ChatGPTS MASSIVE UPDATE, Ai Discovers NEW PLANET, Googles New Robot, Text To Video [AI News#2]

From Boston Dynamics to ChatGPT's major upgrades, this has been a massive week in AI. So let's cover absolutely everything you missed and without further ado, let's get started. So we had Gen 2 release their amazing text video where essentially you need zero driving image. You can just type in a searcher catching a wave and just like that, we now have footage. This is really good because now we don't need any input image and we can literally just use a simple text prompt to, to gain an entire video output. Now this gen 2 is not the only insane thing that there is, because boy, oh boy, trust me when I tell you there have been some other stuff that people have done with this software and trust me, you're going to enjoy this. So this user by the name of Pizza later decided to create a fully made AI commercial that is about a pizza spot. And you can see right here, he used text to video to create the visuals for this and then decided to edit it all together with some nice voiceover. I'm gonna leave a link to his Twitter account because he honestly creates some amazing stuff and it's very, very uncanny at how realistic this somewhat is. And these even sound better with sound. Take a look at this one created by Amar Reshi. It sounds amazing. Boldly calculating the cost of victory. This is a tale of courage and sacrifice, of the indomitable human spirit rising against the tide of evil. This is the story of a world at war. Prepare for a journey into the heart of darkness. This is World War II. So yeah, Gen 2 releasing their text to video model honestly has so many applications. There have literally been so many videos that are circulating online of people creating masterpieces like the one you just saw. So what happens when they release Gen3? Honestly, AI is so exciting. Then we had this amazing research paper and honestly, at first glance it doesn't seem like much, but understand what you're looking at right here, it's truly amazing. It says scaling transformer to 1 million tokens and beyond with RMT. Now why is this actually insane? So RMT stands for recurrent memory transformer and GPT4 actually has 8,000 tokens in its normal way. And of course they also newly released the 32,000 token concept length which they are separately rolling out now. This paper actually allows 1 million tokens. They figured out a way to to get 1 million tokens. So essentially what this means is that GPT4 or potentially large language models in the future are going to be able to remember 1 million worth of tokens, which essentially means that they can now write novels, they can write entire screenplays, they can write entire movies, entire trilogies, if you will, just based on the fact that they're able to retain information. You have to understand that this was one of the huge limitations of large language models, because a lot of the time when you're having a conversation, eventually it forgets what was said. And of course you can't have something effective if it can't write for very long periods of time. So this is something that is going to be truly game changing, but when it's finally deployed, it's going to be absolutely insane. Now, everybody knows what Boston Dynamics robots are. I'm pretty sure you've seen them freaking out everyone. But recently a machine learning engineer by the name of Santiago decided to integrate his own Boston Dynamics spot robot with Chat GPT. Now, it's actually pretty insane because they use this to essentially power the robot so that they can ask it questions and that they can, you know, dictate to it to say something and then it gives it a pretty decent response. I honestly find the entire interaction pretty, pretty cool. And you can check out all of the links in the description that you can see the entirety of the clips because honestly, today there's just too much to get through. This company called Forefront decided to release an upgraded version of ChatGPT where essentially it's ChatGPT that powered by GPT4, it has unlimited uses, and you can have many different interactions with this large language model because essentially what they're doing is they're allowing you to essentially fine tune it to any means necessary. You can see right here that the user is essentially just talking with it on many different tasks. And you can also get images from it. There are many different ways to talk about these images. And of course, you could also essentially have have many different characters that are embedded into the software. So it's a way that people are accessing GPT4 at the current moment for free and it's currently in alpha. So this is something that is very interesting because a lot of people are fine tuning ways to change the experience of ChatGPT. Now, many people know about Auto GPT where essentially you put one prompt in and then the AI goes off and does completely what it wants to do. But this week was pretty insane because we had many different users that were able to use their own autogpt and essentially fine tune them to get them to do things. For example, one example right here was we had a user who got it to essentially code a specific website and Code it with very quick accuracy. Now remember, this is what people usually do with GPT4. You have to ask it many different things and then it gives you the final code. But with this, you ask it one question and then it runs off and does everything by itself. We also had some other, I guess you could say, applications that are being developed and are likely to be deployed to with a lot of Internet access later this year with the likes of Agent GPT being released. It just goes to show exactly what these kind of autonomous agents are capable of and exactly how they think and how they decide to develop their plans and strategies. This example was where someone decried to create Hustle GPT, which is essentially a bot that is trying to make as much money as possible. And I actually did use this and I did have a lot of fun with one that was also released, which is called Goal GPT, which is completely free, which is. You can also use this. I was giving it tasks to essentially create a multimillionaire from scratch because I wanted to see what its thought making process was. And it was very interesting to see how AI with a very, very high IQ formulates its ideas and thinking process in order to achieve its end goal. And it was just interesting to see that inner workings rather than just seeing a final prompt. So definitely something that I recommend you do play around with now. This was truly absolutely incredible. I couldn't believe my eyes when I saw that Adobe released a Firefly update for video. This was the first of its kind. There is honestly too much to dive into this. I'll leave a link to the full video in the description. But essentially what Adobe have done is they've literally changed the game with video editing by adding a bunch of AI tools embedded into the video software where you can use natural language to change the color scheme of the video, you can use natural language to generate music, you can, you can use it to generate captions instantly. You can literally use it to generate store boards. Guys, it is absolutely insane to be a content creator or someone working within media, because the types of content that we are going to be getting in the future are likely to be very automated and also very quick in terms of the production process. Because essentially what Adobe's AI is now able to do is it's able to identify the boring parts of the workload and then essentially just automate those in order to speed up the video creation process. Like right here, you can, you can see that they're able to generate storyboards just based off a script. And you're able to visualize those storyboards using them for whatever short film you're going to be making. So it's likely to just increase the efficiency and production capabilities of people who otherwise wouldn't have these tools available to them directly. We also had, interestingly enough, Microsoft announced that they're working on a new chip called Athena. Now, the reason that they chose to work on this chip is because essentially Nvidia's are very, very expensive, expensive. And they're deciding that they need to do this to offset the cost, because Nvidia GPUs currently cost around $40,000 each. And of course, Google is also trying to work on their chips, which they are claiming that outperforms some of Nvidia. So we can see that even on the hardware side, the AR race is truly heating up, even if it's in the background. Now, this is where I've got probably the biggest news in the video. Now, everyone uses ChatGPT, but recently chat GPT in alpha, rolling out to a limited number of users, was given Internet access. So you can see right here that it says default GPT 3.5 with browsing optimized for speed and can browse the Internet. Now, I haven't seen any actual use cases for this, and I haven't actually seen OpenAI release any statements about Internet access, but it does actually seem pretty legit based on the screenshots I've seen floating around on the Internet and, and based on the kind of plugins that we've seen integrated with ChatGPT before. And we do know that Internet access was a previous limitation based on the data only being available from 2021. So this is a screenshot of what it does look like. I'll leave a link to the thread on Twitter where someone explains exactly their experience of how they're currently using GPT3 and GPT4 with the Internet. So if you don't actually have access yet, well, I guess it's just slowly being rolled out in alpha, but to a select number of users. And I'm guessing they're going to get that feedback and then potentially refine the software. So honestly, if GPT4 does really get access to the Internet, I think things are about to change really, really quickly. So that's probably going to be the most interesting thing that I'm looking forward to. Then we had something really cool, so everybody knows what stable diffusion is. And then we got the largest and best model yet in stable diffusion. We got Stable Diffusion xl. Now, what's great about this is not just its Image generation capabilities. The best thing that we are getting now with AI, we are now getting the ability to, of course, have better prompts, but text is something that is now actually working. So legible text was one of the largest barriers to AI image generation that we didn't have before, and now that is something that is a reality. So like I said, many different things are happening in the space and it is moving very, very quickly. So I would recommend checking out Stable the fusion because the new feature is definitely really awesome. Although it does take a few prompts, it still is a large step in the right direction. Then we got one of the scariest pieces of technology I've seen recently. Now, everybody's been scared about AI voice cloning and people managing to clone your voice, but this one just knocks it out of the park because their speech synthesis is just far too human. Like. Like, it's insane. Just take a look at these audio samples. I used to work at a fire hydrant factory and, you know, you couldn't park anywhere near the place. I got a face for radio and like, what did they say? Voice for print. Now remember what you heard just there was completely 100% artificial intelligence. Now, I do know that the clarity of that voice wasn't the best, like elevenlabs, but it's what the inflections are doing that we need to pay attention to. It's the ums, it's the ers, it's the little tones in your voice that make it sound a lot more realistic, which means that it's going to fool people even more, which does mean that it's going to be increasingly harder to detect these AI voices. Now, something that I do want to add, although this technology is quite impressive. Impressive. You have to remember that, like I've said many times before, Google has been advanced in many different AI fields, but they just, for some reason, won't release their technology. Compare bark to what Google released in 2018. Go. Happening. Hi, I'm calling to book a woman's haircut for our clients. I'm looking for something on May 3rd. Sure, give me one second. Mm. So, yeah, five years ago, Google was sitting on a piece of technology that had AI voices that sound exactly like humans that could literally make phone calls and book appointments. So what are they doing with the technology? We don't know at the moment. So there's still some more developments that we do need to get to. The coolest ones was VideoGPT. And video GPT is essentially what it says. You know, you can literally talk with with Videos in a chat session. And what's really good about this is that this kind of technology can be used to identify many different things. Now, it's not just something that can be used to identify basic stuff. Essentially what you can actually do with it is you can query it, so you can ask it, you know, where is it? What are they doing? How does that work? So essentially you can get information out of the video and you can get some kind of understanding rather than just saying, okay, the trees are green, the floor is, you know, black or whatever. So it literally tells you what people are wearing, it tells you what they're doing, it tells you why they're doing it. And it's something that is free at the moment. So this is something that you can actually use. And it's actually pretty, pretty accurate because I tested this in a video, which I will leave a link to in the description, and it was something that I didn't think got enough attention. So it's definitely something that you should check out. But it just goes to show what could actually happen. When OpenAI finally released their GPT4 with images and. And then of course, finally in the future, perhaps releasing that one with videos, because of course, as we know, images make up videos. Google came with another announcement talking about their new browser called Project Magi, which is essentially a browser that is focused on AI answers and having the AI integrated into the browser. So it's definitely going to be something that competes with Bing because of course, as you know, Bard wasn't technically what we all expected and what we all wanted. So I'm guessing that Google is trying to shift their focus on onto what they already have the competitive advantage on. And then of course, integrating an AI with that to provide a much better service than Bing does. So it'll be interesting to see when this releases this month, because that's when it's scheduled for. And let me know if this is going to be something that you're going to be testing out. Then we had a University of Georgia research team confirming the evidence of a previously unknown planet outside the solar system. And essentially what they did, they were able to use machining learning tools to detect it. Now, the reason this is actually really cool news is because they already previously scanned that specific area and they didn't find anything. But the machining learning tool that they used actually was able to detect this planet outside the solar system, which meant that these tools that we're using can previously find things that we may have missed. Then, of course, we had Google's DeepMind their new project, which was essentially to teach some robots from scratch how to play football. And they were actually pretty effective this they're able to get up pretty quickly. They're able to score goals, they're able to perform defensive tactics. And this was all done using deep reinforcement learning, which is where, you know, you train a robot or you train an AI on a specific number of times, usually millions of times, and then eventually, after those times, it becomes really good at performing a specific task, or I guess you could say in this task, which was to play football. So it was definitely really interesting to see how these small robots are able to get up quickly, but perform strategic defensive behaviors that are very, very interesting to see. And honestly, I think this is going to be more interesting in the future, when we have much more capable robots performing much different tasks.