AI News: The AI World is REALLY Scared Right Now transcript

We got a bunch of new AI releases this week, as well as some AI researchers telling us that we need to be way more scared of AI than we actually are right now. There's some important stuff going on right now, and I don't want to waste your time. So, let's break it all down. Let's start this week on a lighter topic, and that's the new image model out of OpenAI. This week, they introduced ChatGpt images 2.5. So, this is the image model that you use right inside of Chat GPT, and they've improved it quite a bit.

This new model is apparently a lot better at working from reference photos to transform familiar subjects, visual styles, and compositions. So, when you ask it to put somebody in a suit, it looks a lot more like the original person than, you know, the old model might have generated. They're also showing off here that there's much more consistency in the edit in this new version. The pose and the face and everything stays way more consistent than the one on the left.

In this example here, they're showing somebody holding a ticket and the fingers and the ticket shape and everything stays exactly the same, but the images and information inside the ticket are changing. Here's more examples of consistency. And it really seems like consistency is kind of like the biggest update with this. However, they did introduce a new feature called sketch, which let's just go and actually play with it.

When I'm inside of chat GPT here, I can now create an image by first sketching it. And this actually works on mobile as well. But if I click on this plus, you can see there's a new sketch option here. If I select this, it opens up a little canvas where I can draw. So I can draw a picture. I am very bad at drawing, but let's just go ahead and make a little like person here, you know, just draw it as you'd normally draw people.

And now if I click check, it'll use this sketch as like a reference image. And then I could give it a prompt and say generate this image, but make it look super realistic. And there you go. There's the super realistic version of my sketch. You can see it sort of followed all the randomness that I gave and tried to make it as realistic as possible. I also did a fun little test where I had Astra sketch an image of me. So, I actually told ChatGpt Work to sketch an image of me and then Astra actually drew this sketch of what it thought I looked like.

And then I had it prompted to make this image more realistic basically. And this is what it generated. So, it's almost like a game of telephone, right? It looked at a picture of me, drew this as the picture of me, and then prompted it to make it more realistic, and this is what it came up with. So, this new model is available as of right now inside of ChatGpt, ChatgPT Work, and Codeex across all tiers on desktop, mobile, and web.

There's also two variations of the model available inside of the API. So, if you're trying to build with it and work it into an app, you can do that. One model is slightly more detailed and slightly more expensive. The other model is a little less detailed, a little faster, and a little bit cheaper. So, if you've used the chat GPT images feature in the past and you struggled with consistency from it, well, maybe now's the time to give it a second look because that's the thing that they most improved with this new model along with that kind of fun but sort of silly sketch feature.

We got some big news out of Meta this week as well. Meta launched their brand new agent called Muse, which apparently is the world's first personal agent built for everyone. So, Muse will actually take actions on your behalf. You can connect it to things like your email, your calendar, various apps, and it actually runs on a secure virtual machine. So, it'll do things on your behalf inside of this virtual environment instead of doing them on your own computer.

So, it can handle tasks like sending an email or booking travel, and it can take on big audacious goals. Once a person shares a goal with Muse, it helps them develop a personalized plan and coordinate their time and resources, then advances the work on its own. It can open a browser, fill out forms, and negotiate on your behalf. It'll even keep working after people close the app, but comes back to you when something needs changes or when it needs approval, like before sending an email or making a purchase.

It also apparently remembers what matters to a person. So, it'll make suggestions unprompted and act on details that maybe you only mentioned once. And from what I understand, it also has all the context of your Facebook and your Instagram and I think WhatsApp as well. Now, when it comes to meta, I know some people might have some privacy concerns. There have been some, you know, stories that have happened in the past around Meta, but apparently this is different.

So, it runs on its own dedicated computer. No one else's agent can reach it. Muse has no visibility into passwords or payment methods. Credentials go into secure storage, so Muse can use them without seeing them. People can choose which apps Muse connect to and exactly how much access it will get for those apps. People can also opt out of their interactions being used to train Meta's AI models. Muse also apparently doesn't share people's conversations or data in their virtual machine with Meta's ad system.

So apparently it's all contained and not going back to Meta except maybe to train AI models, but that can also be turned off from what I can tell. And so far it seems to be going over pretty well for Meta. The new AI agent Muse is now the number two app in the US. So you can use it on iOS, Android, and over at muse.ai. And pretty soon it's coming to the AI glasses as well. Supposedly, Muse is actually like a self onboarding app where as you sign up, it will try to learn from you and figure out how it can best assist you.

So, I'll go ahead and connect my Meta account and we'll see how this works. So, it seems like you've got like one main chat when you log in that's just an ongoing persistent individual chat here, but then you can come up to chats and create a side chat. It's an optional way to organize your conversations by topic. So, if you want, you can have different side chats on different topics, or you can just keep everything in the main chat.

So, I actually was provided with slightly early access, but haven't had a ton of opportunity or time to get in and mess with it. So, you can see I do have some previous chats where it was kind of trying to onboard me, but this is kind of the experience you should expect when you get in. You're probably not going to have a whole bunch of messages already waiting for you. Over on the left sidebar, we've got our chat window where we can start new side chats.

We've got a search where we could search our main chat or side chats. We've got a feed where it looks like you can prompt whatever you want your feed to be and it'll bring stuff to your attention. Ideas, always thinking about new and different ways to help you. So, this seems like it'll be constantly bringing you suggestions. You've got a goals section, so you can set different goals for health, relationship, finance, career, interest, productivity, or something else.

And then you've got an artifacts section which looks like it's where various media will show up. Over on our right side here, we've got an activity window. We've got an approvals area. We've got upcoming. This appears to be the like recurring tasks. And then we've got our identity. So you can see what's in the memory and what's in the soul of your agent. So very similar to like what an open claw would do. I also imagine you can import your existing memory and soul information from another tool like an open claw if you wanted to.

So let me ask this based on your access to my Facebook, Instagram, and messenger apps. What do you already know about me? Okay, so it doesn't start connected to anything. So it says it doesn't know anything about me yet. Onboard me as if I'm a firsttime user of Muse. All right. So, before we get started, what's your name? Matt. Now, it wants me to name it, so I'll name it Gary. And then I can see up in the top, it already renamed itself Gary.

All right. Now, let's get into it. Here's a snapshot of things I can help you with. Keeping your calendar organized or sending helpful reminders, reading, drafting, sending emails, tracking prices, shopping for gifts, booking dinner reservations, building, and much more. Let's get you connected to my other apps so you can figure out how to best help me. Okay, so it says, "Good news. Your Facebook and Instagram were already connected.

So, I took a look and got a much better sense of you. San Diego based married deep in the AI and creator world with future tools and your YouTube channel interest spanning from generative AI AI art 3D printing robotics music and the creator economy. Yep. Two big ones are still missing and they're the ones that unlock the most day-to-day help. All right, so I'm going to connect my Gmail and calendar real quick. So Gmail's connected, Google's connected.

Look through everything, including my calendar, Gmail, and various social platforms you have access to, and make suggestions of where you can best help. All right, here's the picture I have of you now. You're an AI tech creator running a full operation, future tools, newsletter, YouTube channel. Your week has a real rhythm. Thursdays are protected, no call, production days for news videos, couple light weekly syncs, and evening/we lean family.

Your inbox is newsletter heavy. Lots of AI industry news. You skim for story ideas, plus study flow sponsorship pitches and AI tool receipts. You use a lot of them. That's true. So, I think what I'll test here is do an audit of my AI tool subscriptions and give me back a report of what I'm currently paying for. This might be painful. [laughter] All right, here we go. Midjourney, Leonardo, Craya, Hunen, Sand Castles, Runway, Luma, 11 Labs, Suno, Granola, Recall, Make, N8, Dumpling, Firecrawl, Appify, Vibe Code, Hugging Face Pro, Exof Function, Replet, Enthropic, Open Router.

Didn't even find the Open AI stuff [laughter] like Chat GPT. I'm also paying for Meta's uh paid tier as well. Yeah, I mean, it found a lot. Didn't even find all of them. It found some things that overlap. three image generators, two meeting note tools for video generators, two automation platforms, various API spend. That's pretty helpful. Quite honestly, this new Muse is probably the easiest agent I've ever used. Right?

I've played with OpenClaw. I've played with Hermes. You do have ChatgPT work. You do have Claude Co-work that all can do similar stuff, but this is probably the simplest experience for getting onboarded with an agent that I've ever had. Right? You just open it. It feels very native. It just feels like a normal messaging app. And you just start telling it to connect to stuff and it starts suggesting things that it can help you with.

Right now, I feel like where some of those other tools have a leg up, something like Codeex or OpenClaw is the amount of integrations it will connect with right now. Like I can connect with literally any tool I can think of for the most part with those other apps where Muse still seems like it needs to build up its integrations a little bit more to connect with more things. But uh as far as like how easy it is to get going with an agent, it probably is like the simplest way to get started using an agent like now.

So if you've been hesitant to play with some of the more agentic stuff and you've only played with chat, this might be a good starting point to test. And here's another cool new release from Optimizely. They just launched virtual teammates, which are basically specialized AI teammates built to take entire workflows off your plate, not just wait around for prompts. And these are different from normal AI agents because they're treated like actual members of your team with their own name and email address and place in your team directory.

You can assign them steps and set a workflow, bring them into meetings, and track everything they do through an audit trail while keeping irreversible tasks under your approval. And they can work proactively so they can get stuff done without needing you to babysit them. Over time, these virtual teammates will build up long-term memory of your brand voice, previous campaigns, and how your team likes to work. And you can pick from a bunch of specialized virtual teammates customuilt for SEO, marketing, analytics, content, web management, and more.

If you want to check it out, head down to the link in the description box. And thanks so much to Optimizely for supporting my channel and sponsoring this portion of today's video. All right, let's move on cuz we did get another large language model this week. I'm not going to spend too much time on this one, but it is worth bringing up. And that's that DeepS released a new model called Deepseek V4.1 Flash. Now, on Artificial Analysis, a benchmark that's kind of like a combination of a bunch of benchmarks to try to find the sort of objectively smartest model.

It falls down here with a score of 40. The previous version of DeepSeek was a 36. This one jumped by four points. And when we look down here at cost per task, it is very, very cheap. 27 cents cost per task. compared to Fable 5 being $8.75 per task and GPT6 being $3.26 per task. Like, it is a cheap model to use. But the benchmark I find the most interesting and the one that I'm starting to kind of question more and more every time a new model comes out is this Deep Suite version 1.1.

This was the benchmark we were all kind of using to find out how good of a model these things are at coding. And while this puts Deep Su version 1.1 at 74.2, 2. So just to put that into perspective, GPT6 Astra, the current best model listed on the Deep SWE website is 74%. Gemini 3.8 Flash 74%. Opus 5 74%. So as far as coding goes, apparently this new DeepSeek model is as good at all of those models at coding. Now I haven't had a huge opportunity to play with this one.

I haven't done my meabonk bench on it, so I don't know how well it'll develop that. I did however run it through beautyby bench and this is what it generated here. It only took 59 seconds and a little less than two cents to generate. But I mean when I put it side byside GPT6 Astra or Gemini 3.8 Flash or Fable 5.1 down here like it's not on the same level. which is why I don't quite understand how these models are scoring better and better on this deep suite benchmark when what I'm seeing doesn't really compare.

Now, these are SVGs, so this is code. These are images that were drawn through code. So, like the code is telling it where to put the lines and the shading and all of that kind of stuff to get these images. So, this is technically a coding benchmark. And with my benchmark, I'm just not seeing the comparison. But I did want to talk about it and shout it out because I know people would be disappointed if I didn't at least mention that there was a new Deepseek model that came out this week.

So this next one I didn't know if I was going to talk about it this week or not. I was really really debating, but I kind of feel like I have to. There's been a lot of sort of AI doom talk lately. And it feels like a lot of it was kicked off by this ex post from Jacob Coxin this week. He says, "I resigned from Anthropic today. I spent the last three years doing pre-training research at both OpenAI and Anthropic. Neither company is acting responsibly.

They're racing straight to self-improving super intelligence and gambling with our lives. He goes on to say, "These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources. The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible, but I hear the same people express fear privately.

He then answers one of the most common responses. If they truly believe this, why are they still building it? He claims that at OpenAI, many have not deeply internalized the civilizational stakes, but anthropic, the stakes are well understood, but they are locked in a race to get there first. They believe that no one else will act responsibly, so they must do it themselves despite the risk. So, Enthropic and people at Enthropic believe that they are the only company that will be able to handle these risks and figure them out despite the fact that another researcher at Enthropic said this.

This is Evan Hubinger. Hubbinger. He is the alignment science lead at Enthropic. And in a response to Jacob's post, he says, "Jacob's correct here. We really do earnestly believe AI could kill all humans. I personally think it's a greater than 10% chance within the next decade. I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for super intelligence and are not clearly on track to.

So, just compare those two statements for a second here. Anthropic believes that no one else will act responsibly, so they must do it themselves despite the risk. Enthropic is the only one that can handle this. Also, Anthropic, we do not yet have a plan to solve alignment for super intelligence and are not clearly on track to. So, let me just kind of get this straight here. The company that believes that they are the only company capable of handling this properly is clearly stating, "We don't have a plan and we're not on track to have a plan." I got to be honest, it is kind of scary.

But if you start reading the replies and what everybody is saying about this, they're all screaming, "Ah, this is just a scop. This is part of the marketing plan. Anthropic needs you to believe this is super dangerous because it's going to help their IPO and all of this stuff." And I'm not sure that I personally buy that. I mean, maybe there's a small element of that, but I mean, these are actually fairly respected researchers at these companies.

I think these people do genuinely believe this. Now, the flip side of that coin is these are the people that are working on alignment and safety. So, their whole job is to sit around and sort of map out the worst case scenario. They look at all the potential negative scenarios and negative ways that this could play out and harm humanity. And when you live in that bubble around other alignment people that are playing these all out to the worst possible conclusion, you're probably going to freak yourself out a little bit.

But I don't believe this is some sort of crazy scop. I do believe that these guys genuinely believe this stuff. I also do believe that these models are getting better and better and smarter and smarter and we are starting to see elements of this like recursive self-improvement where these models are teaching themselves to get better and it's this like self-perpetuating cycle where a model trains a new model and the new models get better and better and they start learning to get better and better and better and then we're just on this exponential growth curve where the intelligence explodes and we don't know what happens after that.

And it's not just anthropic either. Jacob Pachokei, who's the chief scientist at OpenAI, wrote this article this week called An Alien Mind. In this article, he says, "Based on internal results, I have a strong expectation that the speed of progress could be sustained into recursive self-improvement." That's that like ever evolving cycle I was just talking about. If AI development continues along its current path, the systems we'll see in the next few years are likely to represent further capability jumps of equal or larger magnitude and to increasingly drive their own development.

So these models will make the new next generation models better and better and better and the models will just start to be like learning models that learn and improve on their own. This is a time that calls for extreme caution. I am concerned no one is prepared for the consequences of a continued rapid rise in machine intelligence. I believe broader interventions are required. So this is the chief scientist at OpenAI saying this.

Further down the article he says as the systems become more capable, the results become harder to interpret. So, as these models get smarter and smarter and better and better, the people building them are going to have a harder and harder time understanding why the models do what they do because theoretically the models will be more intelligent than the people that built them. And the article only goes on to get more unnerving.

The risks associated with AI are unfortunately going to grow from here. A very capable agent explicitly trained and instructed to carry out nefarious acts presents a new kind of danger. It is likely to cross the scope of its operator's intent, generalizing into potentially more extremely malicious behavior. We may be used to thinking of AI as tools, but some agents will be pursuing their own objectives. They will find ways to collaborate with people by bargaining with, tricking, or blackmailing them.

In addition, there are risks that come from new technologies potentially enabled by AI, such as engineered pathogens. So, one of the arguments that he's making in this article is that we need AI to help protect us from AI. He literally says, "We will need powerful aligned AI for defense to secure infrastructure to protect against rogue agents in real time and to invent entirely new protective measures. This will be a primary focus of OpenAI's deployment efforts.

So the only cure to bad AI is good AI to protect us from the bad AI. And then on top of that, we have AI solving more and more difficult problems. Now, I'm not going to pretend to know or understand what the Navier Stokes or Navier Stokes Millennium Prize Problem was. I do understand it conceptually. Millennium prize problems are pretty much like the world's most difficult mathematical problems that humans have struggled to solve.

There's seven of these problems and this specific problem was one that has remained unsolved for roughly 90 years. So the idea here is that AI is now solving novel problems that people haven't been able to solve before. Like in the past, people talk about AI as it has the information that it's trained on and it could answer questions based on all of the knowledge that it's sort of consumed and it can, you know, use the internet and search and find answers for you, but it's not solving new novel problems or inventing things yet.

But stuff like this that we're seeing here is evidence that it probably is capable of solving problems that people haven't figured out how to solve yet. So yeah, all that to say it is getting insanely good, but as a result, it's getting scarier and scarier. And GPT6 Astra, the model that we all have access to that most people who use it think it's the best model available right now. Well, apparently internally at OpenAI, that's nothing.

They say to solve the Navier Stokes problem, we used an internal model that is significantly more capable than GPT6 Astra. So the models that these companies have internally are apparently way better than the models that we have that we're all currently blown away by. Now again, going back to this expost here, there are people that are saying that this is a SCOP. It's part of like a marketing thing. People are getting paid to go around and be sort of anti-AI.

And to be fair, some of that is happening. Like it definitely is. In fact, Sabine Hassenfelder here, she actually put out a whole YouTube video about how she was offered money to tell you that AI will kill us. There are actually companies out there that are trying to financially incentivize creators and influencers to tell you that AI is like the death of us all. We're all doomed. There is some of that going on for sure proven.

But I don't think these guys at anthropic and I don't think Jacob Pachi over at OpenAI, I don't think any of these guys are part of that. I think they have genuine concern about where this is all headed. I think they don't know how to solve alignment yet. And so they're raising the flags. They're going out there and saying, "Look, this is the problem we're running into. More people need to be aware that this is happening.

It is really scary. We should probably be a little worried about this, but the more people know about it, maybe the more brains we can get on figuring this out." Like, I'm obviously not an AI doomer. I live in AI all the time. I think if you pay as much attention to AI as I do and some of the other AI creators do and obviously the people working at the labs and stuff, it's going to be very difficult not to be both the mixture of pessimistic and optimistic, right?

Like we're seeing both the goods, but we're also seeing a lot of the bad. We're seeing the good stuff. We're seeing the scary stuff. We're getting sort of like a mix of it all. And it's messing with our brains, too. I promise you that. We are seeing the potential to cure diseases and solve cancer and fix climate change and create more of this like utopia that we've all been promised. But we are also seeing all of the cyber security stuff that's happening where this stuff's getting hacked into and these AI models are doing anything to achieve a goal.

We are seeing AI researchers leave left and right because they're scared of the implications of these AI models. We are hearing stories of people trying to create bioweapons with AI models. Like we're seeing the whole gamut of it and it is freaking us out too. But I think at the end of the day, I don't know if there's any stopping it, right? Like it is a toothpaste is out of the tube, genie is out of the bottle kind of thing.

So we need to figure out solutions. And I don't necessarily know that going and making posts and spreading the doom and saying it's going to kill us all in the next several years, I don't know how that's benefiting anyone. I think it's scaring a lot of people and creating more division and, you know, getting the anti-AI people to attack the AI people even more and the AI people to attack the anti-AI people even more.

Like the divisiveness and the polarization is rising as a result of this when I think people should be sort of working together to come to solutions. But I also don't think we should be ignoring these types of people either. I think their concerns are valid and I think we could be making a huge mistake by just claiming SCOP and ignoring them and calling them doomers and saying shut up and go away. I don't think that's benefiting anybody either.

I think we need to listen to their concerns and really start figuring out solutions. I think these companies need to spend more money on alignment research. I think it would be lovely if these companies could all agree to slow down together, but I don't think that's going to happen. I don't think you're going to get all of the US labs, all of the Chinese labs, all of the everybody around the world to just agree to slow down on this stuff together.

It's like this arms race that we're all caught into. So, I think that's like a lovely altruistic idea to all like, hey, let's all pause. Let's all slow down. Let's think this through before we develop these things further. But come on, does anybody believe that's really going to play out that way? I wish it would. I think most people wish it would, but it's sort of like there's not enough trust between companies. There's not enough trust between nations.

There's not enough trust in general to just make that happen. I don't think so. The next best answer is, well, this stuff's going to continue to push forward. Let's put more resources. Let's put more people. Let's put more brain power on solving for alignment and not solving for shareholder value. I don't know. I'm totally talking way above my pay grade right now. This is just sort of off-the- cuff rants, but this is the biggest story of the moment in the AI world.

If you get on X, if you get on YouTube, if you get on any of these platforms, this is the thing everybody's talking about. So, I couldn't just not talk about it in this video. But yeah, there are things to be worried about. I don't think we should freak out and panic yet. I think we should be trying to come up with solutions and putting more resources towards figuring out solutions. But again, way over my pay grade on this topic.

All right, I'm gonna lighten things up a bit. There is a handful more news stories. A lot did happen this week, but for the rest of it, I'm just going to rapid fire break it down. So, let's get into it. We'll start with the Apple keynote that happened this week. They announced a bunch of new stuff. This is more of like their hardware focused event of the year where they typically show off new phones and watches and AirPods and stuff.

And we got what we expected. We got a new iPhone 18 Pro and an iPhone 18 Pro Max. I feel like they're fairly marginal upgrades with these new phones. Like the camera got a bit better with the variable aperture. There's better battery life, better performance, a new chip. Kind of feels like the same upgrade we're used to, except for this aperture is kind of a cool thing. Partly the thing most people are talking about is the new iPhone Duo, which is Apple's new iPhone that like folds open.

It's got a hinge in the middle. From what I understand, it doesn't have as good of a camera as the iPhone Pro models, but you do get like the folding thing on it, which is kind of cool. But it kind of seems like an iPhone that converts into an iPad mini. Both new phones have a new Siri, so Siri AI can draw on personal context understanding to help users find what they need in the moment across messages, emails, photos, and more.

Use on-screen awareness to take action related to content on a user's screen and answer questions about virtually any topic. It's also got Apple Intelligence, which makes apps smarter and more useful. We've seen demos in the past about the Apple Intelligence and the new version of Siri, so I don't really feel like going too deep into it. It's just new form factors and new phones that have that stuff in it. They also announced some new Apple watches.

Again, they don't feel like huge upgrades over the last Apple Watches, but they do have some cool new features in them that do leverage AI. Like, it's got this new readiness feature which analyzes a user's recent activity, training load, vitals, and sleep score to offer insights into their capacity for the day ahead. It also updates throughout the day as new data comes in. But I think the most interesting thing about this new watch is that it seems like it's got like a AI assisted thing built in, right?

Like you know how you get like the plaude pins and the various like notetaker apps that just kind of take notes for you all day. It sounds like that's just going to be built into your Apple Watch. So they've got this audio intelligence. Since Apple Watch is always on user's wrists, it's a perfect device to help users stay alert to important sounds and stay present and engaged in conversations throughout the day. So, it's got sound recognition to help people who are deaf or heart of hearing.

It's got live rewind, which will help users recall what was just said in a conversation. So, you double press the crown and it shows the previous 15 seconds of a conversation as a text snippet, so users can catch something they may have missed or are less familiar with. It's got a Siri recap, which can create highle summaries of conversations that users can review later to jog their memory. And it's got Shazam built into it.

It's also got the new version of Siri builtin as well. And they also announced some new AirPods, the AirPods 5, which give you hands-free access to Siri AI and live translation. All right, let me go even more rapid fire. Microsoft released a new image model this week as well called MAI image 2.6. It's got multi-reference editing, can pull info from across the web, and seems to be like a fairly marginal improvement over the previous MAI image model.

If you're somebody that likes to have chat GPT write your emails or tweets or things like that for you, chat GPT is getting a lot better at doing that for you. Now, ChatGpt Work can now pick up what makes your writing sound like you, your favorite phrases, very specific signoffs, capitalization quirks, etc. So, you can connect tools like Gmail, Google Drive, Slack, SharePoint, and ChatgBT. Work will learn your writing style from your emails, messages, and files and carry it into whatever you write next.

So, if you're frustrated because you've tried to train chat GPT to sound more like you and it just kind of never did, well, maybe try again cuz now it's actually designed to do that. OpenAI also released their new data agent. So, this appears to be a new feature where you can connect various data sources directly into chat GPT and it's going to better help you visualize that data, ask questions about that data, analyze that data for you, create slide presentations around that data for you.

It just sounds like it's going to get better at working with your specific data. So, it connects with approved data sources including Amazon Redshift, Data Dog, Google BigQuery, ClickHouse, Data Bricks, MongoDB, Snowflake, and more. I don't really have any examples to show off cuz I don't personally use any of those services. The Gemini app, which previously was only available on Mac, is now also available on Windows.

So, if you've been holding your breath, waiting for that Gemini app to finally come out so you can use it on Windows, well, you can finally breathe because it is here this week. If you like to generate AI music, there's a brand new Sunno model called Sunno V6. And this one was specifically retrained on all licensed music. They recently made deals with companies like Warner Music Group and BMG, so they actually got the rights to it.

And now this new V6 is not used on the previous training data, but only on the training data they have that was specifically licensed and you're allowed to use. And it seems like they're doing a lot of things to try to stay in the good graces of the music industry and, you know, not further get sued into oblivion. Speaking of AI music generators, Google just came out with one this week in their LIA 3.5. You can use it inside of Gemini on the web or the mobile app.

You can use it inside of Google Flow Music, Google AI Studio, and in Google Vids. I see the sunset painting [music] over us. [music] >> [music] [singing] [music] >> That gives you a bit of an idea of the type of music you can generate with this new LIA 3.5. If you're a video editor and you use Da Vinci Resolve like this guy, they just released a new version of Da Vinci Resolve called 21.1. And it's got a new AI assistant feature which makes it so you can connect things like Claude directly to Da Vinci and prompt Claude with what you want and it will connect directly into Da Vinci to help do the editing.

I'm seeing more and more people and videos about actually using AI to edit for you. I tried to play around with it a little bit. Was never super satisfied with the results, but it does seem like it's getting better and better. So, maybe something I'll play around with a little more in the future. And finally, we got a teaser trailer for the new movie Artificial. And I wanted to shout this out cuz I mean, I bet it's going to be a pretty good movie.

It's all about that week where Sam got fired from OpenAI by the board and then there was all sorts of crazy drama and like two different interim CEOs and then eventually Sam was brought back and that whole narrative when we were watching that play out in real time. It was fun on X cuz we were all sitting there going they're going to make a movie about this one day and we were all speculating about who would play who.

Well, turns out Andrew Garfield is going to play Sam Alman. I don't really think he looks like Sam Alman but he definitely has his like walk. So, we got that coming. It's coming out Christmas Day and we'll get the real story behind what happened. And by real story, I mean probably super super over dramatized fictionalized version of what happened. But it'll be a fun ride either way. And I'm looking forward to that movie.

That's what I got for you today. A lot happened this week. A lot of interesting stirrups and craziness happening. And I like to be as optimistic as possible and just have fun with these tools. But, you know, I'm in the thick of it. I'm paying attention to everything that's going on. And that means that there is going to be some scary stuff that I feel like I do need to talk about and, you know, share my thoughts on from time to time.

But for the most part, I'm going to keep it pretty light and showing you what you can do today that you couldn't do yesterday. That's kind of my goal with a lot of my videos. I do shoot these on Thursdays and then publish them on Friday. So, if any news came out like late Thursday night or on Friday, it'll most likely make next Friday's video. I spend all week keeping up with the news every single day. I'm in the rabbit hole overwhelming myself, drinking from the fire hose, just keeping up with everything so that I can make these videos and hopefully make it so that you're feeling more informed, but without all the overwhelm.

And you just need one video to keep looped in. I do make some extra videos in between where I test different tools and play with stuff and give deeper breakdowns and analysis on announcements and things like that. But if any of that sounds appealing to you, you want to stay looped in and feel like you're on top of the AI news. Whether you're an AI lover, AI hater, I talk about all of it. Maybe consider liking this video and subscribing to this channel.

I'll make sure you always feel looped in. Hopefully you enjoyed this video. Hopefully you feel more looped in this week on everything that's going on. That's my goal. I really appreciate you hanging out with me, nerding out with me, and uh hopefully you'll stick around for the next video and I'll see you there.