Transcription
Jet GPT's newest feature lets it read your emails and calendar. Deep Mind's latest research can read our DNA. And we have even more fresh drama coming out of OpenAI. All that and more in today's AI news breakdown.
But to kick things off, let's talk about what Anthropic just released. They rolled out a feature that lets you actually build, share, and bind other people's apps directly inside of the Claude platform. So, not only will Claude code things up for you directly inside of its platform, but it lets you actually use the app in the platform and then share the app with others so that others could come and use that same app. So, no need to worry about where the app is going to be hosted and what sort of software stack you're going to build it on because Anthropic now sort of takes care of all of that.
If I head over to cloud.ai over in the left sidebar, you can see there's a menu item for artifacts. If I click into artifacts, you can actually see other apps that other people have built and mess with those apps. For example, Sakura Serenity. Let's just click into this. And you can see that I can use this little game that somebody created here. All it lets me do is move around and jump. But after this was built, it was able to be shared and now I can use it directly within Claude's platform.
So, let's build our own bot real quick. Let's say I want a bot all about Gary Buucy. Yeah, let's go with that. I want to know the latest news about Gary Buucy. I want to see images of Gary Buucy in a photo gallery. I want to have an Ask Buucy bot where I can chat with an AI Gary Buucy. I'm going to type up a little prompt real quick and ideally we'll get exactly what we're looking for.
So, I've defined what I'm looking for. Basically, a web app all about Gary Buucy. This is like the fan app for anybody who just wants to know everything they can about Gary. I'm going to go ahead and submit this and we'll see what it comes up with for us. It starts with giving me a game plan like main features, latest news section, random quote generator, Ask Buucy AI chatbot, image gallery, and a Cameo booking link where you can book Gary directly. It's got all of the steps and technical implementation. So, we're going to say, "Yep, let's go. Here we go." It opened up a new coding window. We can see it write the code. And about a minute or two later, we've got the Gary Buucy fan hub. We've got our latest news: Gary Buucy makes surprise appearance at a film festival; Latest motivational speaking tour announced; Celebrating his iconic career moments. And a link down here to book Gary on Cameo. Did it actually use the right link? Let's see. Open link. Yep, it used the right link to book him directly. Let's see. Let's click on Buucy quotes. Have you ever noticed how stained glass windows vibrate in different colors? They're singing. Gary Buucy. Let's get a new quote. Coffee doesn't ask silly questions. Coffee understands. Be like coffee. Timeless wisdom. Let's Let's do one more. Imagination is the hood ornament on your car of creativity. All right, let's check out our photo gallery. Ah, it didn't pull in any images, unfortunately. No big deal. Let's try our Ask Buucy bot. Hey there, amigo. It's Gary Buucy here. Ready to share some of life's magnificent truths with you. What's on your beautiful mind today? What is your best piece of advice? Look at this. Is it even integrated at LLM? I believe. I'm assuming it's Claude. Oh man, let me tell you something that'll blow your mind like a cosmic tornado of truth. My best advice, fear. False evidence appearing real. You see, amigo, fear is just your ego trying to keep you small, but you were born to be magnificent. I mean, come on. This is this is some wisdom from Gary Buucy himself.
But what makes this new feature at Anthropic so powerful is I can click publish, publish and copy link, and now anybody can go and use my Gary Buucy fan hub. In fact, we'll put the link in the description if you want to go check out Gary Buucy Fanhub. What's even cooler is they rolled out this new feature in both the free and the paid plan. So even free Claude users can do what I just did.
11 Labs rolled out a new voice assistant this week called 11 AI and it connects via MCP servers to all sorts of different services so that you can have a conversation and get it to give you information based on the various sources it's connected to. So let's go check it out. You can find it over at 11.ai and as of right now they have easy integrations with ASA, Google Calendar, Hacker News, Linear, Notion, Perplexity, and Slack. And you can add even more integrations if you know what you're doing and know how to connect MCP servers. I've gone ahead and connected Google Calendar, Perplexity, and Slack. And when it comes to voices, we've got a nice variety of voices that you can choose from. And I even trained my own voice into 11 Labs, so I could have myself talking to me if I wanted to. But you know what? I'm going to go ahead and use Burt Reynolds. A little disappointed that it's not Gary Buucy, but Burt will do. Hi there. It's 11. Tell me what's next on your list. How busy is my schedule next week? Here's a summary of your schedule for the next week. On Saturday, June 28th, you have the Kora Swim Meet from 7:30 a.m. to 12:30 p.m. Uh, are there any messages awaiting for me to reply instead of Slack right now? Okay, I found some messages from Dave. The most recent messages are mind if I call you two sex can even be a phone call. Would you like me to get more messages from Dave? No, no, that's good. Thank you, Burt.
And since we're talking about 11 Labs, I should mention real quick that 11 Labs just rolled out a new mobile app as well that lets you do text to speech directly from your phone. Now, at the moment, the app seems pretty bare bones. I can add some text. I can change the voices here. Let's go ahead and use the Matt Wolf voice that I trained. A wise man will make more opportunities than he finds for the good old American life. And I'll use Burt Reynolds for the secondary voice. So, let's go ahead and click generate and see how it sounds. Subscribe to Matt Wolfan. Check out future tools. I don't know why it did in only one voice. But at the moment, you can do stuff like that. And you can see the history of past audio chats you've done both from the web app and the mobile app.
We also got an update to Chat GPT this week. If you're one of the paying users of Chat GPT, there's now these Chat GPT connectors for Google Drive, Dropbox, SharePoint, and Box. And let me tell you, using these connectors is actually pretty dang handy. So, you can access these new connectors by logging into your Chat GPT account, going up to your profile, clicking on settings, and then you can see all of the connectors here with new access to things like Google Drive, Box, and Dropbox. So now files that you put inside of those places, Chat GPT can have access to. Pretty cool new feature. If you want to pull information in from one of those places inside of Chat GPT, click on tools, select search connectors, and then select your source. Now, whatever chat you do here will search that source when chatting with it. Pretty cool new feature from Chat GPT.
Another feature out of OpenAI this week is the new record mode. Basically, what this is is, let's imagine you're on a Zoom call or a Google Meet or a Microsoft Teams meeting, something like that. You can actually record directly with Chat GPT now and it will help summarize the calls and give you the to-dos and things like that after the meeting. Now, you've already got a feature like that inside of tools like Zoom and Google Meet, but now I guess you can do it a little bit more discreetly, like people won't see that you're actually recording the conversation and getting summaries. I'm not sure like the main benefit of using Chat GPT over using it directly inside of whatever the meeting app is that you're using, but you could do that now if you want. So, if you have the Chat GPT desktop app, you'll have a new record button down in the bottom right. If I click on this record button, we can see that it's starting to record everything that I'm saying. It will also listen to anything that's coming back at us. So, if I'm watching a YouTube video or I'm on a Zoom call or I'm in a Google Meet, it will actually record both sides of the conversation so that at the end of it, I can actually get details in a summary about what I talked about. When I'm done, I click stop, click send, and we can see that it pulled in details about what I just said. Overview of recording software features. Matt Wolf explains recording software that captures both sides. So, it's not spitting back the complete transcription. It's just summarizing what it heard and giving me back details and notes about it. So, another pretty handy new feature from OpenAI for Chat GPT. Both these features that I just talked about are on the paid plans, but most of the features that come to the paid plans eventually migrate to the free plan at some point. So, I imagine it's only a matter of time.
For those of you that love to play around with new image models, we got a new model out of Higsfield AI called Soul. And it's a super realistic model, but it also seems to be kind of designed with a bunch of presets in mind. So, check this out. If I head on over to Higsfield AI, we can see this new Higsfield Soul here. And there's all sorts of sort of pre-designed looks here. Things like hair clips where it took an image and put hair clips on it or elevator mirror selfies where it took images and made them into elevator mirror selfies. But as you can see, the image is ultra realistic. Let's go ahead and generate. And it appears to be only text to image at the moment. So, let's do a tall white man with a short brown beard wearing a Padres hat and let's generate that. I'm also still on this elevator mirror version here. So, it should be a white dude with a beard wearing a Padre hat but in an elevator mirror and it really screwed it up. It put basically a Dodgers hat which is just unacceptable. Unacceptable. But, I mean, the realism's kind of there.
The company Udio this week that helps you generate music with AI rolled out some new features in a new tool called Sessions. Sessions automatically identifies musical elements like choruses and bridges from audio waveforms, allowing for the editing of lyrics and sound through a visual workstation. Now, I will say that this Sessions feature isn't super intuitive. Like, it doesn't immediately pop out at me on how to actually use it. But I do have some songs I generated previously. And if I click on these dots, there is an edit option where I can like replace a section or edit lyrics. We've got this little slider here where I can select different areas of the song. And so I'm assuming this is the Sessions section here. All right. So, let's give this a quick try here. Here's what the existing song sounds like. He breaks it down, makes it clear as day. Now, I want to try and change some of the lyrics here and see how good it sounds. See if it still sounds like the original song that it made, but with the new lyrics. I will say this UI is kind of clunky. Like, I can't seem to unselect this section for some reason. Is it because this is what I've got selected here? Yeah, that's it. Okay. So, I want to edit the lyrics. Let's change AI demystified in the funnest way. Does it immediately change it? So, I clicked the blue edit button after making my changes, but it seemingly made like a different song. It makes it clear as day in the funnest way. So, it changed the words. It also added the word undefined at the beginning before it started singing. So, I don't know. I guess I just don't know how to use this tool yet, but something fun to play with if you're into making AI generated music.
We also got a new app out of Google this week called Doppel, which I believe is like short for doppelganger, but it's an app where you can upload a photo of yourself and then virtually try on different clothes. So, I went ahead and downloaded the app. I can upload an image of myself here. I'll go ahead and use this image. It's making my doppel. And now it's giving me various looks that I can try like this or this. I mean, I'm not quite sure this was designed for men, but let's see what this would look like on me. Let's try this look. Perfect. Amazing. So, that's what it generated. Uh, and then I think, yep, I can actually animate this. So, let's go ahead and uh animate this. And this is what it looks like. What the heck? They turned into pants when I turned around. Anyway, let's see. Let's try a look from a screenshot. I'm going to try to use this image and see if it'll give me this look. Let's see if we can break it. So, yeah, just as a reminder, this was the image I gave it and I, you know, I guess it found the clothes the guy was wearing, but it omitted the dinosaur. Kind of a bummer. Not bad, though. And here's what it looks like when it's animated. I can uh, you know, sort of turn around and look at the camera. Pretty fun little app, actually. Uh, but it won't put me on a dinosaur for whatever reason. AI is not advanced enough for that yet.
We got some fresh AI drama this week as well. Starting with some controversy over the acquisition that OpenAI did with IO, Johnny Ive's company. So apparently Johnny Ive and Sam Altman met with another company called IYO, and got demos of this hardware product that they were making. The product that they were making was like this earbud that I think had like cameras in it and used AI and could understand what was going on around it. Kind of like what is in the movie Her. Sam and Johnny met with this company acting like they might potentially want to acquire it, but they turned it down, decided not to acquire it. And then a few weeks later, OpenAI buys Johnny Ive's company, also called IO, but this time spelled just IO, and announced that they're going to be making hardware together. Well, the company they met with, IYO, did not like that. And now they're suing OpenAI, saying that they stole the concepts after meeting with them. However, OpenAI and Johnny Ive are disputing this, saying that we're not even building a wearable device. So, the claim that they're trying to make is BS because the thing that we're going to put out is not going to be an earbud like what they put out. And well, this is a story that's still unfolding. And we probably won't know exactly how it unfolds until OpenAI actually reveals whatever it is these two guys have been working on together.
And in more Sam Altman related drama, Sam Altman joined the Hard Fork podcast, which if you're not familiar with that podcast, it's a podcast owned by the New York Times. New York Times is a company that is currently suing OpenAI and trying to get OpenAI to share all of their data and save all of the customer user data for potential future discovery in court. And there's a lot of drama going on between OpenAI and New York Times. So, it was really interesting to see Sam Altman come out and be interviewed on a New York Times podcast. And let me tell you, it was awkward. We think that privacy and AI is just this like extremely important concept to get right for the future and we care a lot about the president. Well, thank you for your views and I'll just say it must be really hard when someone does something with your data that you don't really want them to.
All right, there was a handful of other things that happened that I want to talk about real quick. So, let's jump into the rapid fire section. There were a few key court cases this week that actually went in the favor of AI that could set the precedent for how AI lawsuits are handled in the future. So this week, Anthropic won a key ruling basically saying that the fact that they trained on a whole bunch of books was fair use, but there was some nuance to it. Basically, Anthropic claims that they bought a whole bunch of books, ripped them out of their bindings, scanned them all in, and then trained on the data that was in those books that they legally purchased. And the judge says, "Yes, you can do that. That falls under fair use. You purchase them. You use them to train your models. Cool. We'll allow that." That part really, really sets the precedent around fair use in the future for training AI models. However, the judge didn't completely let Anthropic off the hook because it also turns out that there was probably a lot of pirated books that were also included in the training material. And Anthropic is sort of still on trial for that portion of the training data. So, it kind of seems like the judge said, "It's cool if you train on this stuff because you paid for it legally. You owned it. You scanned it in and used the training data. But this stuff over here, which looks like you probably just scraped from the internet from maybe illegal torrent sites or whatever to get this book data, that's probably not cool. We need to further flesh that out in court."
And that wasn't the only judgment that came down this week that was sort of in favor of AI. Meta also had copyright claims. And well, the judge wasn't quite as friendly to Meta as it was to Anthropic, but Meta did still win the case. They only won the case because the plaintiff that was suing Meta apparently just sort of worded the case wrong. On Wednesday, Judge Vince Chabria ruled in Meta's favor, saying it is entitled to summary judgment on its fair use defense to claim that copying these plaintiff's books for use as LLM training data was infringement. However, it goes on to say, "This ruling does not stand for the proposition that Meta's use of copyrighted materials to train its language models is lawful. It stands only for the proposition that these plaintiffs made the wrong arguments and failed to develop a record in support of the right one." So basically the argument was that the people who wrote these books claimed that Meta's Llama can regurgitate what was in these books, making it so that people wouldn't need to buy the books because they could just ask Llama to regurgitate it for them. However, they couldn't prove that in court. It was only able to give like little snippets and sound bites and it wasn't able to regurgitate the whole thing. And that was the claim of the plaintiff. Well, the judge basically said, "Well, it can't recreate these books word for word in a way that makes it so people wouldn't want to buy these books. So, your argument is not valid." But again, he also went on to say, "I'm not saying that scraping all this data to use in these books is legal. I'm just saying that you weren't able to prove that these AI models are able to regurgitate these books well enough that it would prevent people from actually buying the books." But all of this is sort of heading in the direction of the courts basically believing that sort of using information from books that were purchased legally or acquired legally are likely to be able to be scraped and used on account of fair use. And it worked so well that Getty actually dropped a copyright claim against Stability AI. So Getty was suing Stability for sucking in its images and being able to regurgitate similarish images. And well, Getty saw the ruling that Anthropic got and the ruling that Meta got and went, "We're not going to win this case" and dropped it. Now, they didn't technically drop the entire case, but they dropped the portion that was related to the fair use of the training data.
Google launched a new open-source Gemini CLI that stands for command line interface. And basically, it's an AI agent that can write code for you directly from within your terminal on your computer. and it leverages the various Gemini models and a lot of developers have been saying it works really, really well. I haven't tested it yet. It's a little too in the weeds for me, but apparently people are loving it and those who actually use it, they can actually use the new Gemini 2.5 Pro model for free and get 60 model requests per minute and 1,000 requests per day at no charge. So, just like your regular vibe coder who's using this tool to code up projects is probably never going to
Run out of requests and be able to use Gemini 2.5 Pro for free to build their apps. And the fact that it's open-source means other people can sort of build off of this CLI and improve it and make better versions of it. So, a pretty cool move by Google in my opinion.
Google DeepMind this week also dropped a game-changing AI model called Alpha Genome, and it could totally reshape how we understand human disease. This new tool can read a million DNA letters at once and predict in fine detail what happens when a mutation occurs. So, from cancer research to synthetic biology, this Alpha Genome could really accelerate breakthroughs across genetics and medicine. Google already gave access to scientists through its API if they're using it for non-commercial research.
Now, I'm not going to claim that I completely understand how this model works or what it's going to be used for, but it sounds pretty impressive for scientists and researchers and doctors and things like that to help make new discoveries to help with the human body be better at its job.
Google DeepMind also just released Gemini Robotics on device. It's like this AI brain for robots that runs entirely on the robot itself. It lets the robots follow natural language commands, learn new skills from just a handful of examples, and perform delicate tasks like folding clothes or assembling machinery, all without actually needing to be connected to the internet. And I mean, that all sounds awesome and like a big leap for making robots smarter. But I also can't help but think this is how the robots take over. Like, if their brain was in the cloud and we needed internet to run the large language models, well, we could just disconnect the Wi-Fi. I've seen iRoot. I know how this plays out.
If you have a Ring camera, you're about to get a nice little AI upgrade with your Ring camera. They have a new feature called video descriptions, which generates detailed descriptions of what your Ring camera sees and delivers it in your notifications. So this means you'll get an app notification instead of saying there's motion at your front; it'll say something like, "Two people walking on the driveway with a white dog," instead of a simple "Person detected." So that could come in handy.
And back to the topic of legal, there's something really interesting happening in the United States right now where the US government is trying to make it so that states can't regulate AI for the next 10 years. The rule would withhold federal broadband funding from states if they try to enforce AI regulations in the next 10 years. Now, it's not clear if this is actually going to pass or not, cuz it seems like there's both Democrats and Republicans who oppose this. And if you want my opinion, I kind of understand where they're coming from. We're kind of in this like cold war with the US and China where we're both racing to have the most powerful AI and hit AGI and then ASI first, and all of that's happening. And the federal government's worried that if we try to regulate too much too quickly, then it's going to hinder our progress and give other nations the ability to catch up and surpass us. That's obviously why they're doing it. But 10 years is a long time in my opinion. Like the pace of AI and how it's advanced just in the last like 3 years is mind-boggling. We can't even imagine what AI is going to look like 10 years from now. So to put this like blanket restriction for 10 years feels a little bit like an overcorrection in my opinion. I mean, maybe set it for like a year and then when a year comes around, re-evaluate and see if you want to extend it. But setting it for 10 years, I don't know.
Now, I don't want to end on that note. I like to end on positive news. So, let's talk about this. In this major breakthrough, scientists in Berlin have developed an AI model that can diagnose brain tumors with over 99% accuracy without needing surgery. So, it reads subtle DNA signals in fluid samples like spinal taps. And the system is called CrossN. And it can identify over 170 types of cancer. It's faster, it's safer, and it's more precise than traditional biopsies. And they're already preparing to use it in clinical trials in Germany right now. I love seeing these news stories where AI is making this real positive impact on the world, especially in areas like health. And this one, well, it could truly change the way we detect and treat cancer forever. So, love seeing stuff like this.
But that's all I got for you today. I hope you enjoyed this. I hope you feel more looped in with all of the AI world and drama and news and things you can do. I'm having a blast paying attention, keeping up with it all for you and then turning around and breaking it all down for you once or twice a week. And if you like stuff like that, make sure you like and subscribe, and I will make sure more stuff like this shows up in your YouTube feed. Thank you so much for tuning in. Really, really appreciate you. Hopefully I'll see you in the next one. Bye-bye. Thank you so much for nerding out with me today. If you like videos like this, make sure to give it a thumbs up and subscribe to this channel. I'll make sure more videos like this show up in your YouTube feed. And if you haven't already, check out futuretools.io where I share all the coolest AI tools and all the latest AI news. And there's an awesome free newsletter. Thanks again. Really appreciate you. See you in the next one.