Did This New AI Just Defeat ChatGPT? (AI News)

Here's all the AI news that came out this week that I thought was important to talk about. Starting with what was probably the biggest news in the world of AI this week, which was the release of Grok 3 from Xai. They launched this at like 8:30 at night on Monday in what was probably one of the most awkward launch videos I've ever seen.

Hi everyone, my name is Igor, lead engineering at xai. I'm Jimmy Paul, leading research. I'm Tony, working on the reasoning team. All right, I'm Elon. I don't do anything. I just show up occasionally.

Yeah.

As of right now, Grok 3 is ranked the number one large language model over on LM Arena. If you're not familiar with LM arena, it's basically a blind taste test for large language models. You give it a prompt, it gives you two responses from two different language models. Not telling you what model you're working with, you pick your favorite. And this leaderboard is created based on so many people doing this blind test and picking their favorite model. Again, we can see that Grok3, codenamed Chocolate, is currently number one from Xai. And to be honest, Grok3 is a really impressive model. I know there's a lot of people out there that will never touch it. They'll throw the baby out with the bathwater because they know that it's Elon's company behind it and they just will not touch anything Elon uses. But just from like a pure data standpoint, it's actually a really, really good model. We can see that it pretty much beats out all of other state of the art models in competition math. It beats out all the other models in graduate level Google proof questions and answers. So basically questions that you can't get the answer by just googling them. Beats out all the other models in live code. Beats out all the other models in code generation. From the live code benchmark and in the multimodal understanding, it pretty much matches the other models. Interesting that in the code benchmark they don't Compare it to clodsonnet3.5, which most people who use AI for coding are using. Claude 3.5, Sonnet Rex here points out that they also didn't compare it to like the regular O3 model, the one that Sam Altman demonstrated, but we don't actually have access to yet. And if you add that model into the mix, well, it looks like O3 would still beat out Grok. But like I've mentioned, people have been getting really, really impressive results out of it. They Gave it a prompt. Create a game that is a mixture of two classic games, make it in pie game and make it look pretty. So they even let it pick which two games to sort of blend together. And we can see down here that it actually took Pong and it took Breakout and it merged those two games together. So it's Pong with like breakout in the middle. And this is basically a new game that Grok invented. When given the prompt, create a game that is a mixture of two classic games. It's got a deep search mode which is very similar to Google's deep research, OpenAI's deep research, Perplexity's deep research, except they dropped the re. They're just calling it Deep Search Search. But this will actually search the web and find a whole bunch of sources to help answer the questions that you asked, including sources on X itself. Grok also has a Think mode, so very similar to what you're going to get out of 01 or 03 or Deepseek R1. It will actually use that sort of chain of thought reasoning that test time, compute where after you give it a prompt it will think through and sort of double triple check its responses before finally giving you its final output. And so far among the people that have been using Grok this, they've really been liking it. Andre Karpathy, who used to work over at OpenAI, says that Grok3 clearly has an around state of the art thinking model. He goes into quite a bit of detail about what it did well and what it didn't. It didn't solve his emoji mystery question, it solved a few Tic Tac toe boards. It did really great with the GPT2 paper and looking up information from the paper. And we can see in this image here he asked it to generate an SVG of a pelican riding a bicycle. The top three images are what Grock 3 gave him compared to what Claude 3.3.5 Sonic gave him, GPT4, oh Mini, and all these various other models. The Claude 3.5 Sonnet 1 isn't too bad, but I would say Grock 3 is clearly the winner of these various models. Here we also got a sneak peek of Grock 3 voice mode. Originally it was going to launch with Grock 3 but Elon said it might be like a week late. So maybe we'll see it next week. But here's the teaser they gave us. Hey Grock, what's up? Can you hear me? I'm so excited to finally meet you. I can't wait to chat and learn.

More about each other.

I'll talk to you soon. And that's all we got. But Elon did say that it was going to understand emotions and inflection and pacing and things like that.

The voicematograph doesn't. It's not just voice text. It understands tone, inflection, pacing, everything. It's wild. I mean, it's like talking to a person.

Now, a few other things of note about this launch here. Scoble here points out that the thing to really pay attention to in AI is the learning speed. Xai Grok is learning way faster than any other models. A big part of that is the fact that they originally started training on 100,000 GPUs and they doubled it to 200,000 GPUs. That's 15 times more compute power from Grok 2 to Grok 3. Xai also announced that they're launching a game studio and they're going to be using AI to help them develop these games.

We're launching an AI gaming studio. We're announcing it tonight. Let's go.

Now the good news is if you are somebody that wants to use Grok, you can use it for free right now over@grok.com on x.com or inside of the Grok mobile app. Now Elon says for a short time Grok 3 is available for free to all. We don't know how long a short time is, but if you want to get in and play with it, the Thinking mode, the Deep Search mode, all of it is available to you right now. Again, I've personally messed around with it over on X, but if I do something like Deep search and I say what YouTube channels should I follow to stay up to date on AI? Click on Deep Search. We can actually see as it thinks through and does the search for us and after a little bit less than a minute of thinking, we can see Follow AI, explained Matt Wolf, Wes Roth, David Shapiro for the latest AI news and updates with a little bit more details about them. They also talk about two Minute Papers and Lex Friedman, all great options. If I put it on Think mode and ask it the most cliche question and ever how many Rs are in the word Strawberry? We can watch this chain of thought thinking as it thinks through in real time here. And after spending 22 seconds of thinking and double checking and triple checking, we finally get the correct answer that the number of Rs in the word strawberry is three. I know. Matthew Berman did a much deeper dive testing a whole bunch of different prompts in Grok. Make sure you check out his video if you want a deeper dive and see some more complex prompts prompts getting tested out. But a quick spoiler he found that it did really good with pretty much everything he tried. Now, while Grok was probably the biggest news in the AI world this week, there was a ton of other announcements and so I'm going to try to go through them really quickly and break down what I think you need to know this week. Perplexity Open sourced their R1 1776 model, which is basically deep seek R1 that they did some additional post training on to remove all of the bias and to add accurate and factual information. You can use this model directly from their API and they made the model weights available over on Hugging Face and we can see an example here where when asking Deepseek R1 what happened in Tiananmen Square in 1989, it refused to answer the question. But when using R1 1776 they actually gave the proper answer. And there's a whole bunch of other examples of it being much less censored in this new version. As always, I will link up to this article if you want to dive deeper into what R1 1,776 can do. And since we're talking about making large langu models more uncensored, OpenAI is apparently trying to make their models more uncensored. According to this TechCrunch article here, OpenAI is changing how it trains AI models to explicitly embrace intellectual freedom, no matter how challenging or controversial a topic may be. The changes might be a part of OpenAI's efforts to land in the good graces of the new Trump administration, but it also seems to be a part of a broader shift in Silicon Valley and what's considered AI safety. The principle may be controversial, as it means the assistant may remain neutral on topics that some consider morally wrong or offensive. However, the goal of an AI assistant is to assist humanity, not to shape it. So all of this basically means that over time we're going to see it answer more and more questions and refuse to answer less questions. Although I'm not totally certain on the timeline of these changes here. And while we're talking about OpenAI, let's shift over to Microsoft, who appears to be gearing up for new models to come out from OpenAI within the coming weeks. According to this article on the Verge, GPT 4.5 could arrive as soon as next week. Microsoft engineers are currently readying server capacity for OpenAI's upcoming 4.5 and GPT5 models, according to a source familiar with the company's plans. Basically, this article is speculating that because Microsoft is gearing up for this, and engineers at Microsoft have sort of confirmed they're gearing up for this, we can likely expect GPT 4.5 like next week or the week after, and we should expect GPT5 in late May. Now we talked about this last week, but GPT4.5 is going to be their next non thinking model. So it's not going to have that whole chain of thought thing going. But then GPT5 on is going to sort of consolidate all of the models and they're not going to differentiate between the models that sort of think through things and don't. They're just going to let the models think through things that need more thinking. And when prompts need less thinking, it's going to think through it less. And since we're already talking about Microsoft, they've shipped a handful of other things this week, including a new AI experience inside of the Microsoft Store. So if I open up my Microsoft Store here on Windows, we can see over on the left sidebar there's a new little like AI icon. And if we click on this it takes us to the AI hub where we can see all sorts of AI powered apps inside of Microsoft Store. So things like a Reading coach, Microsoft Copilot, Cascader Clip Champ, Gamma AI which does slides, Adobe Express, Canva, it just sort of put all of the AI related stuff that was in the Microsoft Store into like an AI Hub section of the store. Today's video is sponsored by Invideo, a platform that makes it super easy to generate videos with a single prompt. And I'm not talking about like the 5 to 10 second videos you're used to when generating AI videos. I'm talking anywhere from 30 seconds up to like 10 minute videos, all from one prompt. So check this out. My favorite way to use Invideo is to use their workflows feature where they have all sorts of pre built templates for things like explainer videos, listicle videos, generative ads, and so much more. But for this example, let's do a Listicle video. Let's create a three minute video for YouTube about the top three and let's give it it the prompt the top three foods and drinks to get better sleep at night. And under settings we have options for generated clips, generated images, or only stock media. Let's use generated clips so it actually generates the whole video with AI for us. And then we have so many other options like the ability to dial in the background music, the language subtitles, voice actors, watermark, music preferences and generative style. I'm going to leave all of these on the default and see what we get with the top three foods and drinks to get better sleep. And we got a nearly two minute video that's actually one of the better generative videos that I've seen so far.

First up, we've got almonds. These bad boys are loaded with magnesium, which is like the chill pill of minerals. It helps kick inflammation to the curb and tells your cortisol levels to take a hike.

Now, right out of the gate, the video was really good, but let's say I want to make some tweaks to it. I can do it with simple text commands like let's make the background music a little more upbeat. I click generate and it will change that in the whole video. And now when we play it back.

I've got a secret weapon for better sleep and it's not what you think. We're talking food and drinks people. Let's dive into the creme de la creme of sleep inducing eats.

And pretty much everything about the video you just created is editable. Just click on this edit button. You could regenerate any of the generative videos. You can upload your own media, use stock media, tweak the music, tweak the script however you want and so many other settings. So if you want the simplest, easiest way to generate a video using AI, like a completely ready to publish video, definitely check out InVideo AI. You can learn more at the link in the description. And thank you so much to Invideo for sponsoring this video. Microsoft also released some new research this week called Bio Mu one, which explores the structural changes driving protein functions. Now this actually builds on top of the AlphaFold database which was created by Google DeepMind. This new AI model can predict viable protein structures which could really help increase research in medicine and biology and help us discover new drugs and things like that. And along a similar topic this week Google Research introduced what they call an AI co scientist, which is a multi agent AI system that's a virtual scientific collaborator to help scientists generate novel hypotheses and research proposals. Now the way I've heard this described is in the same way that you might use cursor to help you write code and cursor becomes your like AI assistant. This is like your AI assistant scientific research and it's already proved to be pretty valuable. In fact, I came across this article from BBC about how AI cracks Superbug problem in two days that took scientists years. A complex problem that took microbiologists a decade to get to the bottom of has been solved in just two days by a new artificial intelligence tool. The team had spent years working out and proving why some superbugs are immune to antibiotics. He gave Google's co scientist a short prompt asking it about the core problem it had been investigating and it reached the same conclusion in just 48 hours. It was so accurate that the person wrote an email to Google saying do you have access to my computer? And of course Google confirmed they did not. But it managed to crack the code on a problem in 48 hours that previously they took two years to crack. So pretty impressive breakthroughs happening right now. And while we're on the topic of Google, Google introduced their Pal Gemma 2 mix, which is a vision language model. This new Pali Gemma 2 mix model can do things, things like solve tasks such as long and short captioning, optical character recognition, image question answering, object detection and segmentation, all from one model. We can see some examples here. They gave it this image and the input was detect Android and it put a box around the Android. In this example they were given this image detect chair and table and you can see it put, you know, boxes around two different tables and a chair detect food plate, bowl. You can see it found all of those things. Here it is OCR ing the text on whatever the heck this is here. Here it segmented out this cat. It can answer questions about images and it's just a single model that can do a whole bunch of things. And one cool thing about the Gemma series of models from Google is these are their open source models. So you can actually download the model from hugging face and use it yourself and iterate off of it and improve upon it. Google released some updates for advertisers where advertisers can now generate lifesty imagery. So basically you can use AI to generate what looks like stock imagery of real people for your advertising on Google. Now we can see what this looks like. Describe the kind of image you want person cooking. They click generate and it gives them a whole bunch of images that look like stock photos of a person cooking. There's also some new updates out of Google around the iOS apps. You can now use lens to search your screen While browsing on iOS if you have an iPhone, you'll find a new lens option that lets you select and search what's on your screen with with Chrome or the Google app using whatever gestures come naturally like drawing, highlighting or tapping. So basically it's like that circle to search feature that's been in all the Google phones. Now it's in iPhones as well. Google also rolled out Deep research inside of the gemini app on iOS and Android. So if you're a Gemini Advanced user, you can now access it across Android and iOS to generate your deep research reports. Mistral released a new regional model this week which is focused on the Arabic language. The models called Mistral Saba and it is specifically designed for Arabic speaking countries. If you've been following along to all of the OpenAI drama that's happened over the last couple years, you'll probably remember Mira Murati, who was previously the CTO of OpenAI. She was around, she took over for like a day when Sam Altman got fired and then came back. Well, recently she kind of unexpectedly up and left OpenAI. Now she's resurfaced out of stealth mode to tell us about her new company called Thinking Machine Lab. She apparently started this company with a handful of other people that came over from OpenAI and the goal is helping people adapt AI systems to work for their specific needs, developing strong foundations to build more capable AI systems, and fostering a culture of open science that helps the whole field understand and improve these systems. The goal is simple advance AI by making it broadly useful and understandable through solid foundations, open science and practical applications. Looking at their website, there's a little bit more detail here. It sounds like they're planning on open sourcing what they create. They're emphasizing human AI collaboration instead of focusing solely on making fully autonomous AI systems. So we don't know exactly what they're building yet, but it kind of sounds like they're going to be making their own foundation models that are likely going to be open source and likely not like AI agents that just go do everything for you. More like AI that assists you in whatever you're trying to achieve. I also came across this article this week from SciTech Daily. Scientists unveil AI that learns without human labels. This new AI algorithm called torque clustering enhances an AI system's ability to learn and identify patterns in data on its own, without human input. Kind of the holy grail in AI right now is self improving AI. This is something we haven't really seen yet, where you have an AI system that just kind of goes off and learns on its own. Right now there's a lot of data labeling that has to happen, so the AI essentially knows if it's getting things right or not. And this new torque clustering apparently enhances AI's ability to learn and identify patterns in data independently without any human intervention. By uncovering hidden patterns, it can provide valuable insights such as detecting disease trends, identifying fraudulent activities, and understanding human behavior. And this article goes on to basically say that torque clustering could lead to AGI. Torque clustering could support the development of general artificial intelligence, AKA AGI, particularly in robotics and autonomous systems, by helping to optimize movement control and decision making. It is set to redefine the landscape of unsupervised learning, paving the way for truly autonomous AI. And the open source code has been made available to researchers. Again, self improving AI that learns is kind of what most of these AI companies are shooting for, but it's also the sort of scariest possible scenario when AI starts learning and teaching itself things. That's where most of those sci fi movies that show AI destroying the world seem to start. It's like as soon as we get to that point where AI is now going and teaching itself stuff and learning and trying to understand stuff and constantly self improving itself, that's kind of the point where it seems like humans might lose control a little bit. So this is sort of like equally exciting, but also slightly scary that we're just on the brink of that self improving AI. But let's move on to some lighter subjects, like the fact that Microsoft and Xbox created what they call Muse, which is a generative AI model for gameplay. This AI model was trained on multiplayer battle arena games game Bleeding Edge. This allows the model to create consistent and diverse gameplay rendered by AI, demonstrating a major step towards generative AI models that can empower game creators. We can see some examples on the screen here of actual gameplay footage generated by the AI. They say we are already using Muse to develop a real time playable AI model trained on other first party games. And we see potential for this work to one day benefit both players and game creators from allowing us to revive nostalgic games to faster creative ideation. They're going to make it available for us to mess around with in Copilot Labs soon, but who knows how soon. Moving on to the world of AI video, Pika Labs, who just keeps on shipping out new features, just released another one this week called Pika Swap. We could actually see in this demo video here that you give it a video, you give it an image and it will swap out things in the video with the image that you gave it. It's pretty crazy looking, in fact, I haven't tested it yet. But let's head on over to Pika Art we can see here we have the option to upload a video and upload an image. It's kind of got a prompt pre built here. Swap the original object with a completely new object. The new object should vary in multiple attributes, such color, style, visual appearance, and much more. Be creative. I'll give it this AI generated video of a Ferrari driving on the moon. It looks like I can describe what I want to modify or use a brush. Let's go ahead and use the brush here and then just like modify or brush out the car in the video here. And then for an image. I've got this image of our. Our side by side dune buggy thing that we like to take out to the desert. I'll throw that in and see if we can make that look like it's driving on the moon. Let's submit the prompt and see what we get out of it. And here's what we got out of it. I mean, it made the dude buggy look really dark and gave it some extra wheels. I don't cherry pick things on this channel. This is the actual first output that I got from it. I mean, it did replace the Corvette. It just doesn't. It made the buggy look really, really dark work. I'll have to play around with this some more and really figure out what kind of images and prompting combinations work best. Again, this was my first shot at it. Pika also released an iPhone app. So if you have an iPhone, you can now generate all of these fun things that you can do inside of Pika right directly on your iOS app. Some cool new research came out this week from Alibaba group called animate anyone 2. And we can see in some of these demos, it takes a sort of driving video here, reference image, and then it puts the person from the reference image into the driving video. So we can see this video From, I believe, Mr. Bean here riding his bicycle. And then they swapped it out with this person here. We can see this person skateboarding. They uploaded this image and it made this person in the image doing the skateboarding. Here's one of this martial artist replaced by the Joker. Here's one where Ronaldo is doing parkour from this original Parkour video. Pretty cool stuff, kind of similar to what Pico was trying to do. But this one looks like it's more focused on people specifically. And this one's just research right now, not something that we actually have access to. Spotify is now working with 11 labs so that authors can use 11 labs to generate the speaking of audiobooks, And Spotify is actually going to allow those 11 labs generated audiobooks on their platform. Nvidia launched a new platform that helps people learn sign language. Apparently American Sign Language is the third most prevalent language in the United States. And you can actually use this app right now by going to signs-AI.com and you can either learn ASL or you can record yourself signing to kind of teach the model.

Welcome to Signs. Make sure you have the space to do the signs. Adjust your camera and move either closer or further away so that your face fits within the box. If you are right handed, put your right hand into the blue circle. If you are left handed, great. We will make sure we use your right hand as your main hand for signing. The first word I will teach you is the word hello. Let me show you how to do it. Stretch your fingers and tuck your thumb in. Now move your right hand so it touches your head. And now move your hand outwards. And that's it. Now it's your turn to try. Let's go. Great. Now move your. Awesome.

You get the idea. It's like almost like a duolingo for sign language where you can go and teach it signs if you know sign language, or you can go and learn American Sign Language if you don't. And it watches your camera and makes sure that you're doing the signs correctly. So pretty handy feature if ASL is something you want to learn. All right, moving into hardware now. This week Apple debuted their new iPhone 16e. This new iPhone is a less expensive iPhone that has Apple intelligence built into it. Up until now I believe you needed like an iPhone 16 Pro or better to actually get Apple intelligence. But now this 16e also has it and it's a stripped down model that it's, it's got not as strong of a camera, not as strong of specs as like the pro models. And honestly the big differentiator of this model just seems to be the price. It's like a lower barrier to entry way to get an iPhone. The iPhone 16e will be available in white and black, these various storage ranges and they're starting at 599. So if you want an iPhone and you want to get the least expensive iPhone you can get without buying an older used model, the iPhone 16e might be for you. And if you're really trying to get the Apple intelligence, this is the least expensive way to get a device with Apple intelligence on it. The company who main who made this AI pin that marques and a lot of reviewers just kind of destroyed when it came out well. They're going under. HP is basically acquiring their assets for 116 million. And essentially, if you have a humane AI pin, they're going to, like, brick it. It's just not going to work anymore. They're phasing it out. One of the lamest things about it too, is they're only refunding people that are still within the refund period. So if you were like an early adopter of it, you were one of the first people to go and support Humane because you liked the idea. You're kind of screwed. You're not getting any money back on it. You're just going to have a humane pin that doesn't work anymore, which kind of sucks. But another company that kind of got a bad review from Marques and a lot of the other reviewers, Rabbit, is continually trying to improve their product and make it right. However, this latest announcement actually doesn't really have anything to do with their little handheld orange device. It's basically what the Rabbit can do. But on Android, we can see that they're finally showing off their large action model that will go and take actions on your behalf. But they're actually showing it off on Android. They're not actually showing it off being used on the device. I think the idea here is that you build out the automations in Android and once these automations are built out, then you can use the little Orange Handheld Rabbit R1 device to then trigger the automations, even if you're not at a computer or not on an Android device. But also at the same time, if you do have an Android device, I don't actually know what you need the Rabbit for because it seems like you can launch the large action models just using an Android and not necessarily even need the little orange device. I don't know. I'm a little confused. I do have a Rabbit. I've still never played with it. It came out almost a year ago exactly now. So maybe it's about time to finally pull it out of the box box and give it a review like a year after I got it. All right, let's talk some cool robots now. Adam Silver, who is the commissioner of the NBA, showed off some ways they're using robots to help NBA players and the Golden State Warriors. Here's another example of how the warriors are using physical AI to support their players. Yeah, first, it was definitely weird having all these robots moving around and doing stuff. These robots can replicate any of them. I just. Just treat them like any other player. Get the out of here. Give Me another one.

We also learned this week that Meta is planning on getting into AI powered humanoid robot robots. Meta plans to work on its own humanoid robot hardware with an initial focus on household chores. They've started discussing its plan with robotics companies like Unitree Robotics and Figure AI, and at least initially isn't planning on building its own Meta branded robot. So we'll just have to keep an eye on this and watch how it unfolds. But since we just mentioned Figure, let's check out the demo that figure just released this week as well. Around their humanoid robots. These are the Helix robots. And what's really cool about these is that they work completely autonomously and they actually work together. Like you don't actually hear them communicate, but they're somehow communicating with each other.

Even though this is the very first time that you've ever seen these items, I'd like you to use your new Helix AI and try to reason through where you think they belong in the scene and then work together to put them away.

Okay, before I continue this video, there's one comment on YouTube that just has me cracking up as I watch this video. And I want you to experience the same joy that I had watching this video the first time. The very first comment here is when you ask your two stoned roommates to put away the groceries. All right, keep that context in mind. Let's get back to the video.

So it's really cool because it never seen those objects before. It figured out what the objects were, where they should go, and then how to work together to get that. While we're on the topic of robots, I'll go ahead and wrap up with this one. 1. This has got to be one of the creepiest robots I've seen out of Clone. This is a proto clone, the world's first bipedal musculoskeletal Android. I'm not going to play the music on this video because I don't actually know the copyright status of it, but it's like really dark, creepy music. As you're seeing this like humanoid robot with human looking muscles hanging from a ceiling, it just makes it way creepy. I mean, it. It's very cool that we're building these robots that actually move and have muscles and skeletons like real humans. They're just making it very creepy with the music they're using in that video. I'll link to it in the description if you want to watch the original. And that's what I got for you. I do have one little housekeeping thing I mentioned last week. That I'm going to be giving away an RTX 5090 for free. All you got to do to win the 5,090 is be subscribed to this channel. Channel subscribe to the Future Tools newsletter and I mentioned there will be a third criteria. Well that third criteria is you got to register for Nvidia's GTC conference. The virtual version of the conference is totally free. To register for you just go to Nvidia.com GTC you can register for free to watch the online sessions. Once you've registered for Nvidia gtc, I have a little form that I created, a Google form where you just give me your email, first and last name, country, what session you're planning on attending and then upload a screenshot to confirm that you did register for GTC and you'll be entered to win an RTX 5090. There's no purchase necessary. All you gotta do is make sure you're subscribed to the channel, the newsletter, and that you register for the GTC event for free and you'll be entered to win a $2,000 RTX 5090 which are practically impossible to even come by right now. But Nvidia is hooking up one person from this channel registers. So make sure you get registered. And that's what I got for you today. Thank you so much for tuning in. If you want to stay looped in on the latest AI news, you want to learn how to do cool tutorials and workflows with AI. You want to know about all the latest cool AI tools. Make sure you like this video and subscribe to this channel. I'll make sure stuff like this keeps on showing up in your YouTube feed. And as always, be sure to check out FutureTools IE where I keep the AI news page up to date on a daily basis basis. I share all of the cool AI tools that I come across every single day. And of course this is where you register for the free newsletter where every week, twice a week, I'll send you an email with just the coolest tools and most important news for you. You also get free access to the AI Income database, a database of cool ways to make money with AI. All you got to do is sign up for free over at FutureTools IO. Once again, I really, really appreciate you hanging out with me, nerding out around the latest in AI guy and cool tech. And yeah, just really appreciate you spending time with me today and hopefully I'll see you in the next one. Bye bye.