GPT-5 News, GPT-4 IMPROVES 900%, AI Crashes The Stockmarket ,Microsoft Announcement,[AINEWS#6]

From GPT5 to AI crashing the stock market, there are many different things that occurred in the AI landscape that I can guarantee you you did not see. And simply because there's too much news that you're literally going to miss it. So with that being said, let's get into every single thing interesting that happened this week in AI. So remember how deep fakes were a huge thing in AI at the start? They still are a thing, and they're becoming even more perfected with this format. You can see that using stable diffusion control Net and epsynthesis and fusion, you're able to get a much more realistic version of a deepfake that we saw from this Reddit user. Pretty, pretty accurate, but at the same time also very, very scary because AI is moving very, very rapidly. And when it comes to facial reconstruction and deepfakes, this is always something that is going to be scary. Then we had something that was quite frightening in terms of its economic impact. This morning, an AI generated image of an explosion at the US Pentagon surfaced with multiple news sources citing it as real, and the S&P 500 fell around 30 points in minutes. So this resulted in a $500 billion market cap swing on a fake image, and it rebounded once the image was confirmed to be fake. Now, this was pretty incredible considering that a single image circulating around media outlets was able to wipe off $500 billion in value. That was all down to AI. Which just goes to show in the future, when these tools get even more sophisticated, maybe with video modalities and maybe with certain text modalities or what are we going to experience? This clip right here, I can promise you guys, was one of the most terrifying things I've ever seen in recent times for AI. And that is not an exaggeration at all. Essentially, 60 Minutes hired an ethical hacker to show how easy it was to be scammed, and they literally did it right in front of them using artificial intelligence. It was genuinely mind blowing because they used it in real time to scam one of their colleagues. And honestly, guys, you won't believe it until you see the clip. And we can't play the clip here for copyright reasons, but I can promise you, once you see this clip, you're going to really start to understand concerns around privacy and how this AI technology is really going to affect many of the verification tools that we currently use and how it's simply going to destroy them, because honestly, they were able to get the data so quickly that it truly scares me because the. The thing is here that many people Think that it's only old people that fall for scams, only the elderly and those who are retired. But this was able to work on someone who was relatively young, with a very tech savvy background. So when these kind of things become commonplace, we're going to need a radically new solution very quickly. Then we had a user on Twitter that was able to create a wearable that lets AI see the world. So you can hear using OpenAI's whisper, which was released last year. Then it can see using an image to text model. Then it can also speak using Elevenlabs human like voice model. And the punchline is that it was 100% coded by GPT4. Now the reason I actually brought this up was because I thought it was super interesting to see a project that was able to be coded 100% by GPT4. But not only was this really interesting, it also leads onto this point right here. So you can see that this is Humane's new product. Now many people may have seen this TED Talk already but, but for those of you that haven't, Humane is a company that is building artificial intelligence powered products that are essentially wearables that try to essentially assist us in our daily lives and replace the phone. So if you don't understand why this Humane company is a big deal, well around I think 90% of the people that work there are former Apple employees. And these weren't just your regular run of the mill Apple employees. These were some of the people that that part of the key changes that made Apple what it is today. So it's definitely going to be interesting to see how this company manages to use artificial intelligence to create that next gen product that we've all been looking for. Live demo at the TED Talk was honestly quite impressive, although some weren't as impressed. I was definitely impressed because I know that this is only going to get better in the future once these AI models do improve. Then we had Blockade Labs introduce their sketch to 3D World. So essentially remember before when we covered how Blockade Labs allows you to use a simple text prompt. Now they've released something essentially which gives you more control over the final driving product. So what you can do right here is you're seeing the ability to sketch out a certain environment and then of course you're able to control the specific layout of that environment with your initial sketch. Now I do think that this is really creative and really interesting because stuff like this is what's going to power the next generation of creative individuals because we are now allowing ourselves to control how those images are placed and where certain things are in relation to what we wanted before. And this kind of stuff is definitely fascinating because it's still very early in the AI race and we don't really know what's yet to come. Then we had someone taking advantage of being able to clone yourself. So an influencer launched an AI girlfriend Voicebot, and they made $72,000 in revenue in just one week. And the average user spent more than an hour chatting with her at $1 per minute. So essentially the user could talk with her. And because she was able to clone her voice with, I'm presuming, 11 labs, they essentially were able to charge these users at $1 per hour. And because these users want to interact with her, and of course you can't interact with a thousand people if you're one person, especially on a personal level. This user was able to generate $72,000 in simply one week. And it provides us with an insight to how businesses are going to change in many different industries that we really didn't think to think about. Then, of course, we had Kodi any to any generation via composable diffusion. And this is something that I'm seeing increasingly on different research papers where essentially what they're able to do now is they're able to generate any to any combinations from video, image, audio and text via composable diffusion. So essentially you can take one modality and switch it to another. And this shows a very effective method of doing that. Now, the research paper that they did write was very, very, very interesting. It honestly is quite fascinating as well, definitely worth the read. But we did also see something from Meta earlier this week where they announced something that was quite similar to this. So I do believe that in the future with regards to things like GPT5, we are going to be seeing large language models evolve strictly away from text into multimodal capabilities that involve image, video, text and of course audio. Then of course, we had incredible scientific progress in the image and text reconstruction from FMRI research in which they were able to get the ground truth videos reconstructed by simply analyzing the brain data, the FMRI data, and this kind of stuff we did hear about some previous weeks before, but they are currently working on refining this technique using AI so that you are able to get a higher quality image. And honestly, so far, by the results that we're seeing on screen, it definitely looks like this is going to be very, very, very effective. Then we had Adobe's fascinating AI tool, which really did take the industry by storm. I mean, nobody expected them to release this tool. But here we are in this compelling, new, entertaining and intriguing tool in which you are able to manipulate images, expand them and use AI to have a generative fill on the image. There are numerous applications in which this is very, very effective and we are able to see the the live demonstrations of this tool from the team at Photoshop. Now Photoshop did also talk about how they are going to be integrating Firefly into all of Adobe's products. But we didn't expect this seamless nature and for it to be this effective this early on. So this was definitely something that did shock a lot of people. When we did make the video on Firefly we didn't expect such a reception, but so far it has seen the likes of many different users use this and with very good results when combined with websites and like midjourney. Then we had Microsoft announce three insane things including an update for Bing with ChatGPT and many other integrations across Bing. Now I can't cover everything they're going to say it best so just watch this little clip to get the quick.

Rundown default and when I come in and select it I can now ask sort of real time queries. For example, let's ask what I should expect to hear about build and dot net and what you can see is the results now are more up to date. They include fresh content and they include citations. In fact, if you can see the links on that page there, you can click those and those will take you straight to a webpage that's sourced by Bing. And notice Bing can read the context of the webpage, understand those ingredients, put them into chat and then I can say hey, give me a shopping list for this and it'll automatically call the Instacart plugin. Take those ingredients that read off the page and put them into an instacart that shopping and with one click I can get those now delivered to my house. This is incredible productivity benefit for people.

And of course what we had here was something that I didn't expect to see just yet. Although we did see some people who are working on early versions of this by themselves running it locally on their own computers like Auto GPT. Here comes Microsoft introducing Windows Copilot. So remember Bing for Windows that everybody is starting to use. Well think about Bing for your computer where you can literally just ask it stuff about your computer and and it can actively run and execute different programs, commands and help you navigate your computer if you are confused. And I do think that this does have a lot of applications because many people don't know where everything is located on the computer. And of course, if you simply want to input a text prompt and get back the data that you want in real time, this is going to be something that really helps with your efficiency rather than juggling around on forums trying to find where certain files are located or having to open up certain settings programs. So Microsoft really did honestly shock everyone with all of this stuff they're going with, because it was literally just after Google's IO event in which we thought that Google had simply taken the cake and they simply showed us that Microsoft are remaining on top. Then what we had was really interesting because we had Sam Altman in an interview talk about future modalities and of course he did mention the future modalities that he's going to be focusing on, which is going to be video. So he, he talks about how text is good, but you can't achieve true AGI with text. And if you do, it's going to be relatively slower to get there compared to if you use video. So check out this clip because it is pretty interesting.

There are a lot of things about coding that I think are a particularly great modality to train these models on, but that won't be, of course, the last thing we train on. I'm very excited to see what happens when we can really do video. There's a lot of video content in the world. There's a lot of things that are, I think, much easier to learn with video than text. There's a huge debate in the field about whether a language model can get all the way to AGI. Can you represent everything that you need to know in language? Is language sufficient or do you have to have video? I personally think it's a dumb question because it probably is possible, but the fastest way to get there, the easiest way to get there, will be to have these other representations like video in these models as well. Again, like text is not the best for everything, even if it's capable of representing everything.

Then of course we had Adobe that are working on a text to 3D image composer. So essentially it's text to 3D model. And so far it does look to be really, really effective. I mean, Adobe seemed to be taking back what Midjourney already had. So thing that we did know about Midjourney was that it actually did take market share from Photoshop. But with this, it seems like Photoshop is once again back in the game. Then we had something which is a completely new tool which many are speculating that will be Bought out eventually by Photoshop. So this is drag gan. Now essentially what this is, is this allows you to manipulate images as if they're 3D models and it doesn't really require the use of Photoshop. In fact it kind of negates that because the way how it manipulates images is incredibly efficient. But at the same time it does have its limitations and drawbacks such as needing a wide array of images to be trained on in order for the manipulation to be effective. Because if it doesn't have that then of course they're going to be some defects present and they're going to be some artifacts in the image. But something like that is really, really interesting because I don't think we've ever seen a tool like that with images where you can simply rotate objects and, and simply have them generatively produced. Then we had this company called figure raise a $70 million Series A round in order to help produce their humanoid robots. Now I got to be honest with you guys, this stuff is advancing really rapidly. And Remember in summer 2023, which is this summer, we are likely to see one of the first humanoid robots backed by a company from OpenAI. Now figure does have a promising future because it's been hiring from the likes of Boston Dynamics, Apple and even the likes of Tesla. So with this series a seventy million dollar round, it's really going to push forward what we are going to be seeing because we already know that Tesla bots already walking around and moving. OpenAI is going to be working on their robot. And of course now in comes Figure. And what's very interesting is that all of these robots do have the same kind of design in terms of that sleek slanted cylinder shape face. So will be interesting to see where these robots are in a year from now and what kind of AI systems do power them. Then we had something that left everyone shocked. There was a new paper which was released in which they talked about the tree of thought where they were able to get GPT4's reasoning improved by 900%. So I'll leave a link into the video that covers that in depth. But there was a ton of stuff and honestly I can't wait for what's in store next week.