AI NEWS You Simply CANNOT Miss

You're never going to believe this, but this week has been an insane week in AI. We've gotten a ton of announcements from Adobe, from Google, from OpenAI, from 11 Labs, from Character AI, and even from Disney. And I'm going to do my best to break it down and show it all off. So let's get into it. This week was the Adobe Max event in Los Angeles. And if you haven't guessed, the theme of the whole event was pretty much AI. They're putting AI into everything. In fact, there was so much news coming out of Adobe Max that I made a dedicated video all about that event. Otherwise this entire video would be taken up by Adobe News. So if you want a deeper dive breakdown of all of the announcements that came out of Adobe Max, check out this video right here on my channel to get the deep dive. Some of the standout announcements were the new Firefly Image 2, which creates insanely realistic images like this one, this one, and this one. I mean, this may have even passed mid journey in realism. They also showed off some cool features in Adobe Premiere where you can edit videos directly by editing the text, as well as a feature that will find the filler words like us and ums and automatically cut them out of your videos for you. They've also rolled out Adobe's speech enhancement features directly inside of Adobe Premiere, so if you have bad audio quality or a lot of background noise, it will automatically clean that up for you. But by far the coolest thing they showed off during the main keynote was the ability to do text to vectors. So directly inside of Adobe Illustrator, we can now create vector graphics that are infinitely scalable and that are completely editable. And if you have an Illustrator account already, it's super quick and easy to do. I'm gonna create a little square here inside of Illustrator, use my selection tool. You can see I've now got a new Generate beta button. Let's go ahead and click on that. Let's have it create a rocket ship. And we'll click Generate. And there you go. I've got a vector graphic of a rocket ship and I've got a few options here too. There's also this one and this one. They're infinitely scalable, so. So they won't lose quality the bigger they get. And I can edit any individual part of this rocket ship that I want to change. Let's say I want to adjust this flame so it's a little bit further back. I can adjust every little element. Maybe I want to change the shape of this wing. They Also added a feature to copy a style from any other image. So if I pull this colorful image in right here, I could click this little style picker here, select my image, and it will actually model this style and try to match the color scheme from this image. So now let's do our rocket ship one more time. Click Generate. Would you look? Look at that. Some rocket ships that kind of match our colorful image over here. Let me unselect them so you can see what it looks like. And here's a cool example I came across on X from Paul from Adobe. So check this out. He starts by showing off some examples of some flat vector images that you can create. He creates a little space dog here. He then uses the generative AI to create a new scene, gives it the prompt 1950s style, alien planet landscape, retro graphic style, and gets these options here. He pulls a dog into the scene, some people into the scene, add some text. Now he has a postcard available to get people to visit vectora. Now, when I made that Adobe video about Adobe Max, I thought all of the big announcements had come out already because I for the most part covered the first day main keynote. But would you believe it, there's even more that came out of Adobe Max that I missed in that video. So I'm going to cover some additional Adobe stuff that I didn't even talk about in that other video right now. So after I recorded my Adobe video about Illustrator and Firefly and all that stuff, they actually had another presentation that they call Max Sneaks, which is basically their sneak peeks of some of the upcoming features that are in the works. Damn. Some of them are really, really cool. So the first thing that they showed off during this Sneaks event actually had a little bit of a twist to it. So they showed off this image here with some people in the background, masked where the people were. And then you can see over here, they use Generative Fill. Generative Fill then automatically removed the people in the background. After they removed the people in the background, they then revealed that he was editing a video all along and that the people in the background were removed from every single scene of the video. This is what they're calling Project Fast Fill. And here was another demo that he showed off during Project Fast Fill where he showed this man wearing a suit, but he was missing a tie. So he went ahead and traced around where the tie should go, used Generative Fill on the right side, typed in tie, clicked generate, and it added a tie. But again, this was a video. So when he pressed Play the Tie stays with the person as he walks along and looks amazing. Here's another example where they have a coffee with some shapes in it and he used Generative Fill to change the shape from a heart to a leaf here. And you can see as the person walks, the leaf actually moves with the motion of the coffee. Next they showed off Project Draw and Delight, which is part of Firefly, but gives you the ability to sketch out what you're looking for in your image. Click Generate and it will actually generate images based on the combination of your sketch and the prompt that you give it. So here is the cat pushing a ball around and this was the original sketch that it started as. It then gave the option to color it and then automatically filled in the colors on the image as well. Here's another one where they drew this outline of a dog looking thing and then gave the prompt puppy. And here are the images that it generated with that. He then pulled his cat and his puppy in and drew a background scene, gave it the prompt scenery and generated a colorful background in a similar style that matched the original images. They also announced a new upcoming tool called Project Neo, which actually looks similar to like maybe Adobe's version of Blender or something like that, where he's adding a bunch of images and dragging them out and changing the shape to get exactly what he's looking for. And he showed off this complex basketball court that he made and this really cool looking blend feature where you can blend multiple objects together and they sort of mold like putty together. So that was a pretty cool announcement. They also announced something called Project Scene Change where he shot this video of himself walking in his office like this and he wanted to put himself in a scene from Bali. And when he composited the two together like he normally would, it looks very strange and unnatural. But they announced this Project Scene Change which actually takes the point cloud data from the actual scene, composites it with the original video of him walking, and gets a much more natural looking scene of him walking in that area. He also demonstrated this table that has a coffee mug and a handbag on it, used the point cloud cloud data from the scene itself and then he was able to make himself walking around on that table next to the handbag and the coffee mug. They also showed off this tool called Project Primrose. It actually makes the dress that she is wearing change colors and change its entire appearance. You can see it changes to silver here or mirror or whatever you call it disco ball. As she's moving around, you can see the look of her Dress completely changing. This is done with Adobe Primrose. The dress itself can also actually like animate, so you can see the dress is changing in real time as they're talking to her. They also showed off an upcoming feature for Illustrator called Glyph Ease where he created a few letters for a font but didn't have the time to create all the other letters in the font. So he selected the letters he already created. He gave it a little bit of a command to make it look like bread, I guess, and then it processed these letters and then changed the rest of the letters to match the font style of the letters that he already created. He then showed what it would look like if some of these letters had a bite taken out of them, reprocessed it, and all of the letters now look like they had a bite taken out of them. Check this out. He even took a picture of handwritten letters that say Rain used Illustrator to select those letters and then created an image that said Organic Honey with that same sort of drip look from the handwritten image that he created before. They showed off Project Posable, which makes it a lot easier to rig up and pose characters that you create. And then once they posed the character, they entered a prompt. The Princess Knight sits on the throne in a mid century castle. She looks confident and I think there's more to the prompt, but it's a little cut off here and you can see it generated an image following the pose that was just created. It's like a much, much more advanced control Net pose sort of option here. You can also take existing images with a pose in them, load them into this tool and automatically change the character's pose to the pose from the image. They then showed off Project Resupply. They took this cropped in image of an elephant here, submitted it through this resup tool and it boosted the resolution of that cropped in elephant. And my guess is that this is going to be added to premiere at some point in the future. That's my hope at least. They also showed off Project Dub Dub Dub, which is an audio translator so you can record something in English, pick a language you want it to dub in and it will dub it out in that other language. They also showed off their project See through, where he showed an image that was taken out of a car window and it had reflections of the inside of the car in it. He selected an area, clicked Remove Reflections. It found all the reflections from the car window and removed them automatically. Here's another side by side image of a dog. You can See the one on the left? It was taken through a window. You can see all the reflections of things going on outside the window. And the image on the right is after Project See through remove the Reflections. Here's an image of some turntables looking through a window. And here's after Project See through removed all the reflections. And then to wrap it up, they talked about Project Stardust, which I did tease a little bit in a previous news video. But this is the one where you can take any image, click the little button down on the bottom to remove distractions, and it will actually take all of the rest of the people that are behind you in this scene out of the image. She then added a dog into the image using, I believe Firefly, but then decided to actually change the clothes on the people in the image. So she clicked a replace button, entered a prompt of or orange sweater. And just like that, the woman in the photo is now wearing an orange sweater. Here's another example that looks like a stock image. She's able to select any individual character in this image and move that character around because of course it magically knows what he looks like behind her. Of course we can do that. And it automatically filled in the area behind her. So yeah, as it turns out, my video from Adobe Max missed a few things. In other Adobe news this week, Adobe also created a new symbol and they're encouraging people who create art with AI to use this symbol on their AI generated content to let people know that it's AI generated. If you're wondering what CR means, it stands for Content Credentials. So far I've only covered just the news that's come out of Adobe this week, and as you can see from my tabs at the top of my screen here, I still have a lot to cover. But before I do, let me quickly tell you about the sponsor of today's video GPT Trainer. GPT Trainer allows you to create your own chatbot similar to Chat GPT, train it on your own data and information, and then embed it directly on your website, all without needing to know how to do any code. You can upload files directly for the context of your chatbot. You can have it use Google Drive, you can use website URLs, website sitemaps, and even add your own pre canned questions and answers. Once inside, you can customize the appearance, create an initial message, suggest some questions for them to ask when the chatbot opens. You can even go back and review all of the chat that your chatbot has had with various people on your website to analyze how it's doing and consistently improve the quality of the responses. The user interface is simple and intuitive and this is absolutely amazing to help with customer support and getting anybody visiting your site the answers that they need without having to wait on an actual human being to reply to them. Not to mention that it easily integrates with probably any of the website builders that you're actually using right now. They also have API access on their starter plan, and they pride them themselves on their own customer support. So you'll be able to get all of your questions answered as you're trying to figure out how to build your own bot. You can learn more over at GPT-trainer.com and use the coupon code matt wolf 25 for 25 off that's m a t t w o l f e 25 for 25 off over at GPT-trainer dot com thank you so much again to GPT trainer for sponsoring this video. I really appreciate you guys. Now let's get back to the news. Google had a couple announcements this week as well, including the fact that if you use their new Generative Search experience, you can actually generate AI images directly inside of their search. If I come to Google's new Generative search and I say draw an image of a wolf in the snow, you can see it says generating. And here you can see it generated a handful of images of a wolf in the snow, including a moose chasing a wolf and a green wolf. I'm assuming that's the glow of the Aurora Borealis or something. Google also announced an update that's coming to Chrome, where it will use AI to automatically organize and reorder your tabs for you. Here's what it's expected to look like with a little organized tab button, and you can see that it will organize the tabs for you and then let you sort them into groups. Google also made headlines this week as being yet another company saying that if you use their generative AI, they've got your back if you ever go to court over it. Google said on Thursday that it will defend users of generative artificial intelligence systems in its Google Cloud and Workspace platforms if they are accused of intellectual property violations, joining Microsoft, Adobe and other companies that have made similar pledges. And in OpenAI news, most of us, if you have a chat GPT plus account, appear to have access to both GPT4 Vision, where you can attach images to your chat, as well as Dall E3 directly inside of Chat GPT. Now, this single feature deserves its own video breakdown and you can expect that next week. But as a real quick demo, let me upload this image of my head shot here and tell it to make a joke about this image. I find that I have to give it this extra disclaimer or it will refuse to do it. This is an image of me so no one will be offended. Joke. Hey Matt, you look like you just heard someone say that they still use a flip phone in 2023. Okay. And unfortunately we don't have the option to use images when we use the Dall E3 feature, but we can now generate images directly with Dall E3 right inside of ChatGPT. Draw an image of a cat eating a taco and we got some nice taco cat images here. Make them aspect ratio. Now we get the same images, but in aspect ratio:16.9. It filled in the sides. Here's the originals, here's the new versions. If you don't have chat GPT plus or maybe GPT4 Vision hasn't rolled out into your account yet, no worries. There's actually a sort of free version of what GPT4 Vision does, called lava or Large Language and Vision Assistant. You can find it over at, well, this URL right here. Let's pull in my headshot, give it a prompt, make a joke about this image. Why did the man with the beard and the blue plaid shirt smile so brightly? Because he was having a great hair day. In other OpenAI news, OpenAI is planning some major updates to their API to lure in more developers, including allowing the vision capabilities for developers through the API, as well as giving additional memory storage, which could theoretically slash costs for application makers by as much as 20 times. This is addressing a major concern for partners whose costs of using OpenAI's models pile up quickly. Now, whether or not all of these tools that use OpenAI's API are going to pass those cost savings on to us or not? Well, that's yet to be seen. This week, eleven Labs launched a new translation feature where it will take audio that was created in English or whatever source language and then translate it to whatever output language you want and it will actually use the same voice speaking in this new language. So if I log into 11 labs here, click on Create new dub. So I found a little two minute video clip on my hard drive from one of my previous videos where I'm showing off some cool AI animation stuff and I want to dub this video. So here's what it sort of initially sounds like in this video. We're going to have some fun. I want to Experiment with a handful of different AI generation tools. So let's go ahead and plop this in right here for the project name. I'm just going to call IT Test Project 1. This is literally the first time I'm trying this target language. I'm going to go with Hindi for this one. There's some advanced settings here. There's only one speaker, so I don't need to change anything there. And I can pick a specific range for the dubbing. Let's just go ahead and dub the first 30 seconds of the video. We'll go ahead and click Create says there's a little over 30 seconds remaining. So seems to be a pretty quick process and here's what we got. I have no idea how that sounded. I don't speak Hindi, but it was a pretty quick process. So if that was accurate, it's in my voice and was super easy to do. And there's still a ton more news I want to share. So from this point on, I'm going to kind of rapid fire all of the other quick news that I think you need to know about. For example, character AI just released a new group chat feature. This group chat feature allows you to chat with an AI bot and have other friends come in and chat with that AI bot with you. But it also allows you to create groups with multiple AI bots. So if you want to put Mark Zuckerberg, Steve Jobs and Elon Musk all in a bot together and have a conversation with all three of those AIs at once, you can do things like that. This week, Disney got into a little bit of trouble because people are claiming that this poster for the new season of Loki has some AI generated assets in it. The claim being that these clocks in the background look very AI generated with some sort of gibberish text. A note about why this this is controversial. Licensing photos and illustrations on stock sites has been a way many hard working artists have been earning a living. I don't think replacing them with generated imagery via tech built on mass exploitation and wage theft is any more ethical than replacing Disney's own employees. Now, from what I understand, nobody was replaced to create this. An artist was still involved to composite and pull this all together. And nobody actually confirmed that this was generated with AI. People are just just sort of speculating that. In fun Disney news, Disney's been showing off their cool new little robot character that was actually trained using reinforcement learning. This little guy has total Wall E vibes. Today my wife sent me this video on Instagram of these droids actually being deployed at Disneyland's Galaxy's Edge to walk around and interact with people at Disneyland. So that is pretty awesome. I'm going to Disneyland next month. Can't wait to see these little guys. Although the writers strike is over, actors themselves are still on strike. They're pushing for this bill called the no Fakes act, which actually is an acronym for Nurture Originals, Foster Art and Keep Entertainment Safe Act. The bill aims to bring in federal protection against the misappropriation of voice and likeness performance in sound recordings and audiovisual works. It also prohibits the unauthorized use of digital replicas without the informed consent of the individuals being replicated, which in my opinion, I think is ultra fair. They're not fighting to never have their voices replicated or likenesses replicated for movies. They just want to be consented before it happens. Which seems like a totally logical, reasonable request to me. There's been some other shade being thrown around at AI this week as well as Charlie Munger, Warren Buffett's right hand man, is basically saying that AI is overhyped, saying artificial intelligence is very important. But there's a lot of crazy hype on the subject. Artificial intelligence is not going to cure cancer, is not going to do everything that we want done. Now, Charlie Munger is 99 years old, so he's right in the fact that it's probably not going to cure cancer in his lifetime. But I totally disagree with the fact that it is definitely not going to cure cancer. I mean, I say at the moment, a combination of AI with doctors and researchers working together is probably how we're going to come to some of these cures. In other AI shade, there's an article on CNBC that says generative AI will get a cold shower in 2024. Analysts predict the article is basically claiming that using AI is so expensive right now and that businesses are having such a hard time being profitable because of how costly AI is, that for these reasons in 2024 we're going to see it really cool off. Now, I don't doubt that some of the hype around AI is going to cool off. I think it's going to come in waves. It's been sort of coming in waves all year long in 20 expect something very similar in 2024. But I also expect some of these big tech companies to figure out ways to bring the costs down. Right now they're figuring out the best use cases for a lot of these tools. And now that we're sort of solidifying the actual real world uses of these tools the sort of next order of businesses for these AI companies if they want to stay in business, is to figure out how to reduce the costs of doing this AI stuff. They're going to need to figure out how to make it more energy efficient. They're going to need to figure out how to run smaller models that need smaller GPUs. I just believe that it's inevitable that when these companies have real clear use cases that the world can really get behind, they will figure out how to make them cheaper and cheaper to use. That's just kind of how technology has worked forever. This week there was also a really good short 13 minute interview with Geoffrey Hinton, aka the Godfather of AI. He was on 60 Minutes. If you're not familiar with him, he was one of the earliest pioneers of this technology. The company he created was acquired by Google. He worked at Google for a long time and then eventually stepped away from Google to sort of spread the word on the dangers of AI. And if you watch this 60 minute interview, he talks about both the potential benefits and the dangers. It's not a total doom and gloom interview.

I think we're moving into a period when for the first time ever we may have things more intelligent than us. You believe they can understand?

Yes.

You believe they are intelligent?

Yes.

You believe these systems have experiences of their own and can make decisions based on those experiences in the same sense as people do? Yes. Will they have self awareness consciousness? Yes. Oh yes, I think they will in time.

Definitely a highly recommended watch if you're really into AI. I have a feeling this interview was also highly edited and a lot of what Geoffrey Hinton said during the interview was likely edited out. But nonetheless I found it a really valuable watch and making sure that I end this video on a positive note. Google Release released an article this week about how they're using AI to combat floods, wildfires and extreme heat. Using AI, they're able to better forecast floods in more places. They're able to better track wildfire boundaries and predict where the fires are going to spread. And they're using AI to prevent heat islands, which are areas that due to the infrastructure of the area just tend to get hotter than other areas. And last but not least, Harvard released an article called Staying Ahead of Virus Mutations, which is a new AI tool that can predict viral ovarian, figure out the variants that are the most dangerous and future proof vaccines. The tool has two elements, a model of evolutionary sequences that predicts changes that it can occur to a virus, and a detailed biological structural information about the virus. Together, they allow Eavescape to make predictions about the variants most likely to occur as the virus evolves. They're currently using it to look ahead at Covid and predict future variants of concern so as I've mentioned, in the past, AI and healthcare care are like the perfect combination. And now we're seeing AI used to actually help us determine the future of viruses and come up with vaccines before they're even necessary. So like I said, a lot happening in the AI world this week. A ton from Adobe, from Google, from OpenAI. It's just been a crazy week. And if you want to make sure that you're always staying in the loop with the latest AI news, check out FutureTools IO. This is where I keep the news up to date on a daily basis. I also curate all the coolest AI tools that I come across and I have a weekly newsletter where I share all of the coolest AI tools and AI news directly to your inbox. You can find it all over at FutureTools IO. So thank you once again for tuning into this video. I had a blast nerding out in this one. So much cool stuff came out this week that was so much fun to play with. Hopefully you learned about something new in this video. If you did, give this video a thumbs up. If you haven't already, maybe consider subscribing to this channel. Subscribe. That'll make sure you see future videos like this one in your YouTube feed. And thank you once again to GPT Trainer for sponsoring this video. I really appreciate you guys. Thank you so much. I'll see you in the next video. Bye bye.