AI News: Real-Time AI is HERE (and It's WILD!)

All right, let's break down the news. Last week on Black Friday I got a new 3D printer. And then on Suburb Monday, I might have gone a little bit overboard on buying different colors of filament. I also just got back last night from a two and a half day trip to Disneyland. This was before Elon Musk's comments about Bob Iger. But you're probably not watching this to hear the Matt Wolf news. You probably want the latest in AI news. And so. So on that front, this week marks the one year anniversary, slash birthday of the launch of ChatGPT. ChatGPT was launched on November 30, 2022. And in that one year time we've gotten a mobile app, we've gotten code interpreter, we've gotten Dall E3, we've gotten GPT4, we've got voice prompting and so many other features. It seems crazy to me that it only came out a year ago because it's come so far in that short amount of time. Also in OpenAI news, Sam Alt is officially back. Now. The news came out that he'd be coming back last week, but it wasn't made official until this week. He was reinstated as CEO. Mayor Moratti was reinstated as CTO. Greg Brockman was reinstated as President of OpenAI. Greg Brockman posted a write up on the blog, basically saying thanks to everybody who works at OpenAI, including the board, for all of their hard work that they've put into it so far. Also mentioning that he wants to build a more diverse board and improve the governance structure. And one of the new board members is going to be somebody at Microsoft. However, there's a catch. They're going to be a non voting observer member, meaning they'll know what's going on with the board and be involved with some of the decision making, but they don't actually get a vote when it comes time to make the decisions. But at least they won't be blindsided again like they were last time. All the drama went down. We don't actually know who at Microsoft is going to hold that board spot. A spokesperson for Microsoft actually declined to comment on who from the company would fill that observer seat. There is a little bit more to the story from all the drama that unfolded last week. I reported in last week's news video that there was talk of this rumored Q Star model which was supposedly getting much better at doing math, which had huge implications towards the ability to possibly break encryption. Well, Sam actually did an interview this week with the Verge and during that interview they asked reports about the Q Star model breakthrough that you all recently made. What's going on there? Sam Altman's response? No particular comment on that unfortunate leak, but what we have been saying two weeks ago, what we are saying today, what we've been saying a year ago, what we were saying earlier on, is that we expect progress in this technology to continue to be rapid. Also that we expect to continue to work very hard to figure out how to make it safe and beneficial. A lot of people are speculating that the way he worded this, without actually denying it, is somewhat of a confirmation of this Q Star's existence. You would think that instead of saying it was an unfortunate leak, he he would say those rumors are just not true. That's not what he said. Now, I don't know if this fully verifies its existence, but it definitely seems to make it a lot more likely. Also this week we Learned that the OpenAI GPT store is going to be delayed until next year. Today they sent out this email that at the end it says in terms of what's next, we are now planning to launch the GPT Store early next year. While we had expected to release it this month, a few unexpected things have been keeping us busy. I wonder what things those could be. In the meantime, we have some other great updates to chat GPT soon. Next up, I want to tell you about the sponsor of today's video, Notion. Now you may already be familiar with Notion, but they've added some amazing new AI features into it, like their new Q and A feature. This is amazing because you can ask it any question and it will use your entire Notion database as context to help answer that question. So for example, I personally use Notion to keep on top of all of the various sponsorships I do. I've got YouTube sponsorships, newsletter sponsorships, both primary listings and secondary listings. I've got featured listings on future tools and I need to keep all of these sponsorships organized and clear in my head on when and where they're all supposed to get published. And Notion is perfect for this. And with their new AI Q&A feature I can find any information I need in a snap. So check this out. Down in the bottom right hand corner there's a little Star button. If I click this, I can start chatting with my entire Notion database. For example, I can ask what is the status of the upcoming Notion ad Placement in the Future Tools Invent Inventory Database and there we go. The upcoming Notion Ad Placement in the Future Tools inventory is ready for production and is scheduled for YouTube on December 1, 2023, which should be the day this video is going live. I can ask how many placements are sold in the Future Tools Inventory Database for December? I get a response that there are three sold ad placements in the Future Tools Inventory Database for December. They're scheduled for December 1, December 13 and December 22. Are there open featured listing slots in the next few weeks for the Future Tools Inventory Database? Yes, there's featured listing slot for December 25th in the future Tools Inventory Database. I wonder why that date's open anyway. The new Q and A feature and all of the AI that they've built into Notion makes Notion such a no brainer. Now again, I use it to keep on top of all of my sponsorships, as well as to keep an ongoing list of video ideas and the status of the various ideas as they're being created. Notion has become my online hub to keep everything in my creator business organized, and with their new AI and Q and A features, it's gotten even more powerful. You're already a Notion AI user. You can get started today with Q and A already in your Notion workspace. If you're interested in trying it out, join the waitlist to get early access. The link is on the screen and in the description below. Thank you once again to Notion for sponsoring this video. Now, as the drama over at OpenAI seems to be kind of coming to a conclusion, a new drama is popping up over at Stability AI. According to this Fortune Art article, Stability AI is looking for buyers as investors are pressuring the CEO to resign. According to this article, CO2 management one of the investors in Stability AI called for CEO Imad Mustach to step down in a letter to Management last month. The demand came just a year after CO2 led a funding round that initially valued the startup at $500 million. Cotu wrote that Mostak's leadership had prompted several senior managers to leave and place the startup in a tenuous financial position. Supposedly Stability approach to both Cohere and Jasper as potential acquirers of Stability AI. However, Ahmad took to X to say in response to my inbox deluge, I'd like to note we have never reached out to anyone to be acquired. Jasper is an awesome partner and Cohere are doing great work. We work with many strategics. More announcements to come. Stability has simple governance Open rocks Referring to Open Source, the first reply says so in a nutshell, Stability AI is stable, and Ahmad responded relatively do we have another AI ousting in the works? Well, we'll have to Let the drama unfold and see. But speaking of stability AI, they're still shipping. In fact, this week they introduced SDXL Turbo, which is a real time text to image generation model. Now I thought this was using the lcms, the latent consistency models, but apparently it's using a model called Adversarial Diffusion Distillation, which is not something I'm super familiar with, so I can't talk super intelligently on it, but I will dive deeper and learn more about it, so maybe I can break it down in a future video. However, it does seem to be similar to GANS or Generative Adversarial Networks, which is also a topic that I'll save for a later day. I have talked about it in previous videos, but don't want to dive too deep into this video about what that tech means. Instead, I'd rather just play around with the tool. And if we head over to Clipdrop, co slash Stable dash, Diffusion Dash Turbo, we get a box to enter a prompt. If I click in here, you can see it creates a box for us. And now we can prompt whatever we want. And as we type it will actually generate what we're typing on the screen. A blue wolf standing on a mountain top. And as I type you can see it keeps on updating the image in real time. Birds in the sky. Now you can see that some birds in the sky. So as I type more, it keeps on updating the image for me. However, if you're not on one of the Clip Drop Pro plans, you'll be able to generate a couple images and then it'll shut you out and say you can't generate any more. But that's cool. I got something for you. One of my favorite AI generation tools is Leonardo AI and they just released a very similar new feature called Live Canvas and Javi Lopez here actually has a really cool demo of it. You can see his prompt is Isometric House, Age of Empires, a waterfall with mountains in the back, house surrounded by forest. And he actually started by dropping in. This waterfall image had just sort of a green background. And then he's drawing this house here and then over on the right. As he draws, you can see the house is updating in real time. As he adds more colors to it, it all gets added and updated to the image on the right. And this is actually something that anybody can use for free. If you head on over to app Leonardo AI slash Live Canvas, you can enter a prompt down in the bottom, just like I was doing over on Clip Drop. A blue wolf Standing on a mountain top. As I'm typing it, you can see it updating in real time. Now, you'll notice over here, there's two dropdowns. Real time mod, interactive mode. You do have to be a premium member to do it in real time mode, but interactive mode is close to real time. It's almost just as fast. And now that I've got a blue wolf standing on a mountaintop, if I wanted, I can start drawing over here and steer the direction of the image a little bit. So let's go ahead and get like a gray color and let's sort of draw a mountaintop here. And as I scribble on this image, you'll see more stuff coming into view. Let's add maybe another mountain in the background, and I'm just scribbling stuff. Let's maybe give us a blue sky. Make it daytime. You can see as I change the background color on the image, it changes the background on my image on the right here. And just dragging the colors around keeps on tweaking the image just by dragging it around in real time. Let's make it a green background instead, or even a red background. Look at that. That makes it this cool red nighttime. Now let's go back to brush, maybe see if we can get, like a sun shining up here in the top like this. Wonderful. Notice that that's, like, supposed to be the sunshine. There we go. Now it's starting to look like the sun. Now it just kind of turned into, like a yellow sky, which is cool too. Add some more color to it. Sure. Let's bring in Hinks. Maybe we can make it look like a sunset. And there we go. We can keep on updating this image all we want, and it will update it in real time. This is super, super cool. Speaking of Javi Lopez, he actually released his own tool called Magnific, which is an AI image upscaler. But it also adds additional details and creativity to the images that you UPSC. For example, here's one of my YouTube thumbnails that's supposed to depict Leonardo DiCaprio with like a green glow in his eyes and a city burning in the background. Now, I did upscale it, and I will show you the upscaled version as I move this slider on the left over. And you can see it added more of a cityscape in the background. You can see it sort of added window lights. You can see the hand gets updated as I scan over the hand and there's more detail in the fingers and the jacket. But where it's really noticeable is if you look at the forehead, the eyes, the mouth, the hair, all of the facial details. Look at the difference, difference it makes. There's without the upscale, there's. With the upscale, the face and the eyes and the mouth and the hair up here, there's just so much more detail added in when you see the upscaled version. Here's a sort of low res image of my face that I uploaded here. If I bring this line over, you can see how much detail it adds back in. It's pretty freaking phenomenal. And there's a whole bunch of extra stuff that you can do over here. You upload your image right in here. At the moment you can only scale it up by 2x, but you can keep on adding it in and scaling it by 2x over and over again. You can optimize it for different things like portraits or video game assets or 3D renders. You can add a text prompt here, change the creativity strength, the HDR strength and the resemblance strength. And then there's some other experimental features here that you can play with as well. It's really, really impressive. Now, speaking of upscaling, I talked about a company called Crea AI last week, which is another sort of real time AI image generator tool. Well, this week they also rolled out their own version of Upscaling called Upscale and Enhan, which if I'm being quite honest, looks almost identical to what Magnific just rolled out. If we take a peek inside of Crea here, you can see you upload your image on the left here. You have the AI strength, you have an upscaling factor. You can plug in a resolution, a prompt and a negative prompt. So Magnific seems to have a few more features, but Crea looks like they really might have modeled Magnific when they rolled this into their platform. It is pretty good though. Here's a very low resolution image of a frog here. If I bring my line across here, look at how much it enhances this image. It is really actually impressive. Here's an image of a cake here. If I scroll across, look at how much more detailed that fruit becomes on the cake. There's some lemon and some water. I scroll across here, you can see the details just really get enhanced. You can find this one over at Kre AI. This one I believe is still in beta. They're rolling it out slowly to people. But if you have access to kre, I believe you have access to this. Again, I don't know if it's Kre or Kreia. I think I've been using those interchangeably. But if you have access to it, you should have access to this upscale inside of it as well. Now, now, speaking of AI art, Microsoft actually rolled Dall E3 inside of Microsoft Paint. And I can actually open up my own version of Microsoft Paint on Windows here. And you can see I've got a new button up here that says CO Create. If I click on CO Create, it opens up Co Creator Preview. It asked me to describe what I'd like to create. Let's do a humanoid robot holding a basketball. And then I get a handful of style options like charcoal, ink, sketch, digital art, pixel art. Let's go ahead and do digital art and click Create. And as it creates down here, you can see my credits went down by one. So there is a credit system and you only have so many credits. Don't totally know how that works yet know how many credits you get or when they refresh. I just started playing with this, but you can see here's some images that it generated. If I click on one, it moves it over to the left side and I can do all of my normal paint stuff that I would normally do inside of Microsoft Paint with this image. Amazon's also getting into the AI art generation game. However, they're kind of coming in from a different perspective. It created a tool called Titan Image Generator and it's actually designed for developers to build AI image creation apps. So it's not really supposed to be a front end app where anybody just gets in and creates creates AI images. It's designed to make apps that create AI images. It also adds invisible watermarks to any of the images that are created and they offer legal indemnity if you get into any copyright issues with any of the images generated from their Titan model. And speaking of Amazon, Amazon also announced Amazon Q, not to be confused with Q, which is the thing that a lot of people speculated caused Sam Altman to get fired before getting rehired. Amazon Q is actually an AI that looks at a company's database of all of their available information so that employees can get answers to questions, solve problems, generate content and take actions using the data and expertise found at their company. So it's meant to be sort of an internal AI for your company that is trained on all of your company's data for the employees. This week we also saw a sneak peek of where AI video is going. In fact, Pika Labs gave us a sneak peek demo of what they've been doing with their app and it generates some amazing AI video clips, like really, really good cartoon and Pixar style videos here. But not only that, it's got some other really cool features that we haven't really seen in AI video yet. Things like taking a video and turning it into a different style video, or out painting, like expanding to different aspect ratios just like we do in Photoshop, generative fill. Or you can even change things like this woman changing her crop top here, or this monkey getting sunglasses. This is one of the coolest advancements we've seen in the world of AI video in a long time, and I'm excited to play with it. I should have my hands on it next week and I'm hoping to make a demo video of it for you and show off more of what it can do. But while we're on the topic of Pika Labs, Pika also just raised $55 million. This is a company that's somewhat flown under the radar until fairly recently. And then out of nowhere they've got one of the best video generation platforms out there. $55 million round was funded by Lightspeed Venture Partners with participation from Homebrew, Conviction Capital, SV Angel Ben's Bytes, and notable angel investors including quora founder Adam D' Angelo and ex GitHub CEO Nat Friedman, both people who were somewhat sort of involved in all the OpenAI drama last week. This new model that they're showing off is called Pika 1.0 and it's looking pretty dang exciting. Now, when it comes to generative AI video, there's really two companies that are leading the way, Pika Labs being one of them, the other one being Runway ML. And this week Runway ML rolled out some new features as well. There's now more camera moves. If we jump into Runway ML here and we click on camera motion, look at all the options we have. Now. We've got horizontal movement, vertical movement, panning movement, tilt movement, roll movement, zoom movement, and there's even added styles now where you can give it a 3D cartoon style or an advertising style, or a duotone or a sticker style. So much cool stuff in Gen 2. And it's kind of a bummer because earlier this week I made a video about how I was trying to make an AI video movie trailer. And as soon as I release that video, better tech comes out for making AI videos. So expect an update to that process pretty soon because these tools are really going to up the game there. And in the last note about AI video, this week the AI video startup, hey Gen, is launching A near instant avatar generator and they just raised $5.6 million in funding. Now, previously, hey Gen had a tool that was one of the best video translation tools I've ever seen. You upload a video in English or whatever language you're starting with, you tell it to translate it into a different language and it changes the video to that language, but it makes your lips move so it looks like you're saying the words in that language. It sounds like your voice. Now it looks like you're going to be able to create new avatars in about 5 minutes using iPhone video. Now, back when this company was called Movio before they rebranded to HeyJo, they reached out and asked if I wanted to have a custom avatar made for me. And in order to do that, I had to send them a whole bunch of green screen footage and it was going to take their team quite a while to generate that avatar for me. Now it looks like you're gonna be able to do it in just five minutes and do it yourself. So this tech is just mind blowing how fast it's moving along. This week Meta also had a couple of research announcements. One of them being this audiobox research which they are not making publicly accessible yet. Yet it's only available for researchers and select institutions. But you can give prompts like a river running and birds chirping and it will give you a output that sounds like this.

It also has this really cool infill feature where you give it an original audio file, select a section of the file and actually replace the audio in that section. So check this out.

That's the original audio. This is the masked out portion of that audio. The text prompt is then given river running and then a dog barks several times and then this is the output.

So it added additional audio to that part that was asked to infill. Again, only available for researchers right now, but looks pretty promising. Meta also released a suite of AI language translation models that preserve the expression and improve streaming and this is one that we actually can use. If you come to this research page, I'll make sure it's linked in the description and scroll all the way to the bottom. You can try the expressive translation demo here and you can see they have some examples.

So glad you are here. I am so happy to see you.

You can see it maintained the voice and the way she said it, but it changed the language.

Please don't leave. I don't want to be here alone.

It's pretty cool. I personally still think hey Jen does a better job at this because it actually changes the way your lips move. So it looks like you're saying it in the new language, but it's still pretty cool. And they do have a demo that you can try. You select a language, it looks like you can do English, Spanish, French, or German right now. Give it the input language, the output language, and then record. Now it won't let me record. Court I've already tried. It says your mic is in use, your camera's in use. You can't do this right now, but anybody could go play with this if they want to. Again, it's linked up in the description. In the world of AI art, Chinese courts make a landmark ruling AI images can be copyrighted. Now, this is not the case over in the us. In the US they've ruled that AI images cannot be copyrighted. The Beijing Internet court ruled this week that a set of AI generated images possessed originality and reflected human human creative intellectual input and should be recognized as works protected by copyright law. The court ruled that because it was a human being who set the relevant parameters for the AI model and ultimately selected the image in question, the final output is directly generated based on their intellectual input and reflects the plaintiff's personalized expression. This ruling is in stark contrast to the stance of Western courts, which thus far have rejected copyright protection for AI generated images due to the lack of human authorship. On this one, I got a side with China. Honestly, I think they got this right. I've said this before in past videos, and I'll say it again right now. When I create thumbnails for this YouTube channel, it often takes me longer than the creation of the video itself. I often create tons and tons and tons of iterations of the image. I change the prompt over and over and over again until I get it dialed in the way I want it. I do in painting. I do out painting. I use Adobe's generative fill to fix areas. I change colors and boost contrast and do all sorts of things things before the image is ready and it becomes a thumbnail. A lot of work goes into some of the AI generated images. And under the current US Copyright laws, my videos are copyrightable, but the thumbnails that I create technically are not, despite the fact that I put so much effort and work into them. This week, the US, Britain, and 16 other countries signed an agreement to make AI secure by design. Now, the agreement itself is fairly vague, basically saying that, hey, AI could be dangerous. We should all agree that we're approaching this safely. Does everybody agree to that? Yes. Okay. All 20 countries agree we need to approach this safely. It says the agreement is non binding and carries mostly general recommendations such as monitoring AI systems for abuse, protecting data from tampering, and vetting software suppliers. So again, no real guidelines. Now this is something really cool. Google DeepMind's AI dreamed up 380,000 new materials Google DeepMind researchers say they've expanded the number of known stable materials tenfold. Some could be useful for everything from batteries to superconductors if they make it out of the lab. So this is a project called GNOME or gnome. I'm actually not sure if the G is supposed to be silent or not, but basically this AI was trained using data from the Materials project, which is a free to use database of 150,000 known materials. And using that information, the AI system came up with designs for 2.2 million new crystals, of which 380,000 of them them were predicted to be stable, which means not likely to decompose or explode and thus be plausible candidates for synthesis in a lab. So this can be really, really huge and I'm looking forward to diving deeper into this and possibly even making a dedicated video about this because the implications are pretty massive, especially in the world of finding new batteries, superconductors and even possibly new like building materials. But again, I don't want to deep dive too much into it, into this video. So it's something that I need to do more research on and will likely do a de deeper dive in a future video. And finally, to wrap up this week, everybody seems to be talking about AGI. Nvidia CEO Jensen Huang said at the NYT Deal Book Summit that he believes AGI will be reached in five years. Now, one of the things about AGI is that everybody that you might ask seems to have a different definition of AGI. So for example, Jensen Huang's definition is exhibits basic intelligence fairly competitive to a normal human human. While Elon Musk seems to be a little bit more optimistic thinking that we're going to achieve AGI in the next three years.

If you say smarter than the smartest human at anything, can write as good a novel as say J.K. rowling, or Discover new physics or invent new technology, I would say that we are less than three years from that point.

He actually defines it as smarter than the smartest human. Slightly varying definitions there. And yes, this was the same interview where Elon told Bob Iger how he really feels about about him. Anyway, that's all I got for you this week in the world of AI news. Again, I was at Disneyland for two and a half days, had to come home and do a whole bunch of catch up on the news and I'm sure there's a few things I missed. Anything that's really big that I missed I will make sure is talked about in a future video. But this is the stuff that I came across in my deep dive research today. Hopefully you learned some new things and found out about some interesting AI news that you hadn't heard of yet. If you love nerding out about AI as much as I do, make sure you check out FutureTools IO. This is where I curate all the coolest AI tools that I come across. I keep the latest AI news up to date on a daily basis here and I also have a free AI newsletter where I'll email you twice a week. Once with all the latest cool AI tools I came across and once with all the latest AI news that I think is important for you to know. It's 100% free. You can find it over at FutureTools IO and if you want to see more videos like this, show up in your YouTube feed. Give this video a thumbs up and subscribe to this channel channel and I'll make sure that you're always in the loop with the latest AI news, tutorials, breakdowns, research and all of the latest advancements in that world. Thank you once again notion for sponsoring this one. Anyway, thanks for hanging out. Thanks for nerding out. I got some exciting videos coming next week that I can't wait to share with you. So stay tuned and I will see you in the next video. Bye bye.