of course there are a thousand less job opportunities. So I mean, this is of course something that is really fascinating. I think this interview comment does actually show us just how powerful these systems are going to be because he did actually say it would have been unimaginable without AI. And of course now it's going to happen. So I'm going to be excited to see what things people are going to be building in the future when they've got like, you know, the reasoning abilities and all of these, essentially like, you know, the power to build your own company with just a few AI systems. I think that that is truly going to change our society and I'll definitely be watching it unfold. Then, of course, we did have this comment from Lian Lecun actually about open source AI, and I think it's a very fascinating comment because he does talk about how essentially having open source AI is very good for the field in terms of accelerating the entire field because it allows everyone to openly share and collaborate ideas and whenever a breakthrough is made, people are able to use it. Now I do think that open source is good, but I do wish that this was the case with some of the models that we do have, like GPT4 and what these larger companies are working on. Of course, unfortunately it is a race. So Google DeepMind can't share what they're working on with, with OpenAI, of course, and the wider community because there is of course that race that, you know, in terms of essentially being the top model or AI system. So I mean, it will be interesting to see how that does change in the future if these open source models do manage to catch up. Because of course Meta's plan is to build AGI and source version of AGI. But I think his comments here are really important because of course there are certain things like regulation coming into the.

Play when something is viewed as a, as a basic common platform ends up being open source, right? That's true for the software infrastructure of the Internet. And the reason is that it's, it's, it's safer, it progresses faster. Like the way the reason why we've seen such a fast progress in AI over the last decade or so is because people practice open research. It's because as soon as you get a result, you put it on archive and you open source your code and everybody can reuse it. And we have common software platforms like Pytorch that everybody uses and so it makes the entire field accelerate because you have a lot of people working on it. That are self selected. That's why. That's the magic of open research and open source. And so the way to keep that progress fast is to keep the research open. So if you legislate open source out of existence because of fears, you slow down progress and something is viewed as a basic common platform.

Then there was this, and I really didn't want to cover this, but you know, at the end of the day it's AI news. So someone claimed to have achieved AGI proof of the first artificial general intelligence. And long story short, I am claiming that this is just, you know, with 100% pretty much skepticism here that these guys haven't provided any proof. So I really don't get what the point of this post is. I mean, it just seems like an elaborate troll. Although it is pretty fascinating to read and it's pretty intriguing, I'm pretty sure if someone really did create the first AGI. And you know, forgive me if I'm completely wrong, but you know, I just don't see any credence to this at all because there isn't a full research paper, it isn't peer reviewed, but still, it did gain some traction. But of course, to no surprise, the majority of the traction it did gain was negative. But like I said, I'll leave a link for you because of course it keeps you up to date with exactly what's going on.