AI News: Brace Yourself for the Coming AI Storm!

There seems to be a storm a brewing in the AI world. If you remember, it was March of 2023 that some of the biggest AI announcements that just totally broke everybody's brains happened in March of last year. This video is coming out on the first day of March of 2024 and I think we're in for another wild ride this month. So let's get into all of the interesting and exciting things that happen in the world of AI this week. This video is sponsored by LTX Studio and a really awesome new AI generation platform. And I'll tell you more about that in just a few minutes here. But this week saw some huge progress in the world of AI video. I don't really want to dive too deep into all of the announcements that came out in the world of AI video because I did make another video this week. You can find it on my channel. It was called I Was Wrong About AI Video. That video specifically breaks down all of the really exciting AI video generation announcements that we had this week. But for those of you that just want the tippy top of the iceberg and a simple recap of what we saw this week, we got even more video examples from Sora and what it's capable of, including some that Marquez Brownlee got to prompt himself. Like this video of a dog walking in a parking lot and a 3D printer making something here. And a product reviewer surrounded by gadgets. And of course Sora is still managing to blow people's minds, but for the most part, people, people just want to get their hands on it. Like me. I want to play with it. Pika Labs also got an update to their video generation model this week, allowing us to add lip syncing to our videos. You can either do a text to speech model, type in what you want your character to say, or upload your own audio file and it will speak out what came from the audio file. Wait, hold up, hold up, hold up. These things can talk for real. Runway also got some minor updates to their platform with We've got a new UI here in Gen 2 and if we click on the motion brush here to sort of get more control over how we want to animate this image, there's this new auto detect area. So we can actually let the AI detect what areas we want to affect in the video. So if I want it to just animate my face, before I could sort of draw over my face and mask it out, now it sort of finds my face for me, so I can just click that to. To animate just my face. And then watch as it messes with just that portion of the image. We got this research out of Alibaba Group. Another thing that we don't have access to, but looks really cool and I'm itching to get my hands on called Emo Emote Portrait Alive Generating expressive portrait videos with audio to video diffusion model under weak conditions. This isn't just a lip sync where you can upload an audio file and it will match the lips to the audio. And it actually understands the emotions behind the audio as well and causes the facial expressions and the face to move a little more closely with the emotion that comes with the audio. Here's an example of Mira Morati from OpenAI speaking and this image actually using those same vocals.

So check this out and maybe we have several of them and maybe they all have different competences.

Here's another example where this image on the left was uploaded and then a music sample was provided. And then on the right we can see the animation that was generated.

And it's just way more than lip syncing. It's eye blinks, it's the head movement. It really sort of follows along to the emotion of the audio that it was given. And there's tons of examples. I can't really play them all because a lot of it is copyrighted music and I will get in trouble. But. But I will link up this research in the description. And once again, I did do a deep dive video on all of the latest AI video announcements that have come out recently called I was Wrong about AI Video. So check that one out if you haven't already. We also got news this week out of Light Tricks, who happens to be the sponsor of this video. If you're not familiar with lightrics, they're the company behind really cool tools like Motion Leap, VideoLeap and PhotoLeap. Well, now they've got LTX Studio, a platform that lets us generate full videos. With AI I'm talking about, you give it a single prompt and it generates multiple scenes with multiple shots within that scene. Consistent characters, controllable lighting, music, sound effects. Everything that goes into creating a video you can generate with one prompt. And once you have all of your scenes and all of your shots, you can edit any piece you want on a shot by shot basis. You have complete control. So if you want to make a green car, read simply prompt it and your green car is now red. Need to rearrange shots within your storyboard? You'll be able to quickly rearrange any clip and put it wherever it needs to be within your video. This is a tool that prides itself on consistent characters, consistent lighting, and can even do text and titles. So if you have an idea for a story and you want to bring that story to life with AI Video, LTX Studio can help you do that with a complete A to Z video creation platform. This is unlike anything we've seen before and really, really shows off how quickly this AI video space is moving. What we're able to do these days is put pretty dang mind blowing and LTX Studio is making it easy all in one single platform. Right now. You can get on the wait list to join LTX Studio when it's available and I have a special link in the description that'll even give you early access. So make sure you use the link in the description for this video and as soon as it's ready, you'll be one of the early ones to play with LTX Studio. Thank you so much to Lightrix and their new LTX Studio platform for sponsoring this video. We've got so many more cool things to share around LTX Studio and, and I'll be talking about them a lot in future videos because this is a platform that I'm really, really excited about. And jumping back to the topic of Sora, for just a second here, this was an interesting news story where Tyler Perry saw Sora and then decided to cancel the creation of an $800 million movie studio that he was working on. So he was ready to invest almost a billion dollars into expanding his studio, saw what Sora was capable of and went, I don't know if this is still the future of the film industry. Let's put a pause on this. This week also saw some research come out of Google DeepMind called Genie, which is an AI that can generate sort of platformer games. Here's some of the examples on the page where you can see these little sort of NES era looking platformer games. And these were all created by giving the AI a bunch of video footage of people playing through games. And then it figured out how to recreate these games without any additional extra input. They were able to start with reference images like this image here and generate a movable character in this world. They were even able to upload, you know, children's art and cutouts and turn those into little playable games that you can play as well. Heck, they even did it with real world images. Now you can tell these are very small resolution and, and they're only running at one frame per second, so you probably wouldn't want to play these games yet. But the fact that it can learn from just a sample of other video games that are out there, create a new game from an image or from a drawing, and then actually put a movable character in that world, pretty impressive. It's the worst it's going to get right now. And if you saw how fast things like Dall? E got to where they are with Dall? E3 now, and how quickly we went from things like Model Scope to where we're now seeing Sora, this is just that early moment for these little platform video games where, who knows, a year from now, some of these could look like they were generated with Unreal Engine 5. Who knows? If you asked me a few months ago, I would say that was a huge stretch, but at the pace that things are advancing and the exponential curve we're on, I don't think it's out of the question anymore. If you want a deep dive and see some more examples of what this genie is capable of, of course I'm going to link the research up below so you can check out some of these videos. But hold on to your papers, my friend, because one of my favorite channels, 2 Minute Papers, just did a breakdown video called DeepMind's new AI makes games from scratch, where he does an almost 8 minute breakdown of this model so you can see more of it. I'll likely do a deeper dive myself in a future video, but if you want to learn more, that's a great start. What a time to be alive. And since we've been talking about multimedia so far in this video, let's talk about Adobe and how they just revealed a new generative AI tool for music. During the Hot Pod Summit in Brooklyn, Adobe unveiled Project Music Gen AI Control, a platform that can generate audio from text descriptions like happy dance or sad jazz. Here's a little sneak peek of what it can do.

I can take an input melody that sounds like this.

And then you can generate the accompaniment. In this case, some film music.

That's cool. That's just from a prompt. Yeah, yeah. Do you want to hear something else? Yeah, let's do hip hop. All right, one sec.

Okay.

Check out this other example where I can control the intensity up or down.

Oh, yeah, no, that was intense.

The article here says that Project Music Gen AI Control may be made publicly available at some future date, but for now it's firmly in the research stage. But something like this would be so cool for background music on things like YouTube videos, podcasts, or even your generative games that you might be making. I love seeing all this cool AI research that's in the works, but I'm always bummed out that we can't get our hands on it and play with it ourselves yet, because that's what we all really want. Let's jump over to AI art for a second because Ideogram is just released IDEGRAM 1.0, which is a huge improvement over the previous version of Ideogram. It's so much better at realism. It's so much better at adding text to your images. I got a little bit early access. I've been playing with this one for a couple weeks now, but it does some really good work. I log into my account here. You can see some of the stuff I generated with 1.0. I generated some images that say, subscribe to Matt Wolf on YouTube. You can see all of the spelling is right. And it actually generated four images. You can see my prompt was a group of wolves holding up a sign. It says, subscribe to Matt Wolf on YouTube. Here's another one that it made. Nailed the text. Another one, once again nailed the text. And the fourth one. So all four of the iterations that it generated nailed the text. I did ask it to generate the word Mr. E flow spelled out in the clouds, conceptual art, 3D render, and well, it missed the F. It says Mr. Elo. And if you look at all of the generations, all of them made the same mistake. It got the concept right, just spelled the word wrong. But if we look at some of the other images that people have generated here, we can see that it could generate famous people in this cartoony style. It does really cool high contrast colors similar to what Mid Journey is capable of. And it's even gotten so much better at realism. Videogram did add some paid plans, but the free plan still allows you to generate up to a hundred images a day. And since we're on the topic of AI image generation, if you remember last week I showed you that Google ran into some issues where people would ask for like historical images and it would get like the race wrong of the image. Well, Google came out and apologized. They said, Gemini image generation got it wrong. We'll do better. We've acknowledged the mistake and temporarily paused image generation of people in Gemini while we work on an improved version. So they realize that when you're asking for historical figures, they should be the actual ethnicity of the historical figure that you asked for. Now, when we can expect these changes to be made, Google claims they should be fixed within weeks. So I would imagine within a week or two of this video coming out, we'll get Gemini images again and we'll go back through and test it and see if we can actually get it to generate what we want it to generate. We're starting to hear more and more AI news bubble up out of Apple as well. Since 2017, we've been hearing that Apple was building the Apple car. Rumor had it that it was going to be a competitor to like a Tesla. However, this week Apple finally announced that project is dead. They're moving on past the Apple car. They're not going to continue to work on it. How is that relevant to AI? According to Bloomberg, most of the team's workers that were working on that car are moving over to Generative AI initiatives. It sounds like there might be some layoffs at Apple over this, but for the most part they're just sort of reshifting priorities and instead of focusing on these autonomous vehicles, they're putting the focus into Generative AI instead. And Tim Cook even even further confirmed that by saying that Apple will break new ground in generative AI this year. Apple had their annual shareholder meeting this week and during the company's Q1 earnings call, Cook said Apple was working internally with Gen AI, but it was taking a slower, more deliberate approach to customer facing incarnations of the technology. Indeed, Apple's only briefly mentioned Generative AI in its recent press conferences and announcements, such as when it introduced new autocorrect and text prediction features in iOS last fall. The rumor is that we're going to get a big upgrade to Siri this year where if you use an Apple device, Siri might work a little bit more like what we expect from something like Chat GPT. Apple is also said to be exploring AI powered features to allow users to automatically generate presentation slides and keynotes and playlists in Apple Music, as well as generative AI powered coding suggestions in xcode, the company's app development platform. We also got the news this week that Microsoft was partnering with Mistral AI, the French AI company that has built some of the best open source AI models we have available today. Mistral AI will have access to Azure's cutting edge AI infrastructure and cloud computers while further boosting Microsoft's absolute dominance in the AI space right now. If you're keeping score, Microsoft invested in OpenAI, they invested in Meta and Meta's open source large language models. They've partnered with Hugging Face and now they've partnered with Mistral, who has both closed and open AI models. So it just seems like Microsoft is sort of spreading the deck, diversifying their risk. They're working with OpenAI and their closed model, they're working with Meta and their open source models, they're working with Mistral, who has both open and closed models, and and they're working with Hugging Face, which is becoming sort of like the central hub for where all of this AI development is happening. Microsoft really has their toes in every single pond. So no matter which sort of scenario plays out in the AI world, whether these large closed models win or open source wins, Microsoft wins regardless, because they're all using Microsoft's compute, and Microsoft hedged their bets and sort of bet on all potential outcomes. And since we're on the topic of Microsoft, Microsoft this week also rolled out Copilot for Finance. This is really a suite of AI tools inside of Microsoft 365 that helps the people that manage the finances for a company. So it helps financial analysts quickly conduct a variance analysis in Excel using natural language prompts. It simplifies the reconciliation process in Excel, provides a complete summary of relevant customer account details in Outlook, enables customers to turn raw data in Excel into presentation ready visuals and reports. So if you're somebody that sort of manages the financial side of a business, this is probably something you're going to want to look into. A lot of these large language models are getting new resources to train their models on as well. This week it came out that Tumblr's owner is striking deals with both OpenAI and Midjourney for training data. Owner of Tumblr and WordPress.com is in talks with AI companies Midjourney and OpenAI to provide training data scraped from users posts. At the moment this is sort of still rumor. This is based on an anonymous source and neither OpenAI or Mid Journey have actually confirmed this. But there is a lot of content on WordPress.com and Tumblr sites that these platforms can potentially train on. Although let's be honest, WordPress.com and Tumblr have become a lot less relevant in recent years, which is also probably part of the reason why Tumblr, Tumblr and WordPress.com want to sort of monetize this data because it's a new play, a new rabbit hole that they can go down by allowing their data to be trained into AI platforms so they can make more money because their platforms seem to slowly be going downhill. And I'm talking about WordPress.com, not, you know, self hosted WordPress.org blogs. That's a different story. Don't think those are going downhill. The Hosted blogs you just don't hear about nearly as much as you used to. And then there's Google. They seem to be working with Stack Overflow to help provide Gemini Cloud with training data. As you might already know, Microsoft owns GitHub. So Microsoft has all of the data that's inside of GitHub that it can use to train the various models that Microsoft has available to them. At one point, Stack Overflow launched Overflow AI, but I don't think it ever really gained any traction. Not many people talked about it. It wasn't really that big of a hit. And Stack Overflow also, let's be honest, kind of feels a little bit like a sinking ship. People don't really need to go to Stack Overflow as much to try to debug coding problems as they used to, because now they have chatbots, they've got GitHub, they've got GitHub, Copilot, they've got Amazon Codewhisper, they have all these other ways to debug that. People just go to Stack Overflow a little bit less than they used to. Stack Overflow needs to figure out a way to stay relevant. And if people aren't going to go to their site, well, maybe they can sell the data that's on the site to someone like Google and actually continue to monetize what they've built in this new era of AI that we're entering into. I talked about this one a few weeks ago, but OpenAI is currently in a lawsuit with the New York Times. The New York Times is claiming that OpenAI regurgitated one of its articles verbatim. OpenAI claimed, well, you had to really, really sort of dial in the prompt and even give it some copied and pasted sentences from that original article to get it to prompt your article verbatim. It says here the New York Times used deceptive prompts to get chatgpt to regurgitate its content. So OpenAI is asking the courts to dismiss this case because of that and the fact that normal people do not use OpenAI's products this way. This week, the CEO of Klarna claims that he's using AI to do the work of 700 people after he actually laid off 700 people back in 2022. The CEO explained that in the chatbot's first four weeks, it had handled about 66 of its customer support, or about 2.3 million chats, and it even scored an equivalent in customer satisfaction when compared to humans, even exceeding humans in some cases, the CEO said. So while we're happy about the results for our customers, the our employees who have developed it, and our shareholders. It raises the topic of the implications it will have for society. And this is one of those cases where an AI legitimately has replaced about 700 jobs. And we're going to hear more and more stories like this in the coming months and years, in my opinion. It's definitely one of the downsides, one of the sort of realizations that I struggle with around AI personally. But at the same time, I think everybody should be aware that this is happening. Everybody should be preparing for this. At the end of the day, you're so much better off knowing what's coming, knowing what tech exists and being looped in than getting completely blindsided when it does knock you out. It's a sad fact that a lot of people will lose their jobs because AI can replace them. But I'm really, really hoping that us as humans, as humanity, find new ways to provide value to other people. And finally, this video has been circulating on X about this transparent laptop display from Lenovo. And I find this really interesting. I think the tech is really, really cool. Like, it's awesome to see this laptop and a transparent display through the laptop. It's very futuristic, very cool looking. However, I have no clue what the use case for this is like. Why would you want a transparent laptop display? We already have transparent monitor. We saw them at CES this year. They're great for like storefront windows that you can turn into monitors. They're great for display cases and things like that. But what's the use case for a transparent laptop monitor? To me it looks like it would just be harder to read. You lose all privacy. Cause if somebody's sitting across the table from you, they could just see right through it. And I don't know, it just feels gimmicky. I love it, it looks cool. But I don't really want one because I don't know what I would use it for or what the actual real use case of it is. I thought I would show it off because I thought it was interesting and well, it looks cool, but maybe you have an idea of what that could be used for. Let me know in the comments if you know what a transparent monitor on a laptop would really be used for because I'm really struggling to come up with the best use case. And that's it. That's what I got for you this week. It's been another really, really big week in AI. The previous two weeks before this were a little bit bigger. A little bit crazier news came out of it. This is another one of those sort of marginal weeks where we got some little updates, updates here and there. But again this video is dropping on March 1 and last year March was a whirlwind. So we're about to head into some exciting times. The storm is just getting started. I'm so excited. I'm here for it. I'm going to be reporting on it, telling you about it, keeping you looped in and I'm really excited to see what rabbit holes this takes us down. One quick announcement. I am going to be at GTC in San Jose. That's Nvidia's event. They have both an in person event and a virtual version. Costs money to go to the in person event. The virtual version you can access free and if you register you're entered to win an RTX 4080 Super GPU from Nvidia. I've actually got it here sitting next to my desk right now, still in the box. If you win, I'm going to send this exact unit to you. All you have to do is register for the virtual event. Using my link in the description, I will get the names of the people that registered from Nvidia and we will pick a winner around the time of the event. So if you're not registered for it, it's totally free. You could win an Nvidia 4080 super just by registering for it. I will make sure again that link is in the description. Also, if you haven't already, check out Future Tools IO where I curate all of the coolest AI tools that I come across. I sort them for you so you can drill down to just the tools that you're most interested in. I keep the AI News page up to date on pretty much a daily basis and I have a free AI newsletter where if you sign up for the newsletter I I'll also hook you up with the AI Income Database, a database of really cool ways to make money using AI. It's totally free. You just gotta join the free newsletter over at FutureTools I.O. and I will hook you up. And if you like this video, you want to stay in the loop with the latest AI news, latest AI tutorials, research and just nerd out with me about really cool AI and tech stuff. Make sure you like this video and subscribe to this YouTube channel and I will make sure more videos like this show up in your YouTube feed. Thank you so much for tuning in today. I really really appreciate you. Thank you so much to LTX Studio and Light Tricks for sponsoring this video. That is a really cool platform that I am super excited about. Check that out. Get on the waitlist if you haven't already. And thanks once again for tuning in. I'll see you in the next video. Bye Bye.