AI NEWS - You’re Not Gonna Want to Miss This

All right, let's get into the AI news for this week. Last week pretty much everybody finally got access to GPTs inside of chat GPT, assuming you're a chat GPT plus subscriber this week, well, it seems everybody was trying them out and ChatGPT got a little overwhelmed. You can see from this Tweet here on November 14th from Sam Altman we're pausing new chat GPT plus signups for a bit. The surge in usage post Dev day has exceeded our capacity and we want to make sure everyone has a great experience. And not only have they paused new signups for ChatGPT Plus. ChatGPT has been really slow lately and it seems to be way easier to hit the usage cap than it used to. When I was recording my most recent video about how to create your own GPTs, while I was trying to demonstrate stuff, I was actually getting rate limited after only 15 prompts. So I'd enter 15 prompts and then it would tell me to wait two hours before I can prompt again. So that 25 minute video literally took me about six hours to record. So so that I could just pick up where I left off and continue prompting again. So hopefully ChatGPT figures out how to scale this and we can get back to normal with ChatGPT usage. Speaking of OpenAI and GPT, apparently OpenAI is starting to work on GPT5, and according to Sam Altman, it could even be super intelligent. Sam Altman recently did an interview with the Financial Times and in that interview he said that they actually need more investment from Microsoft in order to continue building more major AI models. Now, the OpenAI team hasn't actually set a timeline for GPT5 and they haven't really said what the capabilities it might have are, but according to Sam Altman, they are starting to work on it. But personally I wouldn't expect it anytime soon. Training these models takes a lot of compute power and a lot of time. If I had to guess, I don't think we'll see GPT5 in 2024. Maybe 2025. We'll have to see. Saying that I have made predictions about how far out things are going to happen and and almost every time I've done that they happen way faster than what I claim they will. So I'm probably wrong there. There was also actually a ton of news coming out of the Microsoft camp this week, who happens to be OpenAI's biggest partner. This week was Microsoft's Ignite event, which was mostly focused on their Azure cloud computing platform and the advancements they've done with the cloud computing. But there was a few few AI announcements throughout this keynote, including how all of OpenAI's tech is actually going to be rolled into the Microsoft Azure Cloud.

As OpenAI innovates, we will deliver all of that innovation as part of Azure AI and we are bringing the very latest of GPT4 including GPT4 Turbo, GPT4 Turbo with vision to Azure OpenAI service.

Now in one of the coolest moments of this keynote, Satya Nadella actually showed off some something that we haven't even seen yet from ChatGPT and that's the inclusion of adding videos into a prompt. Now this was only a slight like couple segment of this entire keynote, but I can't wait to start using this myself.

Soon you'll be able to connect GPT4 Turbo with Vision to Azure AI Vision, allowing you to prompt with video, images and text. In fact, our customer WPP is already using this today with one of their largest clients.

They also announced what they called models as a service where you can use the Microsoft Azure cloud to fine tune and custom train your own stable diffusion code llama, Mistral 7B and other models directly on their clouds. And since Nvidia is a part of everything and Jensen Huang needs to show up at every single keynote, there was a quick cameo from Jensen talking about how Nvidia and Microsoft are deepening their partnership and how Microsoft is going to start using the H200 chips that Nvidia is putting out soon. They're also completely overhauling Microsoft Teams where you can now create your own avatars and show up onto team calls using your avatar instead of your real face. And you'll even be able to prompt in your own meeting space to look how you want that meeting space to look look when you do your Microsoft Teams meetings. It also sounds like Microsoft is going all in on the Copilot branding instead of Bing Chat. Pretty much everything is now going to be Microsoft Copilot.

Bing Chat is now Copilot. It's a standalone destination and it works wherever you are on Microsoft Edge, on Google Chrome, on Safari, as well as mobile apps coming soon to you.

And speaking of OpenAI and their GPTs, you're going to be able to use GPTs inside of Microsoft Copilot pretty soon, as well as various plugins from Microsoft that they announced in a previous keynote.

GPTs are a new way for anyone to create a tailored version of ChatGPT that's more helpful for very specific tasks at work or at home and going forward you will be able to use both plugins and GPTs in Copilot to tailor your experience.

And this week I even saw from Paul Cuver here that in some Bing accounts the plugins have already started rolling out. Now I checked my account, they're not in my account yet. Paul does give instructions to check to see if it's in your account. You basically go to bing.com chat and if you see this new little plugins tab next to your recent activity, well then it's rolled out into your account. But this is a little bit of a preview of what it looks like. You can see we've got instacart, kayak, klarna, OpenTable, shop and search as initial available plugins inside of Microsoft Copilot. And then the last really interesting thing that I found from this main keynote Satya talked about Copilot Studio.

We are announcing Copilot Studio. You know, with Copilot Studio you can build custom GPTs, create new plugins, orchestrate workflows, monitor in fact your Copilot performance, manage your customizations and much, much more.

So here's a look at what Copilot Studio will look like. It looks like you'll be able to train your own data into it, use custom GPTs, add plugins to it, and it looks like they have this really cool visual builder where you can sort of string together elements that you want your copilot. So it almost to me looks like an even more advanced version of what we're getting out of GPTs by taking a peek at the Copilot Studio website here. Looks like you can use your own data here, integrate it with all sorts of other tools, point it to your own website and it looks like we can try a demo or try building our own free version right now. But let's try the demo here real quick. And let's plug in FutureTools IE and start the chat. And it says, hello, I'm the virtual assistant for FutureTools IO. How can I help? Let's type what tools can help me cut long form videos into short form videos. Looks like it says you can use Wise Cut and Spike Studio to cut long form videos into short form videos. Let's go ahead and open these and see if that's what these actually do. Wise Cut is an online automatic video editing software. Automatically cuts down long videos into shorter impactful content. Spike Studio from long form videos to viral clips in one click. So even their demo is already pretty useful. Microsoft Copilot Studio is definitely something I'll have to make a deeper dive video on in the future and really put it through its motions and see what we can do. Let's see how it compares to GPTs, but again I've got to save that for a future video. And something else that came out of the Microsoft Ignite event, although it wasn't in the main keynote, was this ability to create what TechCrunch is calling deep fakes. Really what it's doing is similar to what we've seen from tools like DID or heygen or Synthesia, where we can give it text prompts and it will create a talking avatar where the lips look like they're moving. Pretty much what we've seen from plenty of other tools. It looks like Microsoft is just kind of creating their own version, but here's a quick demo.

Hello, I am Christina. Let me show you how to create an engaging avatar video in four steps. It begins with a talking script for your avatar.

Actually, don't think the lips move very well to the voice on this one yet. I feel like did and hey gen do it quite a bit better still, but interesting to see that Microsoft is rolling this out as well. Not something I would have actually expected from Microsoft. They also appear to be rolling out an 11 lab style thing where you can train your own voice for text to speech models. Moving on past Microsoft Although Satya and Jensen Huang did mention that they were going to be using H200 chips in their upcoming cloud computers, it was actually only this week that Nvidia announced that they'd even be making these H200 chips. Nvidia H200 is the first GPU to offer HBM3e because we know what that means, but essentially it's going to have faster, larger memory to fuel the acceleration of generative AI and large language models while advancing scientific computing for HPC workloads. So basically Nvidia is making better, faster chips designed for training AI, and Microsoft's going to be using them in their cloud computers. All right, moving on to Google News, Google's actually rolling out a Bard experience for teens. It's essentially what we've come to expect from Bard, but with more guardrails on it. Basically, they trained Bard to recognize areas that are inappropriate to younger users and implemented safety features and guardrails to help prevent unsafe content such as illegal or age gated substances from appearing in its responses to teens. Also, if teens ever ask for facts, it will automatically run their double check response feature instead of forcing you to click on it anyway, which in my opinion seems like they should just do anyway. If you're going to ask it a question and you're looking for facts, why not always run the double check? But apparently they're going to do it for teens. Apparently YouTube is about to start cracking down on AI clones of musicians. They're going to start allowing music labels and the musicians to send takedown requests of music that uses the unique singing or rapping voice of an artist. They're also rolling out a requirement where you must let people know when your content was generated by AI. And you can see it's going to put a little label on the content saying altered or synthetic content. This is something people have asked me about recently. How do I feel since I am in AI and I'm also in YouTube and personally I think it's a great move. I think people should know if they're watching something that was generated by AI. So asking people who are generating AI content to inform you that you're watching AI generated content seems like a good thing to me. I don't really see the downside to the only people who would not like this are people that are trying to trick others into thinking it's not AI generated content. But speaking of YouTube and AI and also music, this week Google DeepMind showed off their new Lyria model and they're trialing this Lyria model in an experiment they're calling DreamTrack. Basically, a limited set of creators will be able to use Dream Track for producing a unique soundtrack with AI generated voice and musical style of artists including Alec Benjamin, Charlie Puth, Charlie xcx, Demi Lovato, John Legend, Sia T Pain, Troye, Sivan, and Papoose. These names I've never heard of before. These are all artists that gave YouTube permission and are working with YouTube and DeepMind to allow people to use their unique voice in new songs. They've also developed a way to watermark AI generated audio that theoretically the normal person won't be able to hear the watermark, but it does something to the waveform visualization so that somebody looking at the waveforms can actually see and detect the watermark. And while that last post was on DeepMind's blog, YouTube's official blog is talking about it as well. This tool's also able to do things like create horns by you just sort of humming what you want the horns to sound like. So you'll be able to generate custom AI music for your videos that, since it's AI, shouldn't have any royalty or licenses attached to them. Once again, something that seems pretty cool and I can't wait to get my hands on it. And finally, in other YouTube news, they've started to roll out some of their AI features people to test out. If you are a YouTube Premium user, you can test out the mobile versions of some of their new AI functionality. Right now you can test out their conversational AI where you ask questions about the video you're looking at. And I guess using the transcripts, it can answer questions about the video you're watching. And then they also have a comment topics summarized by AI. So you can just get a summary of what all the comments say without having to read them all. If you want to get sort of a general feel for what the comments are saying about a video, this tool will summarize them for you. Now if you're a YouTube Premium subscriber, you can go to YouTube.com new and there are buttons to try these features out. So if you want to ask the questions, that's available now, but right now it's only available to YouTube. On Android you can also try out the comment topic summarized by AI, and this one is available on both Android and iOS. One thing to keep in mind, they only let you try out one at a time so that you can't actually turn on both the comment topic summarizer and the conversational AI and have them both running at the same time. They only let you do one or the other. All right, I've got some really cool stuff to share with you in the AI art world. This week, for example, Meta announced two new pieces of research around their Emu Art Generator model. They announced Emu Video and Emu Edit. Emu Edit seems to be their inpainting feature where you can edit very specific aspects of your AI generated images. They also announced Emu Video, which seems to be their version of what we've seen from something like Runway gen 2 or pika labs or Moon Valley or all of the various AI video generators that are available right now. However, taking a peek at these, they actually look really good. They seem to have a lot more motion than what I've seen from some of the other video generators. Like this prompt of a couple dances elegantly in a luxurious hall in a masquerade ball. Photorealistic that looks pretty dang good to me. This wood on fire prompt looks awesome as well. A pizza spinning inside a wood fired pizza oven. Dramatic vivid colors. A majestic white unicorn with a golden horn walking in slow motion underwater. A robot with a TV screen as a face at the Acropolis in Athens 4K high resolution a hamster wearing virtual reality headset is a DJ in a disco. But my guess, based on what I'm seeing here, is that they're all going to be pretty short generations. Still like three or four seconds like we're getting with the rest of the tools now. I don't know when we're going to actually get access to this. It says this work is purely fundamental research right now, but the potential use cases are clearly evident. But it appears to be some fun video creation and AI image editing models that should be fun to play with when they're finally available to normal people. Staying on the AI video theme for a second, here's something really cool to come out of Runway this week. I don't believe it's available yet. I don't see it in my Runway account, but it's the ability to do inpainting in videos and actually change what we're seeing in a video. So what they did in this video was they selected this truck here and made the truck the only thing that's moving. They selected just the smoke and made the smoke the only thing that's animated. The waterfall. The rest of the image isn't animated. You are just picking and telling it what aspects of these images to animate. And that looks like it's going to be really awesome and give us even more control over what we're doing with our AI generated videos. This is going to be game changing. Another cool thing in the AI video world is Deforum. And I've actually did a whole livestream with Deforum with one of the creators of Deforum. And it makes these really cool sort of morphing videos where it sort of animates between various images. Well, now Deforum is doing a Discord bot to make it really, really easy to create these types of animations. In the past you had to use something like automatic 1111 and there was all sorts of settings and configurations and number different knobs to turn inside of the app to get the video you want. And from what I can gather, you're going to be able to use the SDXL model and generate these deform animations pretty quickly and easily directly inside of Discord pretty soon. So something to look forward to as well. If you go to app.Deforum.art beta signup, you can actually get on the wait list. I do not know when this is rolling out, but I love playing with a high video. So this will be something we will deep dive in once we get access to it. Let's check in on Midjourney real quick. Alijules was on the November 15th office hours. They recently released a Niji style tuner. They announced that version 6 is coming out sometime in December. Apparently we're going to be able to generate images directly from within the website within one to two weeks. However, it looks like it's only going to be available for people who have generated over 10,000 images to start. It also looks like version 6 is going to have better text understanding. It's going to be very likely that version 6 will actually be able to write text. They claim it should be ready before the holidays if all goes well. So fingers crossed for that. There's this really cool tool that's been circulating this week called Kre AI k R e a and it actually allows you to update images in real time. In fact, here's one of my buddies, Riley Brown, showing off what it can do. So check this out.

You can actually just add clip art. So if we add a clip art of this umbrella here and drag it down right here. He's now holding the umbrella. Look at that. He's holding the umbrella. And we're gonna enhance it. Look at that image.

Now this Kria tool is still in beta. You can get on the wait list and they are rolling out access to it. I did get access to it to try it out myself. I could give it a prompt like a man laying in the sand on a beach. Then over on this left side, I can start drawing here. So let's go ahead and select a background here. You can see now we've got our sort of beach sand color. Let' add a rectangle. I'll make it blue. We'll put that right here. I'm going to grab an ellipse here. Let's make it. I don't know, let's make it red. I'll make a little head here. I'll grab the paintbrush. You know, draw a little thick figure lying here. This is supposed to be his towel on the beach. And look at that. We're getting a little dude laying in the sand on the beach. There's some other modes here. Cinematic illustration product. Let's go ahead and leave it on Cinematic here. And what's really cool about this is I can grab this image over here on the right, drag it over to the left, left. And now it's using this as my starting image. It enhanced this one over here on the right. And now I can continue to build off this. Let's go ahead and grab a red color here and draw something in the sand up here, Put something in the sand over here. Something in the sand over here. You can see as I'm doing it, everything is changing. Now, I haven't spent a ton of time with this yet, but it seems really fun. You can even flip on your webcam and have it generate AI generated images using what it's seeing on your camera. In fact, here's an example from Karen Chang that I saw earlier today where she's got the webcam mode turned on and you can see as she's moving around, the image on the right side is actually changing with her movements. It's pretty dang crazy. Again, head over to Kreakrea AI, join the wait list. It's pretty cool. This week Notion rolled out a new AI called Q& A. And basically what this does is you can ask questions and it will look across your entire Notion database, all the information you've ever plugged into your notion, it can query that information. It's basically using everything you've ever entered into your Notion account or context for any questions that you ask it. So you can ask it things like what were the action items from last week's website redesign meeting? And it will search all of your notes inside of Notion and bring up the action items from that meeting. If you're someone that uses Notion, a lot of this seems like it will be really, really handy to find stuff very quickly. I'm actually going to be talking about it even more in one of next week's videos, so look out for that. Google's DeepMind developed something they called GraphCast and it's an AI model that's faster and more accurate for global weather forecasting. It can actually deliver 10 day weather predictions with unprecedented accuracy in just one minute. And it can also offer earlier warnings of extreme weather events. It says here, for the inputs, GraphCast requires just two sets of data. The state of the weather six hours ago and the current state of the weather. The model then predicts the weather six hours in the future. This process can then be rolled forward in six hour increments to provide state of the art forecasts up to 10 days in advance. It's almost like the way a large language model works where it's constantly trying to predict the next best word in the statement that it's putting out. Well, this is looking at the graphs of the weather and, and trying to predict the next thing that's going to happen with the weather. It's very fascinating. And finally I'm going to wrap up with this announcement of the go forward care Pod. This is an AI doctor's office. It appears to be like a little room that you walk into. You go into this box and it will actually scan your body to help detect all sorts of medical conditions. It could do genetic analysis, heart health, detect skin cancers, and even help analyze mental health. So in the very near future, this might be how medical diagnoses work. You walk into one of these boxes, it does all the scans it needs to do, and then passes that information along to a actual doctor to help figure out the next courses of actions, any sort of treatments that need to happen. But a diagnosis process may become fully automated by AI in the very near future. And to me that's pretty exciting because it will likely free up a lot of doctors to do even more important work. And I would imagine that the accuracy of these diagnoses are probably going to be pretty solid. Anyway, all of this stuff excites me. I'm a huge nerd. I love this stuff. I hope you love this stuff too. I'm excited to see this roll out. Who knows, maybe one day I'll be making a vlog style video where I'm getting a full health analysis by walking into one of these AI boxes. But there you have it. There's the breakdown of the AI news for this week. Again, I record these videos Thursday, so there may be even more news that I missed from this video, but if there was, I'll make sure it's in next week's video. And if you do want to stay in the loop with all the latest AI as it happens, check out Future Tools IO. I keep the AI News page up to date on a daily basis. I curate all the coolest AI tools that I come across. And I even have a free newsletter where every single week I share with you the five coolest tools that I've come across. I share with you three interesting news articles, a handful of YouTube videos, and a bunch of other really cool stuff happening in the AI world. It's totally free. Just click the Join the free Newsletter button. You can find it over at FutureTools IO. So thank you so much once again for tuning into this video. I really, really appreciate you. If you want to stay in the loop with all the latest AI news, you want to get AI tutorials, you want to learn about the coolest AI research that's in the works, make sure you subscribe to this YouTube channel and like this video and it will ensure that more videos like this show up in your YouTube feed. So thank you once again for tuning into this video. I really appreciate you. I'll see you in the next video. Bye. Bye.