AI News: AI Video Better than Sora?!

In this video, I'm going to break down all of the latest AI news from the past week that I think is important for you to know. I actually just got back in from Las Vegas. I was attending the Cisco Live event this week, so I'm a little bit behind schedule. Meaning this video is probably not going to have the same level of editing as my normal videos do. But obviously the point of these videos is to loop you in on the AI news and not to impress you with me and my editor's cool editing skills. Let's go ahead and jump right into it, starting with something that's super cool, and everybody in the AI world has been talking about it this week, and it is this new AI video generator called Cling. Now, this video generator comes out of China, and supposedly if you have a Chinese phone number, you can actually register for the app and use it right. Now, personally, I've been trying to find a workaround and have yet to figure out how to actually get access to this. But if I do figure out some way to actually use this tool by, you know, getting a Chinese phone number or something, I will definitely make a video and breakdown how you can do that as well. Now, most of the videos we've seen from this tool have been roughly five seconds long. There are a whole bunch of examples of videos that have been generated with this tool, like this horse riding through the desert, or this boy riding a bike, which is actually one of the longer videos I've seen. You can see this video is actually a minute and 31 seconds. And if we kind of scrub through the video, you'll see that the environment changes. It starts to actually have some snow on the sides, and then eventually he's riding completely through the snow. It's pretty good. If you look at his feet, it doesn't look like he's actually touching pedals. If you're looking close enough, you could definitely tell these are AI generated, but they're pretty dang good. Apparently it also has a feature where you can upload an image and then upload a template action, and it will actually make the image dance to that same action. Here's another example where we can see the initial image that was uploaded and then the dance animation that was given, and then it animated that character in the same way. And there's a bunch of dance different examples here that we can click through on their website. But this one seems to be pretty impressive and people are getting access to it. I think it's just hard to get access to it. If you're outside of China at the moment. Here's some more examples that I came across from Angry Tom on Twitter here. A man laying on the beach eating tacos. I mean, that one's looking pretty good. Here's someone eating pasta. Apparently they originally tried to prompt it with Will Smith eating pasta. They couldn't get Will Smith, so they just did a man eating pasta and. And managed to get this. Here's some people eating a watermelon. Somebody walking through a wheat farm. Looks like it could be like a samurai or something. A lion with eagle's wings. That one's looking pretty good. Ants carrying building materials to make a tiny skyscraper. A cute black and white border collie with its head sticking out the car window. Beautiful black cheetah yawning next to leaves and branches. A black and gray dragon flying over. More people eating. I mean, it's really good at generating people eating. But I'm curious, what do you think? Do you think these generations are better than what you saw out of Sora? I don't think any of these videos are actually fooling anybody into thinking they're actual real videos, but I think it's pretty cool. I think it's actually pretty close to Sora. A lot better than a lot of the other video generators we've been seeing lately that claim they can compete with Sora. But I want to keep it on this video theme for a second because we've been seeing a lot of really cool AI video tools being released lately. Another one that I came across is this one called tooncrafter. And I actually found this one because Matved Pro, a good buddy of mine, put out a video about it. And what this one does is you can you give it two frames and it will actually animate between the two frames. So if you know what you want the video to start looking like and you know what the end of the video should look like, it will animate everything in between. We've got this like boat bus thing moving. We've got. I don't know if that's a ring or a gem or something that's glowing and putting off all these glowing lights. But all sorts of cool examples of a start image and a finish image. This one, you just see somebody's head and they turn it a little bit and they actually knew to add a blink in there just to give it a little bit extra animation. And apparently this only works well with cartoon type images. It's not really great for uploading a real image and then another real image and having it animate between the two, but it seems to be pretty good at cartoony anime style videos. What's even cooler about this Tooncrafter is that you can use it right now for free. Over on Hugging Face. I'll make sure it is linked up in the description. You upload your first image and your second image. For example, this first image of a man holding an umbrella and then him taking like another step. And then we look at our animation here and it looks like this man is actually walking down the street with an umbrella. And if you want to run it locally, it's open source and available on GitHub, so you can actually download the code and use this, run it locally on your own computer. Once again, I'll have the GitHub link link below as well because this is a pretty cool AI video tool. Proper Prompter over on X also showed off another tool called Domo AI. This is a really cool tool where you can upload a a video file and turn that video into a cartoon. Like here's a clip from Harry Potter. Here's another clip where they turned this flying car into like a pixel art video. And we can scroll through and see all sorts of other examples that they pulled from Harry Potter and then turn them into sort of cartoon versions of that original scene that was in Harry Potter. Not a tool I've played with myself, but definitely something I want to spend more time with and show off more of what it's capable of. Apparently it even does lip syncing. I will definitely be making another video on this in the future after I've had a little bit more time to play around with it. Proper Prompter also shared another tool recently called Pixverse. They have a new feature called Magic Brush where you can select just a portion of an image and it does a pretty good job of animating just that portion. Here's another example where it's Harry Potter's wand and they have them just animating the hand and the wand. Another one with Elon's face. They made him nod by drawing some little squiggly lines next to his face. Here's one of a rocket shift being animated as it takes off. Here's one of the train from Harry Potter where they actually show the steam coming out and it moving down the tracks. All sorts of cool examples of uploading an image, selecting just an area and having it animate. Now we've had this feature in Runway for a while now, but this actually looks a little bit better than what I feel like you can get out of Runway using the similar feature. So this one's called Pixverse. You can check it out over@Pixverse AI and while we're on the topic of AI video, the Tribeca Film Festival is actually going to screen some AI generated short films that were actually created by Sora as of the recording of this video. The Tribeca Film Festival is actually happening right now, so it'll be interesting to see the crowd's reaction of some of these videos. Is it going to get booed like AI stuff was at south by Southwest? Or are people actually going to embrace the AI videos that are presented? We'll find out soon. Last week 11 labs showed off their new AI generated sound effects feature. Before you can enter any sort of text and have it spoken in any voice, now you can prompt any sound effect you need and have 11 labs generate the sound effect. Here's some cool ones I came across on X1 from Amar Rishi here where he prompted an ogre says stay away puny human. Stay away puny human. To me that sounds like it was straight out of Warcraft 2. And here's my favorite shared by Pietro Scirono, which you could read the prompt right here and let's see what it sounds like. Oh no. I mean, this is some world changing stuff we're talking about here. And since we're on the topic of sound effects, Stability AI also released a similar open source model. They introduced Stable Audio Open, an open source model for audio samples and sound design. This can generate up to 47 seconds of samples and sound effects including drum beats, instrument riffs, ambient sounds, folly and production elements. Here's some of the examples, like warm arpeggios on an analog synthesizer with a gradually rising filter cutoff and a reverb tail. And Blackbird's Song, Summer Dusk in the Forest. That one actually sounds pretty good. Rockbeat played in a treated studio session. Drumming on an acoustic kit. Totally different types of sound effects than what eleven Labs was showing off. The model weights are available over on Hugging Face, so you can download them and use them. Again. It is open source and there are a few Hugging Face spaces. You can also play around with this. Now we can enter a prompt here. Let's go ahead and try. A train blowing its horn while coming down the tracks. And here's what we get out of that. Actually a lot better than I was anticipating. Again, linked in the Description Description this week also saw a couple of really massive events where a bunch of announcements were made. We had the Computex event out in Taiwan where we saw announcements from Nvidia, amd, intel and more. We also had Cisco Live, where Cisco themselves made a whole bunch of AI announcements in a roughly hour and a half keynote. Jensen Huang made a whole bunch of announcements around Nvidia, most of them kind of the same announcements they already made at Nvidia's gtc. But there was a few new details shared. During this event. Jensen once again showed off Earth 2, which is a digital twin of the entire Earth, designed to help better predict climate change and weather and what's going on around the world so that we can get out in front of it. He claimed it can do hyper local forecasting down to tens of meters and it's trained on all sorts of weather data to better predict what's going to happen with the weather next and get in front of potential natural disasters and things like that. He also showed this off, which is pretty darn impressive. And finally, AI models that reproduce lifelike appearances, enabling real time path traced subsurface scattering to simulate the way light penetrates the skin, scatters and exits at various points, giving skin its soft and translucent appearance. Nvidia ACE is a suite of digital human technologies. Jensen also showed off how the compute power that's coming out of Nvidia GPUs is not quite keeping pace with Moore's Laws, but far exceeding it. Now, I've heard some people say this isn't a super accurate comparison, but it's still pretty impressive as a visual nonetheless. Now the other thing to note is that one of the big bottlenecks when it comes to running large AI models is not only the compute, but also the power to run the compute. And Jensen showed off this graph showing that the actual cost has been dropping significantly as well, meaning that with the next generation GPUs that we're starting to get, we're going to be able to get more and more and more GPU compute power with less and less and less energy usage, which is necessary to really continue these advancements that we've been seeing at the pace we've been seeing them. Now, the final really interesting thing that I think came out of Jensen's keynote was where he mapped out their plans for new GPUs for the future. Right now most companies are using the Hopper GPUs. Next up is the Blackwell GPUs. And then eventually in 2627 we're going to get the Rubin GPUs. And he talked about how they're going to release these at a cadence of a new GPU every Year, essentially. Now, the keynote was really interesting. I did watch the whole thing, but there was a lot of overlap with what they showed off at Nvidia's GTC event a couple months ago, with just a few new things, in my opinion, still interesting to watch, especially if you can find one of those sort of condensed super cuts of the whole presentation. Nvidia also showed off this thing they call Project G Assist, which is essentially an AI helper for while you're playing video games. So somebody can be playing a video game and ask questions about what they should be doing or what they're seeing on the screen, which is pretty cool. But it's also something that Microsoft just showed off at Microsoft Build that's going to be integrated into their new AI PCs as well. If you need a Sidekick for playing games, it looks like you're going to have that pretty soon. And after all of Nvidia's big announcements at Computex, Nvidia briefly passed Apple to become the second largest company on the planet. I mentioned in last Friday's news video that they were getting really close. I can see them breaking up into the top two and possibly even becoming the top one pretty soon, but as of this recording, they've actually dropped back down and are number three again. Apple past them, but as you can see, they're very, very close. So we might see a little bit of back and forth going on. It was also announced that Nvidia is going to be doing a stock split. So if you don't want to pay $1,209 per share, Nvidia is about to do a 10 to 1 stock split or a 1 to 10. I don't remember the order, but basically the price is going to be one tenth of what you see here. And current sharehold are going to increase their supply by 10, so the value stays the same for the shareholder. It's just at a much more manageable price for people to invest in. And what does being the third largest company in the world get you? Well, apparently it gets you the opportunity to sign women's chests. Yep, that's an image of the CEO of Nvidia signing somebody's chest after his keynote. You imagine seeing Tim Cook or Sundar Pichai from Google doing something like that. Kind of hard to imagine, but there you go. You can see the before and after of Jensen signing somebody's chest. But Computex wasn't just about Nvidia. AMD made some announcements as well. The biggest announcement coming out of AMD is AMD's next generation laptop processor called the Ryzen AI 300 series. AMD claims its XDNA 2 NPU has five times more compute capacity and twice the power efficiency compared to last gens due to its unique FP16 block or NPU architecture that can process both 8 bit per performance and 16 bit accuracy. I'm borderline starting to laugh here because I don't know what half of that means, but starting July 2024 the Ryzen AI 300 is coming to some of the Copilot PCs that we talked about during the Microsoft build stuff. In some previous videos, the manufacturer Asus also announced some PCs that are going to use the new AMD Ryzen AI processors. Intel also got into the action during Computex and made some announcements including their unveiling of the Lunar Lake client process architecture to continue to grow the AI PC category. Basically all the chip manufacturers are focused on making chips to better process AI. I think that's probably the best way to sum up almost every announcement at Computex. Intel also showed off their AI playground. This AI playground has an image generator that runs locally. It appears to be using stable diffusion models. The images look really really solid that is generating, but when I jump forward a little bit we can actually see the prompt that they used and the model shows Run Diffusion Juggernaut xl. So yeah, they're using stable diffusion, but it seems to be a pretty nice intuitive interface for generating your AI images with a ton of options to really dial in what you're looking for. A lot more intuitive than something like automatic 1111 or comfy UI that have been popular for stable diffusion recently. It also appears to have an enhanced tab which seems to let you do upscaling in painting and outpainting. Inpainting being adding and changing elements within the image. Outpainting being sort of extending and uncropping the image, making it larger and filling in the details as it makes it larger. The tool also has an answer section which seems very similar to Nvidia's ChatRTX. Apparently we can choose which model we're using. This example they're using Microsoft's Phi 3 model, but it puts a chat GPT like large language model directly on your computer that can be run locally without being connected to the Internet as long as you have the model weights downloaded on your computer to chat with. Qualcomm also got in on the action during Computex. They actually have a big event coming up in San Diego where we're expecting more announcements later this month, but they showed off their snapdragon X Elite chips as well as their X plus chips which are also designed for processing AI and are going to be used in a lot of these copilot plus PCs. So if you're keeping track of Nvidia, AMD, intel and Qualcomm have all talked about making new hardware for Copilot Plus PCs to help these PCs specifically run AI processes more efficiently. Now there were plenty more announcements at Computex. These are just the ones that I think people really following closely in the AI world would actually care about. Let's move on to Cisco Live. This is the event that I just got back from and I'm making a whole dedicated video to explain what I learned at Cisco Live. Share the experience I I had share clips from interviews of Cisco executives talking about what they're doing to push the world of AI forward. Lots of cool announcements were made there, but again it does sort of warrant its own video, but at a high level. Here's a quick breakdown of some of the announcements that Cisco made, but before I get into them, it's important to know that Cisco mostly deals in enterprise like infrastructure. So you may not use Cisco directly or one of their tools directly, but you probably use a lot of tools that use Cisco's products. Now there was one term that kept getting thrown around during every single session during Cisco and that was digital resilience. Now digital resilience essentially is referring to a company's ability to handle issues that might arise in the digital world. Things like hacks and cybersecurity and privacy and data integrity. It's really this all encompassing term of is the digital environment that you've created for your business ready for whatever might possibly happen to it. And so Cisco is using AI to sort of improve companies digital resilience. When a large enterprise company runs into an issue, maybe it can't talk to the cloud, maybe there's some issue with the code, maybe it's user error. Cisco is developing a suite of AI tools to essentially know exactly where the problem originated from. They created something called Thousand Eyes, which essentially keeps an eye on all of your digital infrastructure for you, alerts you to problems, and helps you find the source of the problem using AI. So while you may not use Cisco's technology on a day to day basis, as a consumer, the enterprise companies that you are using are probably using Cisco's technology, which is why it's so important that this digital resilience is sort of taken care of and AI is making that easier. Now again I'm going to do a much Deeper dive breakdown of Cisco. This is very sort of 30,000 foot view, but their announcements, while maybe not getting as much fanfare as the consumer facing companies like Microsoft and Google and OpenAI, the pieces that they're working on are still just as vital for our security and safety of AI going into the future. Now, the other big announcement to come out of the Cisco event was that they're launching a $1 billion global AI investment fund. So not only is Cisco trying to develop tools using AI to improve digital resilience, they're also going out and investing in a bunch of tools to help with these same areas. And since we're on the topic of big announcements coming out of events, Monday is Apple's big WWDC event, the Worldwide Developer Conference. This is where they announced the Apple Vision Pro last year. And this year the expectation is it's going to be a ton of AI announcements, possibly even a new revamped AI Siri. Apple's calling their new AI Apple Intelligence, which is clever because it still keeps it AI. It's just Apple Intelligence instead of artificial intelligence. And according to this Bloomberg article, the big announcements are mostly going to beat AI features. Now, I'm not going to spend the next 10 minutes of this video talking about all of the speculation, since we're going to find out in about 72 hours anyway what Apple is actually launching. But pretty much from all of the expectations of all the analysts and people following along, it's going to be a lot of AI. In the announcements from Apple this coming week, remember when I went to the Microsoft Build event and I talked about the recall feature? The feature that sort of is like your history of everything you've done on your computer and you can go back to anything you've ever done, essentially taking screenshots of your computer at a very specific interval. A hacker developed a tool that can extract the data that's collected by Windows New Recall AI. Now, this was done by an ethical hacker, somebody that was actually trying to find and report on the vulnerability. But the hacker did figure out that it's not that hard to exploit. So this week Microsoft put out an announcement about the update on the Recall Preview feature for Copilot Plus PCs. They said they were listening to customer feedback and are making some changes about how this recall feature is going to work. They're updating the setup experience, basically making this recall feature turned off by default. So when you get your computer, if you do want to use it, you specifically have to turn it on. They also say that proof of presence is Also required to view your timeline and search in recall. I don't specifically know what that means, but I'm guessing you have to prove maybe through like the retina scan or something, I don't know. And they're adding additional layers of data protection, including just in time decryption. So the recall snapshots will only be decrypted and accessible when the user authenticates. Even if somebody steals your computer and gets access, they would then need to re authenticate to be able to view what's in the recall. Snapshots are stored locally, so they're not going to the cloud. So Microsoft isn't training on your recall data. You'll always know when Recall is saving snapshots because you'll have a recall snapshot icon in the system tray. Digital rights managed or in private browsing snapshots are not saved. You can pause, filter and delete what's saved at any time. And overall it just seems like they're doing their best to say, look, we are not saving this to the cloud, we're not training on it, we're not doing anything with this. It's your data and it's now going to be password protected and you have to turn it on and we have to verify that there's actually a person at the computer. They're pretty much doing everything they can to say, don't worry, this is a lot safer than maybe we originally made it out to be. And they're even making it safer now. And since we're on the topic, apparently Google likes that feature so much that they're going to roll that out into their Chromebooks as well. Google's Chrome OS chief says that the company is cautiously exploring a recall like feature for Chromebooks. Right now they're referring to it as memory. Now in this section it says can Google eliminate the creepy factor? And pretty much everything they're talking about right here is the stuff that Microsoft just announced that they're going to be doing with those computers. If you remember, my buddy Bilavel Sidhu did an interview with Helen Toner where she basically claimed that Sam Altman didn't run anything by the board. Well, Sam finally responded to those claims, saying he doesn't agree with her recollection of what went down, but not really saying much more than that. This website, the Robot report, claims that OpenAI is restarting its robotics research group. The company shut down its robotics group in July of 2021, but recently started hiring for a robotics engineer. Which is interesting because they have been working very closely with figure AI and this figure one robot. And while we're on the topic of robotics, here's a humanoid robot from unitree showing off what it's capable of. After we saw that borderline creepy video from Boston Dynamics of the robot that sort of unfolded itself, this company gave us an almost equally as creepy way for a robot to stand up and walk around. And then you've got this guy here who's probably going to be one of the first ones the robots take out if the robots ever uprise against humans and take out the people that were cruel to them first. This week also saw a little bit of downtime for a lot of the AI models. ChatGPT, Claude and Perplexity all went down at the same time. One of the best headlines that I saw around this was something to the effect of people actually have to think again as AI bots go down temporarily. But don't worry, they're all back up again. There's a new bill that's trying to get passed in California that a lot of AI innovators are very upset about. The bill is called SB 1047, Safe and Secure Innovation for Frontier Artificial Intelligence Models. Now, the bill creates a Frontier Model Division which would be responsible for setting safety standards for AI models. And since California doesn't actually have the budget to form this, the division would be funded through fees and fines levied on AI developers. The bill is designed to apply scrutiny to models trained with more than 10 to the 26th power floating point operations or flops. And this is meant to denote models that are more powerful than those currently in use. But more and more models are currently approaching this threshold. But the thing that's freaking the most people out is this liability for AI developers. Anyone training a covered AI model, including startups, academics and open source developers, must certify under penalty of perjury that their model will not be used to enable a hazardous capability in the future, including by others. This makes no sense. Developers who don't submit their certification must provide annual assurances to the Frontier Model division, again under the penalty of perjury, that they will mitigate risks of the model or any model built on top of it. Violators could be charged criminally or be subject to civil charges. This would make it difficult to build open source models because developers could never anticipate all the ways others would use their models. This is ridiculous. They're basically saying if somebody goes and takes the Meta Llama model and fine tunes it and figures out how to engineers some prompts that make it do unethical stuff. Well, Meta is the company that's held responsible for it, and people are really starting to speak up about this on social media. Clem from Hugging Face Andrew Ng Yann Lecun recently put out a post about this all of these people that are trying to advance AI are going to be seriously stifled if they really can't produce anything else because they don't know how people might put potentially, possibly theoretically, use it in the future. It just makes no sense. Now I'm going to be fair and say that I haven't done a super deep dive and read all of the legislation myself. I've only sort of been reading about the narrative from the pro AI side, so there could be a lot more to this bill that I'm missing. But from what I understand about it, putting the accountability on the people that created the models and not the people actually using the models does not make a whole lot of sense to me. And that's pretty much all the major news. There's a few other quick things that I'm going to rapid fire at you that I found interesting, like the fact that an artificial intelligence is running for mayor of Cheyenne. There is a candidate named Vic, which stands for Virtual Integrated Citizen, and they're running in Cheyenne's municipal election. Now, I don't think this AI is actually going to end up running because this article does say to be a qualified elector you have to be a person and you have to be registered to vote. I don't believe Vic is a person, so I don't know how that's going to play out. Also, we have the CEO of Zoom saying that he wants your AI clone in meetings. He doesn't believe that in the future humans will have to sit through meetings. AI clones of the human will sit through meetings and then report back to you on whatever you need to know about those meetings. Sounds very similar to the Bumble CEO's idea of AIs dating each other to find out who's compatible for the real person. Also, if you remember the Humane AI pin, pretty much every reviewer who got their hands on one hated it. They're in even more trouble as they're telling people to stop using the charging case that it came with because it may pose a fire safety risk. Humane is getting rid of the vendor who supplies the cases and they're finding a new vendor now, but it's just stacking more onto the saga that is Humane, who by the way is also currently shopping a buyer for the technology. Rumor has it that they're trying to get Samsung to buy Humane. But obviously the initial reviews of the pen, plus the fact now that the charging case can catch on fire, is probably making it look less and less valuable. We also got word this week that now even the Raspberry PI is getting in on AI. Raspberry PI are like these little single board computers that open source and sort of hardware hackers like to play with to build little gadgets from. Well now Raspberry PI devices are going to be able to leverage AI, so you'll actually be able to make little hardware devices with Raspberry PI that also have AI in them now. And finally, the Oculus Rift creator, Palmer Luckey is working on a new XR headset. So he's one of the guys who developed the Oculus before Meta acquired them and then was sort of pushed out of Meta because of his political views and now it looks like he's getting back into extended reality. Apparently he's going to be making some announcements at the upcoming AWE Augmented World Expo which I'll be at. So ideally I'll be in this session and turn around and report whatever he's coming out with because that's kind of interesting to me as well. And real quickly, since we're on the topic of wearables, earlier this week I made a video about various AI enabled glasses, including the first frame glasses from Brilliant. In my review I couldn't get these to work. No matter what I did, I couldn't get them to work. Well the people from Brilliant have actually been in contact with me to tell me that there was a bug in the initial version that went out but it's now been fixed. I just got home and I haven't actually retested them yet, but I will probably be making a follow up about my thoughts now that these claim to be fixed and working, but that's all I got for you today. I mean it was a ton of news. There's a lot happening and next week there's gonna be more with Apple. The week that there's going to be more with Augmented World Expo and later in June we also have a Qualcomm event where they're going to be talking about AI. There's a lot happening right now. It's exciting times. I'm here for it. I'm going to turn around and share it all with you. I apologize for the lower production value of this video, but I figured that being timely and sharing all the up to date news with you was more important than really polishing the heck out of a video. So. So hopefully you appreciate that. I even skipped over some of the news because there was so much. But if you want to really, really stay in the loop, head over to FutureTools IO. The AI news page is updated with pretty much all of the AI news. And make sure you get on the free newsletter where I'll send you just the coolest tools that I come across and the most important AI news that you want to know about. Plus, subscribers to both this channel and the newsletter get entered into monthly contests for cool AI gadgets and software. And you get free access to the AI Income Database, a database of cool ways to make money using AI. So you can find that over at FutureTools IO. Just join the free newsletter. You're gonna enjoy it. It's really good, I promise. Not biased at all. And again, that's what I got for you. You like videos like this? Like it. And I'll make sure more show up in your feed. And if you want to really stay looped in, make sure you're subscribed to this channel. And I will keep you up to date whenever the most interesting and coolest stuff happens in the AI world, as well as some cool tools and tutorials. And I'll just keep you looped in. Don't worry. But thank you so much for nerding out with me. I really, really appreciate it. I will see you in the next video. Bye. Bye.