AI News: The AI SHOWDOWN We Needed!

It's been another busy week in the world of AI, with Google and OpenAI sort of competing to make the biggest announcements of the week. It's been really fun to watch, so let's just go ahead and dig right in, starting with the announcements that Google made this week. Probably the biggest thing that everybody has been talking about this week has been Google's VO2AI video model. I actually managed to get my hands on early access to it and I made an entire breakdown of it testing prompts that you guys gave me on X. You can find the entire breakdown video of it right here on my YouTube channel. It's called this new AI video video tool is the best I've seen hands on with Veo, and that title really says it all. This is the best video model that I've seen so far. It makes really realistic videos. It seems to understand physics better, it seems to understand people movement a lot better. Everything about it is just really, really good. Every single prompt that you give it generates four videos for you. So even if one of the videos is a bit wonky, you still have three more to choose from, making it more likely you're going to get a video that you actually like. Now, I don't know exactly when this is going to rolled out publicly, but they do have a wait list over at labs google./fx,/tools/video fx. If you get on the waitlist, supposedly they're rolling it out to the general public fairly soon. They also have a text to image to video model which first generates an image and then once it's generated, the image converts that image into a video. But at the moment there's no way to upload your own image and convert it to a video. Again, make sure you check out my deeper Dive video all about VO2. Google also made some updates to their Imagen 3 model. It's now a generated apparently brighter, better composed images. And from what I've seen, all of the images are pretty dang good. And this one, most people should have access to it. If you go over to Labs Google and you select Image FX, this is using the newest Imagen 3 model. And just like Veo, it will generate four images with every single prompt and they're pretty dang good. I mean, in my mind I feel like AI image generation is a fairly solved problem. You can pretty much generate anything you can imagine in image form now and and all of these platforms are just adding minor details that make them slightly more improved. But for the most part I feel like we've hit that point where you can kind of generate anything you want. It might just take a few rerolls to get there. Google also showed off a new platform called Whisk, which is this really cool way of blending multiple images together to create a cool new image. This one could be found over at Labs, Google FX Tools, slash Whisk. And you might have seen some of these viral examples floating around on social media, where people are making plushie versions of themselves. You can give it a style like this plushie. Toss an image of yourself in like I just did with this image, and it spits out a plushie version of the image. If I click open in tool, you can see the plushified me. But if you don't like that style, you can upload images of your own styles, your own scenes, and of course, the subject, or you can roll the dice and let it do it for. For you. So let's have it do, like a pin with whatever this scene is here, and we get this cool blended pin that has me with this background, like hosting a podcast. It's really fun to play with, and it's totally free to use right now. Again, it's called Whisk, and you can find it over in Google Labs at Labs. Google NotebookLM also got some major updates this week. If you're not familiar with NotebookLM, it's the platform where you can upload documents and then chat with those documents. But the feature that everybody's really talking about is the feature where you can actually have it turn your documents into a podcast. Well, NotebookLM got this redesign this week, so it's a completely overhauled, kind of cleaner layout, but they also added this interactive mode. So I uploaded a transcript from a podcast with Dario Amadai. Let's go ahead and have it generate a podcast. And now when I'm listening back to this podcast, I can go into interactive mode here.

Okay, ready to dive in? Today we're tackling AI.

Oh, yeah.

Specifically how fast these systems are scaling up.

It's incredible.

And how one company, company Anthropic.

Oh, hey, welcome.

What's up? Hey, I'm just curious. Can we just kind of get right into it? Tell me who this Dario Amodai guy is.

Oh, hey there. Yeah, absolutely.

Great question. I'm glad you asked.

So, Dario Amadei, he's the CEO of Anthropic, this AI company we've been talking about.

And he's kind of a big deal in the AI world.

Right. And what's interesting is that he's not just, oh, go for it.

Yeah. Yeah.

What.

What makes him a big deal? I don't really want to get into all the rambly stuff. I'm just kind of curious. Why is it a big deal? What was he doing before Anthropic?

Okay, so you want the quick version on why he's a big deal?

Got it.

Yeah. Let's uncheck this.

Basically, he's at the forefront of AI development and safety.

Before Anthropic, he was deeply involved in.

Scaling laws, which we'll get into, but basically means understanding how these systems get smarter.

So basically now you can create these podcasts using any content you want. You can upload text files or PDFs or pull in YouTube links or whatever you want. And then while the podcast is going, you click the join button, you can interrupt them and you can ask them questions about the document. It is super, super cool. And, and this is now available to everybody. They've rolled it out into everybody's Notebook LM accounts for free. But you know what? Google just kept shipping this week. They had so many cool updates. They also rolled out a new 2.0 advanced model inside of their Gemini Advanced platform. So in order to have Gemini Advanced, you do have to pay for the Gemini upgrade on your Google account, but you can now use their most advanced large language model. So if I head over to Gemini google.com since I'm a Gemini Advanced member, I. I do have a dropdown here where you can see multiple options including 2.0 experimental advanced. And if we take a quick look over at the Chatbot arena here where they compare the various large language models, we can see the Gemini Experimental model is the top ranking model right now. I'm guessing this is what we're getting when we play around with the new Gemini Advanced 2.0 model. Notice which model is right here, almost tied for first place. This Gemini 2.0 flash thinking guess when that one rolled out. That one came out this week as well. Here's Logan Kilpatrick from Google saying, just when you thought it was over, we're introducing Gemini 2.0 flash thinking, a new experimental model that unlocks stronger reasoning capabilities and shows its thoughts. This is apparently Google's answer to OpenAI's Zero1 model. You can actually test this one out over in Google's AI studio. If you go to aistudio.google.com over on the right, you have the option to choose a model and you'll notice that Gemini 2.0 flash thinking experimental is one of the options available. And as of right now, it appears to be free to use. Let's give it the old Strawberry test. How many Rs are in the word strawberry? We can see that it thought through. You can actually expand the model thoughts here. Identify the target word. The word is Strawberry. Focus on the target letter R. Examine the words letter by letter. S not an R. T not an R. R. Yes. Count equals one and you can see it go through and actually count the R's in the word strawberry. Finally summing it up one plus one equals three and then stating the answer Strawberry. There are three R's in the word strawberry. Obviously this isn't the most advanced test, but anybody that has some really complex logic problems or math, I highly recommend checking it out in Google AI Studio and seeing how you think it compares to OpenAI's One Model. And while we're on the topic of Google, Obviously Google owned YouTube just rolled out a new feature. A lot of people are concerned about having their content that they put on the web being trained for AI purposes. Well now YouTube makes that opt in. So if you are okay with AI training on your YouTube videos, you can check a box inside of YouTube now to allow third party companies to train AI models using your channel content. I talk about AI. I'm a proponent of AI, so of course I said sure, use my videos to train AI. You could uncheck this box or you can select only the parties that you're willing to let train. So if I only want OpenAI to be able to train on my videos, I can select just OpenAI. But again, I'm totally cool with them training on my content, but totally understand the reasons why some people wouldn't be okay with that. Now the interesting thing about all of these things that Google has been releasing is it feels like it's been a day for day competition with OpenAI. Because OpenAI originally announced their 12 days of OpenAI announcements, where every single day for 12 days they would make some sort of new announcement or some sort of new rollout. Now the day you're watching this video is the 12th day of OpenAI and they've probably announced something new and it's probably something pretty big. I would imagine whatever they announce on the very last day of the 12 days is going to be one of the bigger announcements that they're going to make. But I could be totally wrong. For all I know, on the day you're watching this, they released that you can now send fax messages to OpenAI. I don't know, but it's probably going to be something a little Bit bigger than that. But here's the five announcements they've announced since I made my last AI news break down on day seven. They announced projects inside of Chat GPT, which is essentially folders. It's like the same thing as projects inside of Claude, if you've ever used Claude's projects. If I go to my Chat GPT account, you can see over on projects. I've got a test project. I can create a new project. I'm going to call it Test 2 Project. You can tell I haven't really been using this much yet. And this project can have its own custom instructions and its own set of files that it uses every single time you prompt it with something. So if you want to make a project that that is specifically designed just to answer questions about a certain product's like documentation, you can upload all the documentation into the files and then anytime you prompt something, it will use that documentation to help with the response. On the eighth day of OpenAI's announcements, they announced that the search functionality is being released for everybody, no matter which ChatGPT plan you're on. Now, if you're a pro or plus user, you've had this for a little while now. But if you were a free user of ChatGPT, you now have this as well. So let me jump in to the ChatGPT4O model. You can see the Search the Web button. I can select this, and anytime I ask it a question now, it's going to search the web for the answer. For instance, if I say, what does Matt Wolf's YouTube channel talk about? It will do a search and answer the questions here. And funny enough, it actually pulled up a YouTube video that I did on the Hustle Flowchart podcast with Joe Fear, all about what I do on my YouTube channel. On the ninth day, they announced that the O1 preview model, the model that's really good at thinking and reasoning, is now available in the API. So if you're a developer before, you were only able to use the GPT4O model. Now they actually gave you access to 01 to use that in whatever software tools you're building. If you go to OpenAI's playground over on the right, where you select your model, you now have access to O1 mini and O1 preview as part of the API. On day 10, they announced 1-800-ChatGPT, which is essentially a way to use your phone and call in and have a conversation with ChatGPT. I saw a lot of people saying that this is really good for people Outside of the US that maybe don't have Internet access, they could call in and use this. But if you look at the fine print on the video right here, it actually says us only for phone calls. Yeah, I don't really get the thinking here, but it's essentially the advanced voice mode that if you're in the us, you can call in and use. Here's what it sounds like.

Hello again. It's ChatGPT, an AI assistant. Our conversation may be reviewed for safety. How can I help you?

Is your refrigerator running?

Yes, it's running smoothly, so I better go catch it, Right? Classic.

Hey, that was my line.

You set it up so perfectly, I couldn't resist. What's next?

Anyway, you can ask it whatever you would ask. ChatGPT, voice mode. You just call 1-800-CHATGPT. They even showed using it on a rotary phone if. If you own a rotary phone. And on day 11, the day that I'm recording this, they announced that the Mac app for Chat GPT now works with more tools. Now, I'm not on a Mac right now, so I can't show you the new tools that it works with, but basically you can give it access to different tools on your Mac. They did say that Windows is going to get these features soon, but right now it could only view certain tools on your Mac, which I find funny because their biggest partner is Microsoft. You'd think they'd prioritize PC and Windows users first, but no, they prioritize Mac users first. In my opinion, compared to the previous couple weeks of OpenAI announcements, none of these were that great. But I really have high hopes for day 12, Friday, the day this video is going live. I do think there's going to be some sort of big announcement. I don't know what it's going to be yet, so I'm excited for that because this week's announcements were kind of on the weak side, if you. Yes, my opinion. I mean, they're all cool, they're all great advancements, but nothing that really got me excited. Like some of the Google stuff that rolled out Veo Whisk, Notebook, LM Updates, Gemini with Thinking Gemini Advanced. I mean, Google crushed it this week and last week with their announcements. While in the past OpenAI usually kind of overshadowed Google's announcements the last two weeks, I feel like Google crushed OpenAI with the announcements. It's just really fun to watch. I mean, I posted a post up on X and I said, who's winning the week? Google or OpenAI? And pretty much everybody said Google's winning, but in reality, the consumers are winning because we're watching this battle play out. And because this battle's playing out, we're getting cooler and cooler tools by the day. You know, there's a lot of great AI video tools out there right now. Tools like Sora and Veyo. They're all very promising, but they're only going to get you a few second clips. And it doesn't come with audio. They're not really ideal for making longer videos. Video content. That's why for today's video, I partnered with InVideo AI. It's a single tool that will generate scenes for you, compile them into a full video, add audio for you, and basically do the entire process of creating a short video with AI for you, all in a single platform. So check this out. With InVideo's new version 3.0, you can give it a prompt of whatever video you want and it will create the entire thing for you. I'm talking generate the script for you, generate all of the videos inside of that script for you, add sound effects, add music, and even do the voiceover in your own voice if you want. They have simple templates that make it really easy, like create a short video. You just follow the instructions here, an explainer video, animated films, and a ton of other pre built workflows that you can choose from. So you're not starting from scratch at all. And of course you can use your own script if you've already got something written up. Let's say you need an ad for your latest course about how to learn piano. Let's go ahead and use this template for an animated here we'll tell it to create a 30 second animated advertisement for YouTube about a beginner's course on learning to play the piano. Now if you want, you can select longer videos, you can select different platforms, you can select to use only generated clips, only generated images, or only stock footage. Let's go ahead and use generated clips and we can set the style. Let's do it in a Disney Pixar style. Now I could go through and select this type of background music, which language I want it to be in, whether I want it to have subtitles. I could pick specific voice actors, add a watermark, select my music preference, or I could just let in video, figure all that stuff out for me. So let's see what it comes up with for me. For my beginner's piano course, I'll click continue here and we can see it pre generated a prompt for me. Create a 30 second animated advertisement for YouTube about a beginner's course on learning to play the piano. Use only generated clips, use a Disney Pixar style. Let's leave it that simple and generate the video. Now this does take a little bit of time because it is writing the script, generating all of the clips for it, adding the voiceover, adding the background music. It's doing all of that behind the scenes from that single prompt. And after about 15 minutes of generating, here's the video ad that it created.

Ever dreamed of playing the piano with Matt's online beginner piano course? That dream is just a keystroke away. Our fun lessons make learning easy. Play your first song in just one week. Join thousands who've discovered their inn musician. Enroll now@mattsbeginnerpiano.com and unlock the melody within you.

I really like the generation. The video of it was really good, but the music was almost like this weird, creepy music. And so if we want to change that, I can just give it a prompt like change the background music to a more upbeat, happy piano song. We'll click generate again and within seconds we have a new version.

Ever dreamed of playing the piano with Matt's online beginner piano?

Much better. But you know what, I actually have my voice trained into invideo as well. Let's change the narration to my voice. I'll just give it the prompt, change the narration to Matt's voice and click generate. About a minute later, I've got another variation. Ever dreamed of playing the piano with Matt's online beginner piano course? That dream is just a keystroke away. Much better. And this is just scratching the surface of what I can do within video. I can give it a prompt to change pretty much any element I want about the video, or I can click on edit, regenerate any single scene from my video, replace it with stock footage, upload my own video, upload my own music, or choose music from their stock music library, change anything about the script and who's actually narrating the script and so much more. Unlike most of the other video platforms that are available right now, you can control pretty much every single element of the output of your video within video AI. I can even change the whole video to another language if I want and have my ad running to another entire demographic. So if you want to try one of the lowest barrier to entry video tools and one of the most customizable video tools on the market right now, head on over to InVideo IE and get started today. Thank you so much to invideo for sponsoring this video if you're a coder we got a big announcement out of GitHub this week in that the GitHub Copilot now has a free tier that you can use directly inside of VS code. If we take a look at the plans here on the GitHub copilot page, their free plan allows up to 50 uses per month and their pro plan, at only $10 per month allows unlimited uses for context. Cursor is 20 bucks a month and Devin is 500 per month to use. As far as code completions, you can get up to 2,000 per month for free or unlimited on their $10 a month plan. So this is big news for coders who are on an extreme budget and only want to use free tools. And since we're talking about GitHub, which is owned by Microsoft, let's talk about a couple of the things that Microsoft rolled out this week, including real time translation inside of live captions. Now you do have to have a Copilot plus PC for this to work, but you can essentially watch videos in any language and it will subtitle it. And it even works with like video calls. You can be talking on a zoom call with somebody else and it will translate in real time what they're saying directly on your screen. And Bing Image Creator also got an update this week using the latest Dall E model called Model PRC 16. It allows users to create images twice as fast, and you can create images directly from within the Bing search bar if you want. They've also made it much easier to share your images once they're generated. But if I go over to Bing.com and tell it to create an image of a monkey on roller skates for me, notice that right at the top, Bing Image Creator starts to create the image. I don't know if it's bogged down or something right now, but every time I'm trying to generate an image I'm running into an error. Hopefully that's something that Microsoft will work out, but supposedly you'll be able to do this right here inside of the Bing search. The Meta Ray Ban glasses got an update this week as well with Real Time Video. So very similar to what we saw with Project Astra and even with the new ChatGPT with vision now your glasses can actually watch a real time feed of what's going on around you. It's also going to allow you to have real time conversations that are translated between languages. So it says Live AI lets wearers continuously converse With Meta's AI assistant, Meta AI to reference things they discussed earlier in the conversation. You no longer have to say, hey, Meta. To get it to start up. It says live AI also works with real time video. Wearers can ask questions about what they're seeing in real time, for example, what's around their neighborhood. It's basically OpenAI's advanced mode with Vision or Project Astra just built into your Ray Bans. I haven't gotten this update yet in my Ray Bans, but I am excited to test this stuff out. It's also got the ability to say Shazam, this song. And whatever song is going on around you, you, it'll tell you what song that is. So some cool new features that are supposedly rolling out now. Since we're talking about Meta, Instagram started teasing their new AI video editing tool. You know, Meta has an AI video generator, and I was talking about it on a recent podcast with Tim from Theoretically Media, and we couldn't figure out what Meta was going to use this AI video tool for. Well, Adam Mosseri here, the CEO of Instagram, kind of gave us a hint.

We're working on some really exciting AI tools for you video creators out there. A lot of you make amazing content that makes Instagram what it is, and we want to give you more tools to help realize your ideas. And you should be able to do anything you want with your videos. You should be able to change your outfit or change the context in which you're sitting, or add a chain, whatever you can think of. Now, all of that you're seeing here are some of the early research models that we've been exploring, and we're hoping to bring some of these into the Instagram app next year.

So.

So keep an eye out. More to come.

All right, continuing on with the topic of AI video, Pika rolled out their new 2.0 model this week, and they have this really cool feature called Scene Ingredients. If I log into my Pika account, you can see there's a new button here that says ingredients, and it allows you to upload multiple scene ingredients, including characters, objects and locations. It's almost like that wisp that we just saw from Google. But for AI Video, let's say I really like this jacket here. I could pull in an image of myself and then the image of that jacket and then give it a prompt, like a man with a beard walking in a blue and orange jacket. Ideally, I'll get what looks like me walking with this jacket on, and, well, I got the guy walking, but it looks to be the original guy. It didn't really swap the face. I do need to play with this Pikachu a little bit more. My last attempt of using ingredients also turned out kind of wonky. If you were watching my live stream, you saw me get this result. But I'm sure there'll be people in the comments letting me know what I'm doing wrong and how to actually make this new Pika feature work. Because when I see some of these other ones that other people have done, like making themselves fly or putting themselves next to the girl with the pearl earring, they all seem to come out really, really good. Which just makes me think that I just need to get better at prompting it. Oh, and by the way, if you want to try out Pika, they're actually making it free for the next few days. They put out this post on X, a giant pre holiday gift from the Pika team. We're giving everyone free unlimited access to Pika 2.0 from today until December 22. Anyone on any plan can generate as many videos as they want using all the scene ingredients they want while it's cold outside at Pika Art. Pretty cool of Pika. The video model Cling also got an update this week here on X they said we're excited to introduce the cling AI 1.6 model. With the update, we have significantly improved the response to prompt, the visual aesthetics and the physical actions, hoping to bring more consistent and vivid results. So this new model has improved prompt adherence, more consistent and dynamic results. It supports standard and professional modes, achieving a 190% overall improvement compared to with cling 1.5 and has a 720pmodel. Now cling isn't one that I've played around with a ton, but we did again make a podcast recently with Tim from Theoretically Media which will be out within the next week or so, where we actually break down all of the various AI video tools and compare them against each other and we put them all on a tier list. And if I remember correctly, Tim put Cling in a tier so he finds it to be one of the better AI video models. Since we're talking about AI video companies, Runway announced the Runway Talent Network, which is a new media platform where you can discover and hire people that are actively using AI to create art and video for whatever project you need help with. This Odyssey company showed off their Generative World model. They also announced that Ed Catmull, one of the co founders of Pixar, is on their board. So more people from the entertainment industry are entering The AI industry on boards of these various AI companies. But check out this generative world model here. It appears to generate what looks kind of like a Nerf or Gaussian splat, all from a prompt. You can see a 1970s interior destroyed by snow. And it generated this. An office interior from the 2000s. An underground workshop with a muscle car covered in a white cloth. And these prompts are generating these 3D spaces here. And then even cooler, here's this outdoor scene. A street corner with a classic American home wide shot. And it generated this sort of 3D video here. A large artway with bright LED windows. A street in London with a brick wall. A small room overflowing with plants. A creepy haunted house lit by a lamp. Like this is pretty ridiculous. This is so cool. This one even has people in it. A family in the kitchen. And it's snowing. And it knows that the snow needs to be outside and not in the kitchen. Slow moving lava, a serene coral reef. I mean, there's a lot of examples here. I will make sure it's linked up below, but check it out. So it is using a Gaussian splat here. And they actually managed to load it into Unreal Engine. So this one looks like a scene that they generated with their tool loaded into Unreal Engine and then then probably loaded this person in inside of Unreal Engine. Now, if you want to get your hands on it, you do have to apply for early access. I did apply myself, and as soon as I get my hands on it, I'm going to show it off because this looks amazing. And in the world of even more cool research, I came across this genesis project this week. It's a generative physics engine that's able to create 4D dynamic worlds. And this one, you just need to see the project webpage to see what it's capable of. Here it did the this character motion generation. Here, check out this prompt. A miniature Wukong holding a stick in his hand, sprints across the table surface for three seconds, then jumps into the air and swings his right arm downward, landing. The camera begins with a close up of his face, then steadily follows the character while gradually zooming out. When the monkey leaps into the air at the highest point of the jump, the motion pauses for a few seconds. The camera circles around the character for 360 degrees and slowly ascends before the action resumes. That was the prompt they gave it. And you were just watching that video as I was reading it. Holy crap. Here's some more. A Japanese samurai performs boxing a Chinese soldier performs Gangnam Style dance. A Roman soldier walks forward like a zombie. And the hope is to be able to use these models to help train robots. It appears. It says Genesis aims to use Generative Robotic Agent and Physics Engine to automatically generate robotic policies and demonstration data for various skills under different scenarios. Here's some other generations of like chocolate going on a candy bar or something. I wish I actually knew the prompts for these, but like this is like a sponge or something twisting and then eventually initially breaking. Some fruit falling, some chocolate melting. A space station as the camera floats through it. A school of fish. I mean some of this stuff that it's showing that it could generate is pretty mind blowing. And look at how many people were involved in creating this project. This is absolutely wild. The paper for it isn't out yet, it doesn't appear. But it's being released as open source under the Apache 2.0 license. And then this demo here is kind of going viral. So check out this prompt. A water droplet drops onto a beer bottle and then slowly slides down along the bottle surface. But it's being released as open source under the Apache 2.0 license. And then this demo here is kind of going viral. So check out this prompt. A water droplet drops onto a beer bottle and then slowly slides down along the bottle surface. And here's the generation. It's showing a Heineken bottle and you can see it follow that water droplet down the bottle. Here's three more prompts. Camera revolves around the bottle while gradually moving downward. A close up shot from above tracking the motion of the droplet. Droplet and camera is positioned on the right side of the scene and tracks the droplet motion. It's creating the same video from three different views. Camera view A, B and C and it generates all three of these with the Heineken bottle following the water droplet from three separate camera angles. Visualize all the forces applied to the droplet and its velocity. And now it's showing the visualization of that droplet with these arrows showing the effects of gravity, velocity, friction and the support force. It then zooms into microscopic mode. Like this is just absolutely wild. I am completely blown away by what I'm seeing. And they also go on to say it is cross platform. So it will work if you have an Nvidia gpu, an AMD gpu, Apple M series chip, or an Intel GPU or cpu. It'll work on anything. I have no idea how to use this yet, but I'm gonna figure it out because this is something I want to play with more than anything I've come across recently. Recently. But I've got to move on from my nerd out fest here. Let's talk about some AI image news. Leonardo got an update this week which, yes, I must disclaim. I am an advisor to Leonardo, so I do own a tiny sliver of equity. But they have some cool new features. They have a few new models. There's a brand new Fenix 1.0 model which is an improved model over their 09 model. And they've also rolled in flux. So if you really like the realism of the flux model model, you can now use Leonardo to generate with flux. So if I switch to the flux model and give it a prompt like a man standing on the ceiling upside down, I can generate images with the realism of flux. And now switching over to the Phoenix 1.0 model. Let's do just a man standing on the ceiling. I'll get rid of the upside down. That's kind of redundant. Let's see what it gives us. We get some, you know, slightly more cartoony looking, but still aesthetically pleasing men standing on the ceiling. Now, I probably should have put upside down if I actually wanted to get the upside down sound like this one. But you get an idea of what it can generate here. Mid Journey also got a new update with their mood boards this week. And this is essentially a way to train Midjourney on a specific style that you want to see it generate over and over again. So if you head over to midjourney.com and log in and click on Personalize, you get this option for Create mood board. So let's create a mood board here. Once you're in here, you can upload images or add from a link or add from a gallery and then generate images that reference the images you trained into your mood board. It's kind of like Midjourney's version of Elora. Let's go ahead and pull in a bunch of images of my face here and see if I can actually generate based on my look. Let's just call this wolf as the mood board. And we'll click copy code. And this is the code that it gives us that we can now add to any prompt. So I'll copy that code. Let's go to create here. And let's just say a man looking out a window along with my little mood board code. And, well, it sort of followed the style of the images, but it didn't really make it look like me. So it looks like it's more for like the color and composition of the images, but not yet to be able to train it on like your own face. But that's mood boards. Magnific rolled out a new AI image model called Super Real which is designed to create ultra realistic images directly inside of Magnific. Kreia rolled out some new features allowing you to to edit images and do some more kind of impressive inpainting and even pull in other images and you know, put jackets on people and things like that. So they actually put like a Balenciaga jacket on this person and put them in Apple Vision Pro goggles by pulling in images of those things and masking out areas. 11 labs just rolled out a new text to speech model called Flash which generates text to speech in 75 milliseconds seconds. So we're going to be able to have really real time speech with AI because of models like this where I can ask AI a question and the response is going to be near instant because these text to speech models are getting so dang fast. This week, researchers at Florida Atlantic University developed an artificial intelligence system that it can actually read sign language. So theoretically you can be on like a zoom call with somebody who's signing to you and actually get the translation of what they're signing mining. So that seems like a pretty cool advancement and good use of AI. The robotic company Apptronic partnered with Google DeepMind. This week, the AI powered humanoid robotics company announced it entered a strategic partnership agreement with Google DeepMind. It'll bring together the best in class artificial intelligence with cutting edge hardware and embodied intelligence. I'm a big fan of Google DeepMind and not just because they flew me out to London and showed me their offices and let me experiment with a lot of their stuff stuff, but because I think they actually do really cool work as well. So excited to see how this partnership plays out over time. And finally, Nvidia this week unveiled a miniature supercomputer.

Our brand new AI computer. The little tiny Jetson Nano, little Orin computer. This thing that's really amazing is that a long time ago, starting with Xavier, you guys might have know that we created a brand new type of processor. It was a robotics processor. Nobody understood what we were building at the time and we imagined that someday these deep learning models would evolve and we would have robots of all kinds. Everything that moves would be robotic. And now here we are, we're seeing all kinds of amazing robots. Robots on wheels, robots on legs, two legs, three legs. And of course general humanoid robotics are Nearly upon us. This is a brand new Jetson Nano Super. Almost 70 trillion operations per second, 25 watts and 200. It runs everything that the HGX does, even runs large language models and can't wait for all of you to try it. It's available everywhere. Enjoy Robotics.

So as you can see, that's designed to power robots and it's under 300 bucks. Super exciting times we're living in. As 2 Minute Paper says, what a time to be alive. That's all I got for you today. I'm imagining things are actually going to slow down the week of Christmas next week. But in the grand scheme of things, the AI world continues to speed up and we're getting more and more huge announcements by the day. I can't wait to see what 2025 holds. There's going to be some big stuff, some stuff I've already been able to see but I'm not allowed to talk about yet. It's going to be a fun year. Hopefully you're hanging out with me on this channel for the ride. I'm going to be got CES in a couple weeks showing off all the latest greatest tech from that event. So if you're not already, make sure you like this video, subscribe to this channel. I'll make sure really cool AI tech videos show up in your YouTube feed. Also I want to mention real quick, I am doing live streams every Single Monday at 11am Pacific Time. I'm calling it the AI sandbox where we're going to get in talk about the latest AI news that happened over the past week. We're going to play with the tools. I'm going to get suggestions on some of the tools. I'm going to show you early access to some of the tools tools and we're just going to have fun with AI for a couple hours and we're going to do it live and you can ask me anything and I will share advice, share tools, help you with automations in your business, all sorts of cool stuff. The last one we did was a blast. People loved it. You could go watch the replay right now and we're doing it again this coming Monday at 11am and ideally every Monday at 11am Pacific. So thank you once again for tuning in. I really really appreciate you. I I will see you in the next video.

Bye bye.