The AI News You Most Likely Missed This Week!

Oh, hello. Didn't see you there, Sorry, that was really dumb. There's actually quite a bit to talk about this week in the world of AI, but there was one big event that kind of overshadowed all of it, so if you weren't paying attention, you probably missed most of this stuff. But we might as well start by addressing the Apple in the room. This week was what the nerds call Dub Dub or wwdc, the Worldwide Developer Conference from Apple. Now, I already made an entire video about what happened at wwdc, so I'm not gonna go into too much dep depth about that in this video. That video looks like this and you should definitely watch it, especially if you want details about all of the various AI things that they talked about, as well as details about the new Vision Pro, which is rolling out next year. But long story short, it's a very cool looking mixed reality headset, but it's $3,500. We won't get access to it till next year. And it only has two hours of battery life. So if you were to watch the WWDC event on an Apple Vision Pro, you would have ran out of battery before the event was even over. Shortly after the event, Meta released an internal memo about their thoughts on Apple Vision Pro. Some of the highlights of what Zuck said about it was the good news is that there's no kind of magical solutions that they have to any of the constraints on laws and physics that our teams haven't already explored and thought of. They went with a higher resolution display and between that and all the technology they put in there, it cost seven times more and now requires so much energy that you need a battery and a wire attached to it to use it. He went on to say, we innovate to make sure that our products are as accessible and affordable to everyone as possible, and that is a core part of what we do. He talked about how his vision of the Metaverse is fundamentally social and by contrast, every demo that they showed was a person sitting on a couch by themselves. He claims that they have a real philosophical difference in terms of how they're approaching this, basically saying that he is optimistic about the Meta Quest future after seeing what Apple is putting out. Also at the Apple event on Monday, there was a bunch of announcements about additions being added to iOS 17. Many of them were AI additions. Scratch that. I mean, many of them were machine learning additions. Because Apple didn't say the words AI once during the entire presentation, they acted like it was a dirty word that should never be uttered. One of the machine learning add ons was that voicemails would be transcribed in real time so you can read who is actually leaving you a message and if you want to answer it, you can. This is a cool new feature to iPhone, something that Android has already had, but being an iPhone user myself, this is something cool to see. They made it so you could turn your own images into stickers and use them as reactions to text messages you get. They also added automatic transcriptions to audio messages in text messages, which I think is a great addition because I really really dislike getting audio messages. They're also using AI stuff slash large language models to improve autocorrect on the typing feature. So basically the more you use the keyboard, the more it learns the way you type and gets better and better at autocorrecting what you're trying to say, which has led to this funny meme circulating Twitter so what did you do with generative AI? Microsoft we built a copilot for the web. Google we created Bard. Apple, we fixed autocorrect. They're also using generative AI to add extra slow motion scenes inside of backgrounds on the iPad. They also improved PDFs on the iPad where you could take a picture and it will automatically find form fields that you can fill out and make them fillable.

IPad is the best device for PDFs thanks to new machine learning models. IPADOs can identify the fields in a PDF so you can use autofill to quickly fill them out with information like names, addresses and emails from your contacts. This even works on documents you scan with a camera.

Going into this event, a lot of people were speculating that there might be some sort of update to Siri because Siri was one of the first AIs that a lot of people used and when they finally got to mentioning Siri, here was the big announcement. Instead of saying hey Siri, Siri can now be activated simply by saying Siri. And with that, my foe just turned on Siri on me. And that was pretty much all the AI related news to come out of the WWDC event. Again, I did make an entire video breakdown of wwdc. I highly recommending watching that for a deeper breakdown of everything they covered. And now let's get into all of the stuff that also happened this week that was kind of overshadowed because of the Apple event. So this was actually news that dropped last Friday. But for Friday I was traveling back from the Augmented World Expo and I also record these videos on Thursdays. So this didn't make last Friday's video But last Friday, Google announced some new research called Style Drop. This allows users to generate new images that follow a specific style of their choice given only a single style reference. Now, this doesn't seem to be available for everybody to use yet. This is sort of research that they've put out. But I imagine the code will be available fairly soon. But here's an example of what it can do. Here is this input image for reference of a watercolor of some flowers, and then any prompt you give it will generate the image in that same style. So a baby penguin, it's got that same watercolor style. A banana, a bench, a boat, a cow, a moose, a robot. It all matches this watercolor style here. Here's another watercolor style of a cat with this initial reference image. Now, all of these images are generated in this similar style. Here's a different style. Here's the style of a starry night. Here's another one where it started with this little phone graphic here. Here's another one. Looks like a little clip art kind of image. Here's one that looks like a sticker. And here's. And another example. So many cool examples of how this can be used. Here's another one. Another one, even more. It's only a matter of time before this feature is added to tools like automatic 1111 and Leonardo and Dream Studio and all of the various tools that run stable diffusion under the hood. I can see this being especially useful for things like logo designs and icons on a website. Here's something else I came across this week, which is cool, but also kind of creepy. It's a tool that allows you to replace the face on any video. So here's an example that I came across. You can see here's the initial video and then the face just shifted to somebody else's face here. And here's another one of a stock footage of this guy kind of dancing and then boom, the face shifts to somebody else's face. And here's an example with the intro from the Office where somebody took all of the characters from the Office and then replaced the faces on all those characters with their own face. This is actually pretty hilarious, especially the scene coming up right there. That's scary. Now, this is actually something called Roop R O O P. You can find it on GitHub and it is open source and it's available to download. You can actually install it on your local computer. This is one that I'll likely make a future video tutorial on because it's one that I'm going to play around with, but there are instructions right on the GitHub page. I'll make sure that's linked below in the description. You give it a face that you want to insert into the image, you give it a target video, and then you just let it run and it will swap out the face onto that video. Once again, this stuff is kind of fun to play with, but also getting a little bit scarier. We are getting closer and closer to that point where you can no longer trust video footage. We already can't trust images or audio anymore, and we're about this close to really, really not being able to trust video very much anymore. And I know everybody that watches my videos is probably only going to use it for good, but I'm really worried about all the people that have plans to use this for not so good reasons. Speaking of which, the news came out this week that Google and Facebook are being urged by the European Union to label AI generated content. So essential, if anything was AI generated, they want laws in place where whoever posted that content needs to mark that it was AI generated. So anybody who views it knows it's AI generated. Personally, I don't think that's a bad idea. However, I just don't understand how it's going to be enforceable. I feel like this is going to be a constant cat and mouse game where people get better and better and better at making fake images, fake audio, fake videos. Tools get better and better at spotting them and labeling them as fake, while the people that are creating them just figure out ways to get better and better and better at circumventing these tools that try to figure out if it's fake or not. So while I agree that it's smart to have things in place where we're labeling everything as AI, I just don't know how this can possibly be enforced for the long term. Also in very related news this week, it's already happening. Now that the political season is upon us here in the United States, we're already seeing fake AI generated images circulating to sort of discredit the other campaign. These images have been circulating as part of a campaign for Ron DeSantis against Donald Trump, with images of Trump kissing Fauci and hugging Foushee, and they're being used to try to discredit and disparage Trump. I'm not going to go down the political route. I'm just pointing out that politicians are starting to use fake images to try to discredit other politicians. And this is exactly what Sam Altman was super worried about when he did his Senate hearing last month. I love AI, I love the tech, I love the tools. But this is the world of AI right here that really kind of bums me out sometimes. Now in other news, Hugging Face just released another large language model of their own is a 40 billion parameter open source model that in their testing actually outperformed all of the other open source ish models. I say ish because Llama I don't think is technically open source. You can actually come to Hugging Face's site and test out the model yourself and have chats with it, but it's not quite ready for us to run it on our own computers yet because right now it requires 90 gigabytes of GPU memory. And I don't believe there's really any consumer GPUs right now that are going to give you 90 gigabytes of GPU memory. But if you want to come play around with this model and see what it's capable of again, they do have a demo where you can play with it right here. Explain it like I'm five. It's explaining the Everett interpretation. It's saying that if you have a coin and you flip it and it lands on heads, then there's another universe where the coin lands on tails. These two universes are parallel to each other, meaning that they exist simultaneously but independently. In my very short amount of testing it was actually a pretty impressive model that gave pretty good responses. I'm personally still using ChatGPT and GPT inside of the OpenAI playground for most of my large language model chatting needs. But I am excited to see what the open source community does with some of these models that are being put out and made available for others to iterate off of. Also this week this image was leaked from Instagram. I don't believe really anybody knows the authenticity authenticity of it, so take this with a grain of salt, but there was this image released that says Chat with an AI. Bring AI to your chats for a more fun and engaging experience. Answers questions, gives Advice. Chat with 30 AI personalities and find which one you like best. Inspire creativity, find the best way to express yourself and get help writing messages. This is really all we have to go on right now. I don't totally know if this is a chat bot that's going to be inside of your Instagram DMS or if this is going to be a chat that's going to help you write social media posts better. If it's going to do both. Really all we're going going off of is this one screenshot that got leaked on Twitter. So we'll have to wait and see what Instagram rolls out. Just something to keep an eye on. All right, let's get into something really, really fun. This video dropped the other day from a company called DAZ3D. Now, if you're not familiar with DAZ3D, they're a company that helps you create character models. It's a free tool that you can download and create characters from and then take those models into tools like Blender, Unreal Unity, et cetera. Well, this week that company announced that they're actually adding AI into the mix. And here's the demo video that they released. They typed, make me a middle aged Asian guy. And then it generated that image. Now make him older and shorter. It shortened him, gave him gray hair, made him older. Let's give him a uniform that makes him look like he's from a sci Fi film. They gave him some sci Fi film uniforms. Let him switch him out a little bit. What if we made the character a woman with mid length hair, changed it out to something that looks like a Barbie doll. Take out hair. Let's give her a buzz cut like the original character. All right, now she's got a buzz cut. No, go back to the older man. All right, goes back to that older man. That's awesome. Can you show me four stages of him turning into an alien? Generates four different stages of that guy turning into an alien. Can you put number four on an alien planet? Okay, puts him on an alien planet. Great. Now export it to Unreal Engine. And there you go. The character is in Unreal Engine. And because it's got so much data and so many other models that it's been trained on, you can pretty much type anything you can imagine and have it generate a character that looks like that. Now combine this tool with a tool like Convey and we're getting that closer to anybody being able to create their own video game. Now, you may not be immediately familiar with Convey, but remember this demo video of somebody coming up and having a chat with a character in a ramen shop and the conversation taking place in real time. This was a presentation that was at the Computex event, and Jensen Huang showed off this technology of somebody having a real time conversation. Well, the company that made this conversation happen, that gave this character here a backstory, gave him a personality that created the conversational style for this character. That company was Convey. And this is a tool that anybody can use right now to create characters that will have unique one on one Conversations with you in real time inside of a video game. Now, what's also cool is that I recently sat down with the founder Pernindu of Convay, and we had a conversation. He gave me a private demo, showed me behind the scenes, showed me how all the tool works and how anybody can set it up in their own game environment. And I'm excited to actually release this conversation in this demo in a future video. The point being, imagine taking a tool like the DAZ 3D character creator, where you can prompt any character you can imagine and get the look of the character you want, and then create the personality and storyline and conversational style with a tool like Convey. And you can take those two and pull them into Unreal Engine or Unity. And now you've got a game character with its own personality, with its own way of thinking, with its own conversational style, and you're well on your way to creating very realistic NPCs. I know it's a buzzword, but it's really democratizing things like game development. Also this week, Runway released Gen 2 to the public. In the past it was in a private beta where only a handful of beta testers were able to play around with it inside of Discord. Well, now it's available for everybody to use. Over@runwayml.com you can generate up to 90 seconds of for free before having to sign up for a paid plan. And anybody can demo this now. Just go to runwayml.com, click on try now and click on introducing gen 2 and you can generate any video you can imagine. Now it only generates four second videos right now, and you do have a limited amount of free credits that you can use before you need to upgrade. This is probably the most coherent text to video model that's available right now. Let's do a realistic race car drifting around a track. Go ahead and generate this and see what we get. It's actually pretty quick as well. And it took less than a minute to generate that. And here's what we got out of it. Now the car is kind of going in the wrong direction. It's sort of drifting in reverse, but not bad. I mean, it's pretty coherent. You could tell it's a car. You could tell it's on a racetrack. It's got the smoke coming off the wheels. It's a fairly decent video. It's just kind of going in the wrong direction. Oftentimes with all of this generative AI, it takes prompting it and generating it a few times before you finally get what you're looking for. Exactly. If you follow me over on Twitter over at mreflow, I actually did share some other examples of what I created using Gen 2. Here's one there. If I scroll down a little bit, here's some other ones that I generated. So if you want to see what I made and how I made those, make sure you're also following me over on Twitter at mreflow. Here's something else that I came across on Twitter this week from this company, Flare AI. It's a company that you can take your physical products and sort of blend them with AI to get real cool product shots. And their new tool lets you put it in people's hands. It lets you put them in different environments. It lets you even choose the skin tone and the way the person looks. So check out this super quick demo that they put up.

Generate Human Models on Flare AI to sell your products. Our AI humans are highly diverse and customizable. We provide thousands of faces to choose from in a variety of dynamic poses. Gone are the days of coordinating expensive product photo shoots that can cost thousands of dollars. Simply choose a pose template, position your product, and then hit Generate. The opportunities are endless. You've got you covered. Whether you're targeting different demographics or aiming to highlight specific physical attributes like skin tone, hair color or age, it's about more than just convenience and affordability. Our goal is to use AI to also promote inclusivity at scale that everyone can see themselves represented in your marketing campaigns.

So that looks pretty cool, especially if you have a physical product. Let's say you've got an E commerce store and you need some images for Shopify. You could just take a picture of your product and then put it into all sorts of different poses. Now. Now this demo, it showed it with different people holding it, but Flare also lets you put it with different backgrounds and different environments and all sorts of cool stuff. So you can check that out over at Flare AI. Also this week, the company Baidu out in China released their own AI cloud coding assistant. Watching their demo, it looks like you type in some code and it sort of auto fills the code. It's like you start typing and then it kind of tries to figure out what you're trying to type with the code and then automatically completes the code for you. It looks similar to what tools like Amazon's CodeWhisperer and GitHub's Copilot do. But now Baidu is in the mix as well. There's been a ton of tools out there that will write articles for you and create blog posts and do your copywriting for you and all that sort of stuff. There's so many now that I've actually stopped accepting new tools like this on the Future Tools website because I get probably 10 to 12 of them submitted a day. There's just so many clones of this same concept now. However, now there's this Jetpack AI assistant. And what makes this one different is this one was actually created by Automattic, which is the company that made WordPress. So this is a WordPress add on by the creators of WordPress that help you create blog posts and do the creative article writing that you would do in your WordPress websites. So here's an example of how it works. They press the plus button up here, they click on the AI Assistant button, and then they just ask it write a blog post about Japanese culture and food and then it just goes on and writes the blog post for you. So if you've used tools like Jasper, this is kind of like that, but it's just built straight into WordPress. There is this article over on the Jetpack blog that breaks down all of the features inside of it. It's pretty much what you would expect. You can change the tone of the article. It makes sure to check for spelling and grammar. It's got the conversational UI so you would chat with it like you would a chatbot. And I will make sure this article is linked up below the video so you could read all about it. If you use WordPress and you want to install this one for Your, you get 20 free requests and then after that it's 10 bucks per month to use. And if you are hosting your website on WordPress.com, the AI assistant is available for free of charge on all WordPress.com sites for a limited time. Bard got an update this week and Google announced that it's getting better at logic and reasoning. Not much to cover there. And finally, let's wrap up with this news that Stability AI just launched a new feature called Uncrop. Now, I'm sure you're familiar with Photoshop's generative fill. Well, Uncrop kind of does the same thing. You could take an original image, expand the borders, and it will automatically use AI to generate what is in that outer area of the image. So here's a square image, they make it taller and it adds to the image. And this one right now is actually available to use for free. So if you don't pay for a Photoshop subscription or your free trial ran out, you can Actually use this right now by going over to Clipdrop co Uncrop and you just drag and drop an image in. So go ahead and drop in this image that I generated AI. I'm going to go ahead and expand it on both width and height, see what it generates for me and click next. And here's some example images of how it added to that original image. I don't know what's going on with my fingers in that one. That one doesn't look too bad, actually. And this one, it seems like it mangled my fingers and then added a whole bunch more fingers around me. But that one actually pretty solid image now. Yes, this is an AI generated image. So I know the next question people will ask me is, will it work with real life images as opposed to AI images? Well, let's find out. Here's a random image that I have of Elon Musk on my computer. Let's go ahead and drag this in here. Let's make it a lot taller and wider so it adds to both the width and the height. Let's click next. Sometimes there's a little bit of a queue of people waiting in front of me, but from my experience, it actually moves pretty dang fast. And here's what it generated. Elon holding some sort of book and all of these. It seems to think he's holding something. So he's holding a book. I'm not sure what that is. Maybe also a book. This one looks like he's holding a pillow with a bed in the background. I don't know what he's holding there. Maybe that's a monitor with mountains in the background. But anyway, yes, it works with real images too. It will try to add on to those real images. So despite the big Apple event this week, there was actually a ton of other news. A lot of big stuff happening. It did get overshadowed by Apple, but there's a lot of new tools that you can use out there, a lot of cool research that's happening. And this is just the space that never sleeps. There's always new cool stuff that's happening all of the time. I just love it. Every day I wake up hoping, hey, hopefully there's more AI news that I can go and talk about today, because I just love all this tech. I nerd out about it every day. I'm so excited to be able to share it with you and point you in the direction of stuff that maybe you haven't seen yet. And I'm really appreciative that you check out this YouTube channel. And if you want to even further stay in the loop, head over to FutureTools IE. I do update this news page every single day with even more news than what I share in these videos. And all of the cool tools that I come across I am adding to futuretools IO on a daily basis. And if it's a little bit too much for you, join the free newsletter. And every Friday I'll send you the TLDR for the week of Just a handful of tools, a handful of news articles, a few YouTube videos, and a cool way to make money with AI goes out every Friday. You just got to go to FutureTools IO, click this button and I'll hook you up. Thanks again for tuning in. Really appreciate you. You see you guys in the next video. Bye Bye.