GPT-6 News, Suno And Udio Get SUED, Sora Get Worse, Claude Launches GPTS and More AI Breakthroughs..

So welcome back to AI News. And there is a very interesting piece of news that has the AI community divided. One of the most interesting pieces of news that is remarkable because it will navigate, slash, set the precedent for how future cases are of course done, turned in. And this is the US record labels suing music generators and of course basically stating that they have infringed upon others copyright. So you can see right here, it says US record labels sue AI music generators Suno and Yuzio for copyright infringement. The Recording Industry association of America and the record labels alleged that the leading AI music generators trained on their artists work without permission. So essentially, if you aren't familiar with these platforms, you can. Suno AI and Yuhudio AI are two very, very comprehensive platforms that allow you to basically create any kind of music track that you want. And I'm not gonna lie to you guys, it's not something that is kind of good that's somewhat okay. This stuff is really, really, really good. And I know I said really there, but I cannot emphasize how good this stuff is. Like when I was creating content on Asuno and Yudio for various different projects, I. I was so surprised at the coherence, the quality and at how consistent the soundtracks sounded. It was honestly mind blowing. So I can completely understand why these record labels are suing them. Because if we look at the future and we try to predict where things might go, AI music generators are definitely going to impact that industry in a remarkable fashion. I mean, I can't imagine how many people are going to be using these instead of, you know, the traditional tools that the music industry obviously benefits from. So it says. The music industry has officially declared war on SUNO and Udio, the two most prominent AI music generators. A group of music labels including Universal Music Group, Warner Music Group and Sony Music Group has filed lawsuits in U.S. federal court on Monday alleging copyright infringement on a massive scale. The plaintiffs seek damages of up to $150,000 per work infringe. And the lawsuit against Suno is filed in Massachusetts, while the case against Yu Dio's parent company, Uncharted Inc. Was filed in New York. Now what's interesting is that they state that Suno and Yu Dio did not immediately respond to a request to comment. But earlier today they actually tweeted something that was remarkably interesting. They stated today, we'd like to share some thoughts on AI and the future of music. In the past two years, AI has become a powerful tool for creative expression across many media, from text to images to film and now music at Yudio Our mission is to empower artists of all kinds to create extraordinary music. In our young life as a company, we sat in the studios of some of the world's greatest musicians, workshopped lyrics with up and coming songwriters, and watched as millions of users created extraordinary new music ranging from the funny to the profound. We have heard from a talented musician who, after losing the ability to use his hands, is now making music again. Producers have sampled AI generated tracks to create hicksons like bbl, Drizzy and everyday. Music lovers have used the technology to express the gamut of human emotions from love to sorrow to joy. Groundbreaking technology entail change and uncertainty. Let us offer some insight into how our technology works. And this is where they're basically stating that, look, we didn't copyright anything, we trained it legally. Just hear us out. So they state that generative AI models, including our music model, learn from examples. Just as students listen to music and study scores, our model has listened and learned from a large collection of recorded music. The goal of model training is to develop an understanding of musical ideas, the basic building blocks of musical expression that are owned by no one. Our system is explicitly designed to create music reflecting new musical ideas and, and we are completely and uninterested in reproducing content in our training set and in fact have implemented and continue to refine state of the art filters to ensure our model does not reproduce copyrighted works or artist voices. We stand behind our technology and believe that generative AI will become a mainstay of modern society. So I'm not going to read the rest of this because it's just some technical jargon and honestly, it was probably written by ChatGPT. But the point here is, is that what we have here is a situation on our hand. On one side, this is completely understandable from the music industry. If you are someone in that industry, I can promise you guys, as someone who's worked with people in that industry, taking the time to write a song, to produce the song, to, you know, sing rap or whatever kind of, you know, vocals you're adding to the song, those things take an incredibly long amount of time. And I honestly, genuinely respect those who do music or pursue it as a hobby or passion, because those things take a remarkable amount of time, that is admirable to say the least. Now, I can completely understand that if you do music or if you're in the music industry and your artists have been creating content for such a long time and companies like Yudio and Suno just essentially sample all of your music tracks and download them, train the AI systems on them and then try to create a platform on them, and then that creates a very, very hostile environment. Because the problem here is that the people's music that was trained on, they weren't actually paid anything. Music is a thing where if your music track is used, usually you'll get royalties. Or the owners of these media companies, these massive conglomerates, they essentially ensure that the royalties are paid out to these musicians. Now, I know people are either going to be on the side of the people where it's like it's AI, everything is going to be new and generative, but. But I do think that in this scenario, the musicians definitely have a profound argument. If you do actually take someone's work, train a system of it, and then that same entire system is actually slowly being used to kill that industry. I personally don't think this is a creative use of a technology. Whilst, yes, it is completely fun, and whilst, yes, this is really cool to use, I personally am leaning more on the side of the musicians here because I've been in the firsthand area where musicians are really struggling to make ends meet and they really put their heart and soul into the music tracks. I think the best kind of solution that would happen here is that these new generative AI platforms, whatever platforms they may be, they should be paying some kind of royalty to whatever artist or whatever person that they trained off the most. I think it would be disingenuous to state that these platforms would even exist without the creators that actually created the initial bash of content, for example, the images. You. You can have some AI platforms where you literally have asked them to imitate the style of a certain person and it's able to do it very effectively. And how would you feel if an AI system was replicating your unique art style and you weren't being credited for that at all? Whilst many people were paying $20 a month and these companies were making millions. I'm not someone that's hating the technology. I'm just stating that if we are to actually progress as one, then these music companies most certainly need to actually pay out a royalty of their profits to these people musicians, because honestly, it is their work and they should be paid for it. Whilst, yes, some people might say that this is decelerationist or whatever, I do think that if it was you in their shoes, you definitely would sympathize with them. And I don't think it's that hard to actually pay the musicians their fair share. I didn't really want to talk about this for too long. But I do think that the future is going to be interesting. Either we're going to have a situation where these music companies lose these lawsuits and these generative AI companies are allowed to just do what they're doing. And this creates a very, very interesting environment because we are entering an area where the precedent is now that, look, I can train off all the work on the Internet, and if I create a generative AI tool, then I can completely go ahead with that, which is a very dangerous precedent. Or we have the area where these platforms are forced to pay out and then unfortunately, they can't pay out because they can't afford them because generative AI is very expensive. And then of course, they go bankrupt. Either way, it's kind of fascinating to see where this trend will head. And I'll definitely be paying attention to this. In addition to AI generated artworks, Toys R Us has released the first OpenAI Sora generated brand commercial. This remarks, well, not remarks, marks the first generative AI work that is essentially commercial. So what we can see here is Sora being used in the production of something here. And this is essentially a kind of commercial that is for Toys R Us, which is a kids toy company. And you can see that they have used the Sora platform in order to generate generative visual video. Now, one of the most remarkable things about this is that I thought that this trailer was kind of interesting, slash, kind of cool. But this has actually been met with quite a lot of backlash. If you check the comments, the retweets, the quote tweets, and all the comments surrounding this, they are remarkably, remarkably negative. Take a look. And this proves my point, and this is something that I've spoken about in my private groups before, is the fact that AI generated content is something that is relevant, Remarkably disgusted and shunned in terms of human creativity circles. A lot of the processes that we see in the creative areas actually provide a negative response in humans. And I've spoken about this in my post AGI Community, because I spoke about navigating the nuances in AI generated content versus using AI generated stuff to advance your businesses. And you can see here, this is an example of where natural not to use NAI generated stuff. So essentially you can see right here, it says that, you know what I realized about AI generated images and your marketing. It sends out the message that you've got no budget. It's the digital equivalent of wearing an obviously fake Chanel bag. Your whole brand immediately appears feeble and impoverished. The point I'm trying to make right here is that if you see the public reaction from this, this is worse than if they just made a bad ad. I can promise you now that this trend is going to continue for the foreseeable future until humans develop a more positive relationship with AI generated stuff. And I can guarantee you that's not going to happen until we figure out a future where humans can coexist with AI without them being replaced to some extent. You can see here, this person states, nobody's going to shop at Toys R Us anymore. This ad is just lame and awful. And if you thought this one just has one retweet and two likes, take a look at this further. You can see that this person is here. So saying that let's vandalize this business, yada yada yada. But take a look at some of the other ones that are more popular. This person said Toys R Lot, Toys R us released an AI commercial and it Effing sucks. That is one person. Another person says Congratulations OpenAI. You've ruined toys R Us. First the bankrupt and now shut down and now you. And then someone else here says we're morally bankrupt, artistically bankrupt, and literally bankrupt. You can see here that a remarkable number of people have liked and retweeted this one that says there's something so cynical about creating an ad about the limitless imagination of a child and then rendering it with soulless AI slot. You can clearly see that there is a huge amount of negative sentiment regarding OpenAI Sora and it seems to continue to be echoed across many different social channels. You can see someone also says shame on Toys R Us for contributing to this madness and and also for making such a blank looking commercial. The point here is that if we look at even the comments underneath, people are simply hating this so much. And the point is is that right now it is less effective to use AI in any marketing media channel to where humans can immediately detect it, because the moment a human detects that you are using AI generated stuff, it is immediately just shunned away and just taken as a lesser value piece of content. Like I said before, this is something that of course is quite good for us humans because it means we're going to be in the loop for a while longer. But as I said before, humans like humans. I can argue that this trailer would have probably went viral if it actually was a single child in this trailer and they actually used CGI with the help of human artists to make this. And like I said before, it wouldn't be surprising to me if in the future there's going to be some kind of human LED studios, human LED writing areas where it is more cost effective to use humans due to the nature and the sentiment surrounding human made work rather than AI generated content, even if it is made faster and cheaper. Now in other news, Claude actually got a remarkable update. They actually got an update that actually allows you to organize chats into shareable products. You can see that right here. Basically what they've added is they've added something that's quite similar to OpenAI's custom GPTs. You can see here that you can create a project. You can see exactly what you're basically trying to achieve. Once you click create your project, you can then include all of your project knowledge, it loads it. And then of course you can then use this to integrate different pieces of your information and then you can talk with Claude to be able to get custom responses. This is something that is remarkably effective because now you can use all of your information in one specific area. You can have consistent chats with the Claude Mon model and it's something that is just a lot more efficient than going back and forth and continuing to load different chats every single time. So you can see right here, you can have multiple chats with this. And this is basically like custom GPTs for Claude 3.5, something that anthropic have probably been working on for quite some time. And I'm glad that they've actually managed to ship this because this is something that many users have requested. Let me, let me know if you're going to be using this. I do have a tutorial on the new Art Facts feature. It is available on the channel currently. There was also an important piece of information from Microsoft CEO Mustafa Suleiman and he actually states something about GPT6. You might think it's clickbait, but it absolutely wasn't. He says it won't be until GPT6, in 2 years time that AI models will be able to follow instructions and take consistent action. That is a remarkable piece of information that. But it does line up with the timeline that we saw in the trademarks. Take a look.

It's still pretty hard to get these models to follow instructions with subtlety and nuance over extended periods of time. I think that they can do it, you know, and there's a lot of cherry picked examples that are impressive, you know, on Twitter and stuff like that. But to really get it, to consistently do it in novel environments is pretty hard. And I think that it's going to be not one, but Two orders of magnitude more computation of training the models. So not GPT5 but more like GPT6 scale models. So I think we're talking about two years before we have systems that can really take action.

So the reason that this is a remarkable piece of information is because this gives us an updated timeline on the features that we may have. You can see here he said that it's still pretty hard. Now he's not talking about the intelligence of the model, he's talking about the ability of the model to essentially do actions in a certain space and in novel environments. This doesn't mean that we might not get agents that are able to do stuff in short term environments, for example closed loops or just, you know, fill out a table or really easy stuff like that. What he does talk about is the fact that these agents are going to be basically built into GPT6 if they can get things working by then, because it's a lot longer and a lot more harder in order to effectively be able to do those things with, with the current models. So essentially the reason I think this is also pretty interesting is because when you start to hear them talk about GPT6 scale models, when we don't even have access to GPT5 nor any real piece of information, it can show us that, look, they've obviously worked on GPT5. It's obviously probably nearing the release date. And I do predict that this release date is likely to be around just after November to towards January 2025. So that is going to be the Goldilocks zone of when we can very likely expect GPT5 to potentially surface. So with that being said, I do think it is rather fascinating that AI agents are rather harder than many people have suspected. And this is of course something that I also did suspect myself when I looked at the current landscape of AI agents. One of the things that I've realized is that a lot of these browser agents that you may have seen, for example, what you're looking at is adept AI labs. And essentially they are basically a company that is trying to create these AI agents and they've been rather silent on the Internet for quite some time now. You can see that some of their posts about, you know, the model that they've been able to do things with. For example, you can see Find me a refrigerator for under $1,000. This last informative post was released around 2022, which was at the around the time of GPT 3.5. This was their Action Transformer model and they taught it to use a bunch of Software tools. Basically, what this is is you can see, find me a house in Houston that works for a family of four, and you can see that the model is going to go ahead and use its vision model, going to go ahead and search for whatever home it might find or whatever. And it's basically looking through the Internet. Now, the reason that this is pretty crazy is that this was their last major update, which was around two years ago. And to not have any kind of Act 2, Act 1.5, any newer models, it shows that this problem is very, very hard. Now, I would say that once this model does become available to the public, it most certainly will change the huge landscape of different AI tools, and we're likely to see a complete overhaul of what's available online and in terms of the available AI systems. And I think this is going to be the final stage to where the average person realizes that, whoa, AI is truly, truly here. Now, if we're continuing to talk about frontier models and companies that are training bigger and bigger, more comprehensive models, there was actually a rather interesting piece of news that gives us an insight to the future of AI models. You can see here, Dario Amodei, the CEO of Anthropic, the lab that created Claude 3.5 Sonnet, which is a recently, very, very impressive model, actually speaks about these scaling laws in a new interview with Time. He talks about the scaling laws that say as you train systems with more computing power and more data, they become predictably more capable or more powerful. And of course, they say, what capabilities have you seen in systems that you're not able to release yet? They state that we released our last model relatively recently. So it's not like there's a year of unreleased secrets. And to the extent that they are, I'm not going to go into them. But we do know enough to say that progress is continuing and we don't see any evidence that things are levelling off. Which essentially means that for those of you that think we are close to the end of the sigmoid curve, this is simply not true. It says, the reality of the world we live in is that it could stop anytime. Every time we train the new model, I look at it and I'm always wondering, I'm never sure in relief or concern, if at some point we'll see, oh, man, the model doesn't get any better. And I think if the effects of scaling did stop in some ways, that would be good for the world because it would restrain everyone at the same time. But it's not something that we get to choose. And of course, the models bring many exciting benefits. The point he's trying to say here is that, look, if these models actually did just all stop at about the next level of scale, this is not really going to be a bad thing because it kind of slows everyone down in order to build the models more safer and of course, slow down the terminal race condition. But of course, right now they don't see any evidence that things are leveling off, which means that these models are likely to increase in terms of their raw processing power, reasoning capabilities, and a sheer understanding of what we want for the next level of systems. And of course, breaking news from the chatbot arena. Claude 3.5 Sonnet has actually surpassed GPT4O in the coding arena and it currently is tied one to in the overall GPT leaderboard, where it's competing against GPT4O and other frontier models. So, basically, this is one of the most important pieces of information that I should have spoken about. This is factory AI, a machine that builds the machine. Basically what they want to do is build autonomous software agents, basically autonomous swes. And you can see that they have a new benchmark which takes the lead in terms of the software engineering questions. So. So they also have a new fundraising round which is $15 million Series A LED by Sequoia Capital. And this is fascinating because it shows us that Devon Cognition Labs aren't the only software engineers operating in this space. You can see that they speak about how their droids are the most advanced coding AI systems that exist, thanks to their breakthrough in research, orchestration, retrieval and planning, which is absolutely incredible. We can see here that compared to Amazon, Devin and, you know, the app map, native GPT4O, the factory code droid outperforms them all by 5%, which is a very large margin. And comparing the fact that we had Devon not too long ago and some other, you know, open source systems, this is something that is remarkable in terms of how much progress they've managed to make. Now, I do wonder how long it's going to take us to get to, you know, 90%, 95 or even to 99% on the SWE bench. And the website is honestly rather fascinating because it talks about the autonomy of software engineering, how they're bringing autonomy. So I do wonder how this industry is going to change. We do know that there are, you know, a lot of competing companies trying to bring autonomous software engineers into the space, and I do wonder how that will change the job landscape. And that is something that I'm going to be diving into more on my school community.