OpenAI의 GPT Next 및 캘리포니아 AI 규제 영향

### AI 뉴스 업데이트: Vid Summit과 그 이후의 주요 개발 사항
* 🎥 미국 텍사스주 댈러스에서 열린 Vid Summit 행사에서 **ChatGPT 사용량이 주간 2억 명으로 급증**하여 작년 가을 1억 명에서 두 배로 증가했다고 발표했습니다.
* 🤖 일본 OpenAI의 CEO는 **GPT-3에서 GPT-4로 100배의 도약**을 이룬 것에 이어, 곧 출시될 'GPT Next'는 **GPT-4에 비해 약 100배의 성능 향상**을 제공할 것으로 기대한다고 예고했습니다.
* 📜 현재 캘리포니아에서는 오픈 소스 AI 모델과 기존 사용자 기반에 큰 영향을 미칠 수 있는 AI 생성 콘텐츠에 워터마크를 의무화하는 법안인 **AB 3211이 검토 중**입니다.
* 🔍 이 법안을 분석한 결과, 이 법안은 AI가 생성한 콘텐츠를 규제하는 것을 목표로 하고 있지만, **규정 준수 문제로 인해 기존의 많은 AI 모델이 캘리포니아에서 불법이 될 수 있다**는 것을 알 수 있습니다.
* 📦 Amazon의 새로운 Alexa 어시스턴트는 **Anthropic의 Claude를 활용**하여 기능을 강화할 예정이며, **Claude Enterprise는 이제 개발자를 위해 50만 개의 컨텍스트 창과 GitHub 통합 기능**을 제공합니다.
* 💻 Vid 서밋에서 새로운 AI 도구인 스포터 스튜디오가 발표되어 **유튜버에게 브레인스토밍 기능과 시청자의 선호도에 따른 맞춤형 콘텐츠 제안**을 제공합니다.
* 🌐 코인베이스는 혁신적인 크로스오버를 통해 **AI 에이전트가 전적으로 관리하는 최초의 암호화폐 거래**를 보고하며, 암호화폐를 사용하여 여행 예약과 같은 복잡한 작업을 수행할 수 있는 AI의 잠재력을 강조했습니다.
* 🎓 마지막으로, 런던의 새로운 교육 이니셔티브는 **가상현실 교실에 AI 교사를 도입**하여 학생들에게 개인화된 학습 경험을 제공하는 것을 목표로 하고 있습니다.

곧 출시될 OpenAI의 GPT Next는 GPT-4에 비해 무려 100배의 성능 향상을 약속하며 AI 기능에 혁신을 가져올 것입니다.

so I'm out in Dallas, Texas, for the Vid Summit event. Um, here in my hotel room, going for a dark and moody look for today's video. I'm starting to lose my voice a little bit from talking so much during this event, but I have been keeping my finger on the pulse of AI news, and I want to break it down for you real quick in today's video, starting with the news out of OpenAI and ChatGPT. This week, OpenAI shared that 200 million people use ChatGPT every single week, which is up from 100 million last fall. But outside of the fact that they're continuing to grow, we got a little tease from OpenAI out of Japan this week as well. The CEO of OpenAI Japan was talking about the next iteration of GPT, and at this event, he called it GPT Next. When talking about the GPT models, he said that comparing GPT-3 and GPT-4, the performance is nearly 100 times higher. He then went on to say that the AI model called GPT Next

will be released in the future, and it will evolve nearly a 100 times based on past performance. So, from GPT-3 to GPT-4, we got a 100x increase in performance, and apparently from GPT-4 to GPT-next, we're going to get another 100 times increase in performance. What that actually looks like in the real world, who knows yet. Now, I want to jump back to something that I talked about in last week's news video, where I mentioned that OpenAI, Adobe, and Microsoft support California Bill requiring watermarks on AI content. So, there's this Bill going through the California legislature right now called AB 3211, which requires watermarks in the metadata of AI-generated photos, videos, and audio clips. When I talked about this in last week's news video, I mentioned, This is something I feel like most people can get on board with. It makes sense that AI-generated images, people should let others know that this was generated with AI. However, when I mentioned it last week, I just kind of quickly brushed over it because on the

surface it makes a lot of sense to prevent deep fakes and things like that. People should just let others know if, Hey, this was generated with AI, or Hey, this is a real photo. However, there's a lot more nuance to it that I didn't really consider when I brushed over that article earlier. And I recently came across this Reddit post, which actually makes some good points. And the Reddit post is called California Bill set to ban civit AI, hugging face, flux, stable diffusion, and most existing AI image generation models and services in California. Now, the bill itself didn't say it was going to ban all of these places. This is sort of extrapolating out what could happen if this bill passes. But basically, this Reddit post is saying that this is going to be really, really difficult for open-source AI models because there's so many AI models out in the wild already that already exist. They're already on people's computers. You've got stable diffusion, you've got flux, you've got animate diffusion, you've got all of these AI generation tools.

that already exist, and it would be very, very difficult to retroactively add watermarking to all of these existing tools in the open-source community. So, this Reddit user basically says, If I read the bill right, essentially every existing stable diffusion model, fine-tune, and Laura would be rendered illegal in California. Sites like civid AI, hugging face, etc., would be obliged to either filter content for California residents or block access to California residents entirely. Now, we've already seen some of these companies out there completely block access to, like, the EU because of regulations that these AI companies thought were sort of too restrictive to continue in that area. This person is basically saying that could happen in California with all of these AI-based tools if this bill is passed. Now, additionally, this legislation would ban the sales of any new still or video cameras that do not incorporate an image authentication system. Now, the confusion lies in the definition of newly manufactured. So, they go on to say, If somebody wants to save money and buy a used camera, is it going to become

illegal to buy AED used camera from somebody that doesn't digitally Watermark and say that this was a real image taken with a real camera. Anyway, I will link up to this Reddit post so you can dive a little deeper. What I actually did myself is I jumped over to Claud and I actually uploaded the bill itself. I said, Please give me the bullet points of what this bill aims to do and how it will impact AI companies. I got a quick breakdown of that bill: labeling synthetic content, Providence detection tools, adversarial testing, restrictions on removal tools, requirements for recording devices, etc., etc. The impact on AI companies: increased development cost, compliance burden, product modifications to existing products, potential limitations on functionality, etc., etc. And then what I did was I actually then took all of the criticisms that this Reddit post shared right here. I pulled that into CLA as well and said, Here's a write-up on Reddit that criticizes the bill. How accurate are these criticisms? Compare the actual bill with the statements made in this Reddit post and tell me where the Reddit post is correct.

and where it's incorrect, and so Claude took the two and kind of gave me a quick comparison so that I could kind of find out if there were any issues with the logic in this Reddit post here. And here's essentially what the comparison analysis came back for me with Claude. I'm not going to read it all; I'm just going to pick the important bits out here. So technological feasibility: the post argues that the requirements may not be technologically feasible, especially for open-source developers, and Claude says this post concerns about technological feasibility, especially for open-source developers, is valid. The Reddit post also claims the primary provisions may become effective immediately upon codification; however, the bill does actually say this will become operative on July 1st, 2026. Impact on existing models: the post suggests that all existing stable diffusion models, fine-tunes, and luras would be rendered illegal in California, and the analysis was while the bill would require significant changes to how these models operate, it doesn't explicitly ban them. Instead, it requires them to implement Providence Data Systems, so the post

overstates the immediate illegality, but the concern about the impact on existing models is valid. Under camera requirements, the post correctly notes that the bill would require new recording devices to incorporate Providence Data Systems. However, the post overstates the potential impact on phone purchases, as the bill allows users to opt out of applying Providence data. Anyway, I just wanted to bring this up again and spend a little bit more time on it in this video because I kind of brushed over this bill in the last video, but it sounds like there are much further-reaching impacts than what I initially sort of gave it credit for in that last video. Another thing I mentioned in the last video is that we're going to get a new voice assistant inside of our Amazon Alexas. I mentioned I bet they're probably going to use Amazon's Titan models because Amazon has their own foundation models just like GPT-4 or Anthropic's Sonnet 3.5. Amazon has their own Titan models, so I would assume that the Alexa would use

the Titan models. Well, it came out this week that Amazon's new Alexa assistant is actually going to use Claude, which also makes sense because Amazon is a very large investor in Anthropic. While we're on the topic of Anthropic, Anthropic rolled out Claude for Enterprise this week. It's designed to help organizations securely collaborate with Claude using internal knowledge. The Claude Enterprise plan offers an expanded 500,000 context window. That's a pretty huge context window, more usage capacity, and Native GitHub integration, which will allow you to work on your entire code base with Claude. Now, if you ever tried to code anything with Claude, which I've spent a little bit of time doing, that is kind of one of the issues, is that it sort of loses track of code earlier up in your conversation. Well, with this GitHub integration, you'll be able to just sync your Claude account over to GitHub, and it will just every single time look at the whole code to figure out where to modify things. So, if you use Claude for coding, the Enterprise having GitHub integration is pretty huge.

hopefully, they allow the GitHub integration on, you know, the smaller Pro plans for 20 bucks a month. I'm just one dude over here playing around with this. I don't need the Enterprise plan, but damn, I really want that GitHub integration. Claude also did another solid for developers this week by announcing the anthropic quick starts repo. So, it's a collection of projects designed to help developers quickly get started with building deployable applications using the anthropic API. They put it up on GitHub over on the anthropic GitHub account, and they put up their first quick start here on how to create a customer support agent using Claude. So, they basically gave all this code here, and you know, I'm guessing you can take this code, pull it all into Claude, and you can sort of use this as a starting code base for coding your own Claude customer support agent. You just pull all this code in and then start telling Claude, Here's what I want the code to do. There's also a detailed readme here on GitHub that walks you through step by step exactly how to set

all this up for the first time. So, if you want to build your own cloud-based customer support agent, well, Anthropics got you covered. Having a website is just a great idea for anybody. It's always smart to have a place on the internet that you own, that's not controlled by the big social media companies. That's why for this video, I partnered with Hostinger.com. Not only are they one of the least expensive hosting platforms, they also have a load of AI features that are really cool. If you head on over to Hostinger.com/Wolf, you'll be presented with this page where you can claim a special discount. If we click Claim Deal and then take a peek at this Business Website Build Plan here, you'll notice just a load of AI features. You can purchase just 1 month or all the way up to 48 months in advance. And if we scroll down here, click on this coupon code button, and enter the coupon code Matt Wolf, we can watch as another 10% gets taken off the price tag once you're

logged in. It's easy. Just click Add a website, choose the hosting, your website builder, because that's the one where AI pretty much does all the work for you. We'll go ahead and click Start creating for this particular site. Let's go ahead and use a landing page to build an email list. Let's pretend I'm building the Future tools newsletter. And then for the description, I can simply speak out exactly what I want this site to be. This is a lead generation website for the Future tools newsletter. It should tell people why they should subscribe to the newsletter and give them the option to opt in. Then we click Create website and then watch as Hostinger does its AI magic to build the website for us. Less than a minute later, I have a website built, ready to go. Subscribe button, join button, details about what you'll get when you opt in. Obviously, we'll go back through and change the text and images, but we're ready to start capturing leads by clicking on Edit site. We've got a simple drag-and-drop builder where I can

redesign the site however I want, and then over on the left sidebar, I've got all sorts of AI tools like a logo maker, a section generator, an AI blog generator, an AI writer, an AI image generator. All the AI things are built into this platform over on hostinger.com. It is quick, it is easy, it is inexpensive, and you can make it even less expensive by going to hostinger.com/Wolf and then using the coupon code Matt wolf at checkout. Thank you so much to hostinger for sponsoring this video. A few months ago, one of the original co-founders of OpenAI, Ilas Eses, he was also their Chief Scientist. He actually left OpenAI. He was a big part of the whole Sam Altman getting kicked out of OpenAI drama. After that all happened, he kind of, I guess, lost relevance at OpenAI and eventually left the company to start his own new company called SSI or Safe Super Intelligence. Well, apparently Safe Super Intelligence just raised 1 billion dollars. A company that's only 3 months old just raised a billion dollars for

20% of the company now. Back when Ilia started this company, he posted on Twitter they're not going to be too focused on any customer-facing products. They want to be a research lab. They want to develop the technology. I'm sure at some point they're going to have to have some sort of customer-facing product so that the company can actually make money, money, and you know, get a return for the investors. But as of right now, I think their plans are just to put their heads down and build. And what's crazy is this is a 10-person team that's only 3 months old, and they just raised a billion dollars to be able to do this research. Seems like lately X has been having all sorts of issues. They were recently banned in Brazil. Well, now it looks like X is having issues with the EU. So, one of the things that's really powerful about Gro is that it's got access to real-time Twitter data. So, you can ask it questions, and it will basically search Twitter to see what people are saying about that.

topic that you asked. Well, if you live in the EU now, grock isn't going to actually pull in any data from people in the EU. So, if we ask it questions, people in the EU might have the perfect answer, but apparently X is going to exclude that data due to some of the policies in the EU. Google had a few new rollouts this week as well. They're adding the improved search in Google Photos. Now, this is being released right now in the US and only to Early Access members. If you go to labs.google.com, you can see they have the ask photos option, and then you can join the waitlist, and then ideally they'll be sort of letting people in quickly. I don't know how quickly they're letting people in. I'm not in yet, but I'm on the waitlist. But basically, what this allows you to do is it allows you to search your Google Photos with natural language, like typing things like Alice and me laughing, and it'll find all photos of those two people laughing, or

what did we eat in Bangkok, and it will pull up images of whatever you took pictures of in Bangkok that have food in it. Here's a little animation here, Emma painting in the backyard, and then it does a search of all of the photos of this child painting in the backyard, or what did we eat at the hotel in Stanley, and then it brings up, you know, food from that place that they went. So, pretty cool that you can use this natural language top 10 things we saw in Idaho, and it will just pull in photos from your Google photos related to that query that you gave it. Again, this is only in the US, and it's only if you're in Early Access. If you go to labs.google, you can join the waitlist. Google's also rolling out virtual Tryon with dresses in their AI shopping tool. If you find dresses in Google shopping, you could try them on a wide variety of models. It's not like you upload your own photo, and it puts you in, and it actually shows what you'd look

like in it, but you can pick from different models, I guess. Try to find the model closest to you and then try the dresses on that model, kind of what it's doing. I have a feeling in the future if you want to put your own face and body into it, they're going to have that, but right now it's just these models. So, we've got some news in the overlap of crypto and AI, something I very rarely talk about, but Coinbase sees their first crypto transaction between AI agents. It says here Coinbase CEO Brian Armstrong recently oversaw his first crypto transaction completely managed by artificial intelligence bots. So, a bot designed to carry out a certain task used crypto tokens to interact with another AI agent and acquire AI tokens. These AI tokens are essentially strings of data that enable algorithms to learn from information. So, he went on to explain that the reason AI agents aren't effective right now is due to their lack of transaction capabilities. Without a payment method, they struggle with basic tasks like booking plane tickets or hotels and even managing social media.

tasks. Beyond just creating content, AI agents cannot get bank accounts, but they can get crypto wallets. They can now use USDC on Coinbase to transact with humans, merchants, or other AI. Those transactions are instant, global, and free. So, we can see why the sort of overlap of crypto and AI agents could be valuable, because they can go use that currency to book plane tickets or a hotel or do all these agentic tasks like, Hey, I'm going to Dallas for Vid Summit this week. Book my airfare and hotel. I need to check in this date, check out this date. I only want to fly these airlines, so make sure it's on one of those airlines, and make sure I don't stay at this hotel, and it needs to be within one mile of the convention center, right? You can give it all of that information, and now it can actually go and do all of those steps, even if it requires making purchases on your behalf. Now, that also sounds a little bit scary, like, What if it goes and books the penthouse and a private jet? You know, you

don't want that kind of thing to happen, so obviously there needs to be some decent guard rails in place before this can be rolled out publicly. But we can see where they're going with it all. Here's an interesting one: AI will teach this class a lesson but won't hang out in the Teachers Lounge. So this article on Tech Radar here is talking about how there's a course at London's David Game College, which is bringing virtual teachers to real classrooms. So basically, they're going to have AI teach lessons inside of virtual reality headsets. The idea being that the lesson is tailored to each individual student. They go into this virtual reality world, there's an AI teacher that has the information that they're teaching, they're learning from this AI teacher, and if they're struggling with stuff, the AI teacher will sort of work with the student more on those areas, and each student in the class has the training sort of more tailored to them. Now there are like overseer teachers also just to make sure this is a success and to actually give

real world human feedback to the other students if necessary, but who knows, this could be the future of education. I kind of feel like people will want the option. Like, I bet there's a lot of people that really don't want to learn inside of a metaverse by AI. They want the old school teacher sort of thing, but there's also going to be some people that are like, I'm an introvert. I don't want to leave my house. If I can learn in virtual reality, sign me up. It's the only way I'm going to get a degree, right? So, I can see people loving this, and I can see people hating this. Either way, I think it's going to be fun to watch how it plays out. So, this is pretty interesting here. Robert Yang here, the co-founder and CEO of Altera, showed off what he calls Project Sid, the first simulations of 1,000 plus truly autonomous agents collaborating in a virtual world with emergent economy, culture, religion, and government. Check this out. They basically created this Minecraft world and loaded

it up with a whole bunch of AI agents, and every single character in this world is all AI. There's no humans in this world. In the video, they say that these characters in this Minecraft world set up a market. They agreed to use gems as the currency, and they built a whole economy within the game. This video shows off, I guess, the three coolest emergent things that happened when they set this up. So, go ahead and just check this out. I'll just press play, and we can watch it with Olivia's dream. Olivia worked as a farmer, providing food for the entire civilization until she was inspired by the tales of the village explorer, Nora. Olivia attempted to go on her own expedition, but townsfolk pleaded with her to stay, and Olivia actually listened. She gave up her dream for the village for now. Next, we simulated parallel worlds under Trump and Kamala. For each one, citizens have a shared constitution in Google Docs that they can vote on to amend. Under Trump, the simulation passed new laws to increase the amount of police in the

world under comma. They focused instead on Criminal Justice Reform and removing the death penalty because our agents are social and grow over time. They're impacted by group dynamics but also use their individual power to change the system. I know it looks silly, but this is the first time we see agents can form a democracy and can govern themselves. Lastly, we have the missing villagers. When agents were concerned about missing townsfolk, villagers took initiative by leaving their post and working together to illuminate the town with torches, hoping to create a beacon for lost members. What's surprising about these agents is how, once they got deeply worried about their fellow villagers, they bent together and changed their plan. You should definitely watch the whole video. I linked up the Tweet with the video in the description. It's definitely worth watching because the idea is just, again, to test it with Minecraft but then to apply the same ideas to actual real-world tasks that you would want these guys to accomplish and not just, you know, gaming tasks. We have a new AI video generator in town.

once again, this one's out of China, and it's called Minia Max. Once again, they're comparing it to Sora, claiming that it's the next big AI model. This article here claims that people are making entire short films with this new AI video generation app, but you can actually use it over on this website, hailoai.com. The site's actually in Japanese. I had it translated, but we can look at some of the featured video clips that are available here, like this kid playing a video game and then the character coming out of the screen. And I mean, it looks kind of similar to a lot of the other video generators we're seeing right now, like Cling, like what we're getting out of Runway, what we're getting out of Dream Machine. It looks like it's on par with those. Can I really say that this is a whole lot better? No. I mean, I'd say it's pretty much on par with what we're seeing. Just for fun, I had it quickly generate a man with a beard typing at a computer, and well,

that's not a computer, that's a typewriter. But it got the man with the beard, and it looks pretty decent. What's cool is if you just go to this URL, you don't even have to log in or anything. You just type in a prompt, click generate video. Now, it's not the fastest in the world; it might take a few minutes for the video to generate. But as of right now, from what I can tell, it's free to use. Just enter your prompt, click generate video, and then you can download the video. I'm not even logged in or anything. Luma AI also rolled out some new features this week with their dream machine 1.6. You can now give it camera commands like move left, move right, move up, move down, push in, push out, pan left, pan right, orbit left, orbit right, crane up, crane down, things like that, which you weren't able to do in the previous versions of Luma. So, if you have a Luma dream machine account and you go to enter a prompt, you can type camera and

notice when I type camera, it brings up this drop-down box with all of my options for camera moves, so I can do camera move left and then just add the rest of the prompt that I would have added anyway. Here's what I got when I typed a wolf howling at the moon with camera moved left. There's not actually much animation going on with the wolf, but it's definitely moving the camera left. Something more to play around with. This is literally my first time testing it, but you definitely have more control now over the movements of your camera. Hopefully, the rest of the animations still work as well. Again, if you've got Luma Lab's Dream Machine, go play around with it, see what you can do with these new features. We've got some new AI news out of YouTube. They're developing AI detection tools for music and faces, plus creator controls for AI training. So, this week, YouTube announced a new set of AI detection tools to protect creators, including artists, actors, musicians, and athletes, from having their likenesses, including their face and voice, copied.

and used in other videos. They're also working on coming up with a solution that addresses the use of content to train AI models. Obviously, a lot of creators have been coming out saying that they're very frustrated that all of these companies are using their YouTube transcripts to train their models. So apparently, YouTube is working on some sort of solution for that as well. Spotter, the company who is actually the main sponsor of the event I'm at right now, Vid Summit, they actually announced during this event that they are rolling out their YouTube brainstorming product called Spotter Studio. Now, it's actually funny 'cause the video I released earlier this week was about the tools that I actually use in my life and my business to speed up my own processes and workflows. And in that video, I showed off Spotter Labs, which was the precursor to this new Spotter Studio that was like the beta version. This is the full rolled-out version, and it looks so much better. But if you're a YouTuber, this is super valuable. They've got this outliers.

section here where it looks for some of the outliers from just across all channels on YouTube, but you can also come up here and change this to my audience also watched, and it will show videos that are sort of in your niche that are outlier videos that do a lot better compared to that channel's normal videos. Where this tool really shines is in its ability to brainstorm. I can give it an idea of something I want to make a video around, and it will help me brainstorm and sort of flesh out that idea. Or I can just click the brainstorm button without entering anything, and it will just look at what does well on my channel and brainstorm based on what I normally make videos about. So, without putting anything in, I got AI news update, wearable tech trends, this OpenAI tech is seriously amazing, create stunning AI videos with one command, etc., etc. It knows what types of videos I post and then gives me ideas for them. Let's say I like this design: any website with one AI tool. I can click continue here, and then I can have it rephrase it. I can

have it shorten it. I can have it explode it. I can change the mood of it. I can say this but and give some additional instructions. Let's just go ahead and play with this one here, and then I can click on concept, and it will actually use AI to give me a concept of what could go into this video. Matt Wolf demonstrates how the new AI platform, AI Web Master, can design and deploy functional e-commerce websites just by describing the desired features, allowed, etc., etc. So, it gives me a whole bunch of concepts of what this video could potentially be about. Once I know the concept and I've got a good title, then I can go to thumbnail, and it will actually look at the types of thumbnails that work well on my channel and generate a bunch of thumbnail suggestions, right? It knows I put my face in most of my thumbnails. It knows I'm usually pointing at a screen or have something going on. So, based on what it knows has worked well in the past on my channel, it

generates thumbnail concepts for me. This is pretty cool if you're a YouTuber. It's an amazing way to brainstorm ideas, figure out a deeper dive of what the video will be about, come up with thumbnail ideas. It's pretty cool, and they just announced it at this VidSummit event. They're not a sponsor; I may have an affiliate link in the description 'cause they did give me an affiliate link. It's just a really cool tool, and they announced this news this week. L NOA has a new laptop. It's one of the new AI PCs, and it's got this really cool feature that I thought was awesome, where you basically take your phone and you smack it against the computer, and it pulls up all of the photos on your phone, and then you can literally just drag and drop the photos on your computer. And what's cool is it works on both iPhones and Androids. I think one of the biggest pains in the butt about owning a PC but also using an iPhone most of the time is it's kind of a pain in the butt to

get your images from your iPhone to your PC. Like, I find myself sending them to myself in Slack or something like that just to make it easy. But something like this is really cool. You just tap, tap, download the images, and it transfers really quick. I want this, so I wanted to share it because if I think this is cool and I really want it, I'm sure there's other people out there that'll think this is cool and can see how this is valuable. And finally, this week we got an intro to a new robot. We got the Neo beta from the company 1X, and here's their demo video. Here's the robot. This thing was circulating all over X with people going, Come on, that's a guy in a suit. There's no way that's really a robot. But this is legit, actually a robot. They just, you know, put this bodysuit over it so that you can't see all the mechanics and stuff. But I don't know, it's kind of creepy how realistic it looks and

then it comes back for a hug, and I think the very first post I saw about this is like, This is going to ruin marriages, or something like that. Yeah, yeah, there's a meme right there. I'm not going to explain this meme to anybody, but you know, it's a pretty interesting humanoid robot that kind of looks like a dude wearing a suit. I don't know much more about it yet, but as we learn more, I'll share about it more in future videos. And that's what I got for you this week. Sorry about the weird lighting and possibly not as good as normal sound. I am here in Dallas at VidSummit, learning about AI. AI was a huge topic at this event about how people are using AI for their creative processes on YouTube. Learning about a whole bunch of AI tools that creators can use. There was a whole bunch of sponsors here. I was learning about all their tools. Been a fun experience. I highly recommend VidSummit if you're a YouTuber, because this is one of the few events where I get to

come and talk to other YouTubers. I'm so used to talking to other AI and tech people, and CEOs, and people like that on the podcast. It's very rare that I get to talk to other YouTubers that understand the struggles, and the ups, and the downs, and the various phases of being a YouTuber. So, I really love this event, and if you're a YouTuber yourself, I highly recommend coming out to it. I've met so many amazing people, so many people that watch this channel. If you watch this channel and you saw me at VidSummit, you're one of the people that said hi. Say hi in the comments again because I love meeting people in person, and a lot of you guys I want to stay in touch with after this event. But again, that's what I got for you this week. If you like videos like this, you want to stay looped in on the latest news in the AI and emerging tech worlds, like this video and subscribe to

this channel. I'll make sure more videos like this show up in your YouTube feed. I also share tutorials on how to use a lot of these cool AI tools. I also make video breakdowns of tools for various niches, things like that. If you like AI and you want to stay on the cutting edge of the latest tech, subscribe to this channel, and I would really appreciate it, and you'll see more videos that hopefully you enjoy. Also, check out future tools. This is a site where I curate a lot of the cool AI tools that I come across. I also keep the news page up to date here. There's a lot of news that makes it to this page that doesn't actually make my videos because I don't feel it's sort of on the same level as the rest of the news in the video, but I share it all over here. And then I've also got the Future Tools AI newsletter. It's all free. I'll keep you updated directly in your inbox if you join the newsletter. You can find it all over at Future Tools. And once

again, thanks for nerding out with me. Thanks for hanging out with me today. I record these on Thursdays, so if there's any new news that came out on Friday, I'll try to make it into next week's video. And uh, yeah, appreciate you. Thanks again to Hostinger for sponsoring this video. I'll see you on the next one. Bye-bye.

AI News: Will AI Art Be Outlawed?

AI, No-Code, Tech, Futurism - I'm a tech nerd and talk about tech nerd stuff

Matt Wolfe

OpenAI의 성장과 캘리포니아의 AI 법안

OpenAI의 ChatGPT 사용량은 주간 2억 명으로 급증했으며, CEO는 100배의 성능 향상을 약속하는 새로운 모델인 GPT Next를 선보였습니다.

AI로 생성된 콘텐츠에 워터마크를 의무화하는 캘리포니아의 AB 3211 법안은 오픈소스 AI 모델과 도구에 상당한 영향을 미칠 수 있습니다.

이 법안의 엄격한 규제로 인해 기업들이 캘리포니아 주민의 접근을 차단할 수 있으며, 이미지 인증 시스템 없이 중고 카메라를 판매하는 것이 불법이 될 수도 있습니다.

새 법안에 대한 Reddit 게시물 분석

이 Reddit 게시물은 오픈소스 개발자에 대한 법안 요구 사항의 기술적 실현 가능성에 대한 타당한 우려를 제기하고 기존의 안정적인 확산 모델의 즉각적인 불법성을 과장하고 있습니다.

이 법안의 조항은 2026년 7월 1일에 발효되며, 사용자는 프로비던스 데이터 적용을 거부할 수 있습니다.

또한 Amazon의 새로운 Alexa 어시스턴트는 Claude를 사용할 예정이며, Anthropic은 50만 개의 컨텍스트 창과 코딩을 위한 GitHub 통합을 제공하는 기업용 Claude를 출시했습니다.

클라우드 기반 고객 지원 에이전트 구축

Anthropic은 클라우드 기반 고객 지원 에이전트를 구축할 수 있는 도구를 제공합니다.

com은 제휴를 통해 웹사이트 구축을 위한 AI 기능이 갖춘 저렴한 호스팅을 제공합니다.

OpenAI의 공동 설립자인 Ilia Sutskever는 연구 및 기술 개발에 중점을 둔 새로운 회사 Safe Super Intelligence를 위해 10억 달러를 모금했습니다.

구글의 새로운 기능과 AI-암호화가 겹치는 부분

Google은 사용자가 특정 이미지를 자연어로 검색할 수 있는 향상된 검색 기능을 Google 포토에 도입합니다.

또한, 구글은 AI 쇼핑 도구에 가상으로 옷을 입어볼 수 있는 기능을 도입하여 모델을 통해 의상을 선보이고 있습니다.

한편, 코인베이스는 여행 예약 및 소셜 미디어 관리와 같은 작업을 수행할 수 있도록 USDC를 사용하여 AI 에이전트 간의 첫 암호화폐 거래를 진행했습니다.

교육과 가상 세계에 미치는 AI의 영향

런던의 데이비드 게임 칼리지에서는 VR 헤드셋을 착용한 가상 교사가 학생 개개인에게 맞춤형 수업을 진행하는 등 AI가 교육에 통합되고 있습니다.

이 프로젝트에는 실제 피드백을 위한 감독 교사도 포함되어 있습니다.

한편, 프로젝트 시드에서는 마인크래프트 세계에서 1,000개가 넘는 자율 AI 에이전트를 시뮬레이션하여 자체적인 경제, 문화, 정부를 만들고 실제 업무에서 AI가 관리하고 협업할 수 있는 잠재력을 입증하고 있습니다.

미니아 맥스: 새로운 AI 동영상 생성 앱

중국의 새로운 인공지능 모델인 미니아 맥스는 소라와 비교되며 동영상 생성 기능으로 주목받고 있습니다.

com에서 제공되는 이 앱은 로그인할 필요 없이 무료로 동영상을 생성할 수 있으며, 다른 인기 동영상 생성기와 동등한 수준의 클립을 제작할 수 있습니다.

또한 Luma AI는 드림 머신에 새로운 기능을 도입하여 사용자가 카메라에 명령을 내려 동영상 움직임을 더욱 세밀하게 제어할 수 있도록 했습니다.

AI 기반 YouTube 동영상 아이디어 생성

이 도구는 YouTube의 이상 동영상을 분석하여 평소보다 실적이 더 좋은 틈새 콘텐츠에 대한 인사이트를 제공합니다.

AI를 사용하여 사용자의 채널과 과거 콘텐츠를 기반으로 콘셉트를 제안하는 등 동영상 아이디어 브레인스토밍에 탁월합니다.

또한 썸네일 제안을 생성하고 최근에는 휴대폰과 PC 간에 사진을 원활하게 전송할 수 있는 새로운 기능을 발표했습니다.

VidSummit에서 얻은 인사이트와 AI가 YouTube에 미치는 영향

연사가 휴머노이드 로봇 밈과 그것이 결혼에 미치는 잠재적 영향에 대해 이야기합니다.

댈러스에서 열린 VidSummit에서 유튜버의 창작 과정에서 AI의 역할에 초점을 맞춰 자신의 경험을 공유합니다.

연사는 동료 유튜버들에게 네트워킹을 위해 이벤트에 참석할 것을 권장하고 AI 도구와 뉴스에 대한 최신 정보를 얻을 수 있는 리소스를 제공합니다.