Anthropic's Research on AI Threats and OpenAI Updates

0:00Welcome to the podcast. There is so much happening in tech and AI news today. I'll try to get you the most interesting stuff. I've been scrolling through the comments on X, finding some funny memes, and takes on a lot of what's going on.

0:11Cognilytica Hosts:The thing I want to kick it off with, because it just happened, is that Apple has just released their brand new iPhone Duo, which is like a folding iPhone. And there's some cool AI features that are going to be in there that I think a lot of people are going to be interested in. One of the things, and they'll be rolling this out more broadly, but this is going to be with the iPhone 18, is that Apple has something called reference image features. And this is to help basically fight AI fakes. I'll get into all of that on the podcast today. OpenAI has added something called Sketch 2.5 to the new ChatGPT image 2.5.

0:43So they have a new image model that came out, and they have a cool feature called Sketch. I was testing it out today. OpenAI also claims that they have solved Navier, Stokes.

0:52Cognilytica Hosts:This is, you know, there's like a million dollar prize for this math problem. There's a lot of drama in here. I don't even want to get into all of it, but it's kind of hilarious to me. Anthropic had a researcher named Jacob Coxton who quit went viral on X this morning. I saw he had over 400 ,000 likes on his post saying he quit Anthropic because he believed that the AI model was going to kill everyone in the world or had the potential to kill all of humanity. So this, to say the least, had even more drama in it. We also have Suno that just released V.6, their model. It was all trained on licensed music because they have so many copyright suits.

1:29Cognilytica Hosts:More drama there. Guys, I'm not joking. This is just a wild day for tech. We also have a bunch of hackers that are draining Claude subscribers' token allowances because they're basically getting your password and they're just draining all of your tokens from your account. We also have Meta that has launched Muse, which is a personal AI agent built on a security first pitch. I signed up for it today and I'm not gonna lie. I'm kind of excited for Meta. I think that they're pulling out some interesting new technology and I'm excited for all of that. I also wanted to mention if you haven't already, if you'd like to get access to over 90 of the top AI models in the world, you need to go check out AIbox.ai.

2:06Cognilytica Hosts:That's my own startup. We have an MCP connector. So basically it's a connector that lets every single AI model in the world work inside of any other AI model. All of these AI companies notoriously hate each other. You have Sam Altman and Dario, at the big summit in India and everyone's holding hands and bowing because it's a big tech summit and they're another person's country and they refuse to hold hands. These people hate each other. Elon and Sam Altman, to say the least, also hate each other. So if you, they're never gonna work with each other, they're never gonna let their models work with each other, but if you wanna have all of the different AI models working together, go check out the AI Box MCP.

2:41Cognilytica Hosts:That lets, we have the brand new image model coming out of ChatGPT. You're going to be able to use that to generate images inside of Cloud. Or if you have something like 11 Labs that you use for generating audio, you can generate all of that inside of Claude. Especially for something like Claude, which is a pretty good model, but it doesn't do any of the image, audio, video generations. Or even OpenAI that's killing their Sora video model. If you want to generate videos inside of OpenAI work, you can go and get Google VO3 to generate them. So anyways, all the models work together. It's less than$9 a month to get started.

3:12Cognilytica Hosts:And we have a bunch of tiers where you can get access to a ton of really cool stuff. AI box.ai linked in the description. Okay, let's talk about Apple first, because I think this is an interesting tool that they've rolled out to kind of fight against fakes. Apple just announced something called reference images. They have kind of, this is tied to their iPhone 18 Pro launch. Basically what it's going to do is it is going to capture signed sensor data. So the moment that you take a photo, it's going to capture that. It's going to route it through private cloud compute. So Apple's compute and it's going to produce an unalterable image.

3:45Cognilytica Hosts:Apple is adopting the Synth ID standard to help users detect AI generated or AI altered images. This is something that I think OpenAI and Google have also been working with. The iPhone 18 Pro is going to have an 84 megapixel fusion main camera featuring a variable aperture built from six laser cut blades, each thinner than a human hair. So Apple, we're going to get into all the drama today. There's like Anthropics researcher says it's going to kill all of humanity. OpenAI apparently is stealing everyone's math research. And we have Apple that just said, hey, look, we made a better camera and it's going to be harder to deepfake.

4:20Cognilytica Hosts:So Apple doing pretty good right now. Reference image is also going to start inside of the photos app. There's going to be an API coming for third party developers as well. So other people are going to be able to use this as well, which is going to be interesting. So we'll keep up with that. And then I just have one meme to share if you're watching the video version of this. because we do share this, someone posted a meme of the new Apple folding, the Apple Duo, iPhone Duo with the folding screen. And basically there's a feature on it where it has the main screen that you can see. And then there's a little screen on the backside of it as well.

4:54Cognilytica Hosts:It's kind of like there's two screens on this thing. But the way selfies work with this is if you're taking a picture for someone, they can actually see themselves on the back of the phone and they could tell you to adjust how you're taking the picture. And someone said, this is going to destroy relationships it's over boyfriends are cooked nuked from orbit toasted fried basically because they're you know the boyfriend's taking a picture of the girlfriend and she's like telling him how to adjust the camera because she can see into like the selfie screen on the front and so he's adjusting it and inevitably he is still going to take the picture wrong even with her telling him how to adjust it so anyways that's pretty funny to me so opening it has now added or has launched image 2.5 this is a really impressive model i've been playing with it today one of the coolest features on here besides, okay, I'll just like the quick rundown on it is that I think they're going to start competing directly with Canva.

5:42They have these things called templates built into it. So it's like, if you generate a logo, it's like, you know, it has these pop-ups where it's like, do you want to, do you want to see what this looks like on a t-shirt, on a mug, on a hat? There's like all these different templates and all these different styles and all these different posters and things you can generate.

5:57Cognilytica Hosts:And it gives you ideas. I think they're going straight after Canva with all of this. Okay. One of the cool features they rolled out though was called Sketch. And I was playing with it today. Essentially what you do is you say just the at symbol, say at Sketch. And once that happens, there is a little draw pad that comes up. For me, I went and like kind of doodled a little guy holding a microphone in a red chair with a purple background. And I said, you know, generate image of, you know, this, I kind of explained what was in my image. And it went ahead and generated an image, which was exactly like I had, you know, requested it's perfect the thing that i like about this is sometimes when you're doing a prompt i i do i will admit i do like taking a reference image so i'll screenshot an image i like and say like make an image like this but it has that and this and i tell it like some of the some of the things that should be changed on it this is really cool because you can just doodle it if you don't have a screenshot but it's just an idea in your mind you can do that and opening i actually had some really good recommendations and examples of like when this would become really useful one thing that they showed is like if you just doodled like a picture of the garden in your backyard and or you know maybe you're like I'm trying to like redo my garden and so you do like a doodle where you they had one where it was like you doodled the pool and you kind of doodled everything that you wanted to to be inside of your garden and then it was able to generate first of all it generated like a top-down view of that specific garden so it's like look this is what your garden could look like with like a pool and your umbrellas and all your stuff.

7:27Cognilytica Hosts:And then it actually went and generated like a ultra realistic photo from like, you know, just the perspective of someone standing there. So anyways, if you had an idea and you doodled it out, it could help you with all sorts of schematics and all sorts of angles of this photo generated all sorts of ways. I love this feature. I think this is very cool. I think it's like opening eyes image generator is one of just the most sticky features. It's why I never left opening eye. I went to Claude for like six months heavy with Claude Code and with Claude Cowork. And I never was able to get rid of my subscription to OpenAI because I always needed image generation done.

8:01Cognilytica Hosts:And so I think by investing in it, adding new features and new models, they're doing an incredible job. Like no hype, OpenAI's image generator, in my opinion, is the best. I'm sorry, Google, I've never been very impressed with Nano Banana. I think there was a lot of hype around it. It'd never be OpenAI, in my opinion. So now that they have this new model, they're just kind of doubling down there and good on them, good work. Okay, Anthropic researcher named Jacob Coxton has quit and he warns that labs are, quote, gambling with our lives. So Coxton spent three years on pre-training at OpenAI and Anthropic, and then he said that people building the tech privately believe it could kill everyone by 2030.

8:40Cognilytica Hosts:So, you know, countdown, we have less than four years left to live before AI takes over. Now, I've actually seen some really interesting conspiracy theories about this particular post. The main concept for all of them is basically the same. And that's it. This is a pretty complex regulatory capture move where essentially you can, or essentially Anthropik or someone is paying to have, maybe he was sincere in his like leaving, but it was, his post was very quickly amplified. So this is basically someone that has zero followers, zero, you know, like history posting on X before today. And as soon as he posted, he immediately had a whole bunch of people that lead like AI regulatory companies retweeting him and reposting within like 20 minutes.

9:25And so this kind of helped the post blow up, get a lot of notoriety. And so a lot of people are saying that it is, you know, just people amplifying it because they want to, you know, they want to regulate AI into oblivion. But in the conspiracy theory realm, one of the things that someone said about this was before he ever made the post, he went to the Wall Street Journal and had them write, I believe, like an article or an op-ed about this. And after it launched, then he tweeted. So it's like coordinated and he's obviously working with the press. So some people believe it's more sophisticated because it's kind of a, you know, quote unquote, random.

9:57He said, this is his tweet, which was kind of a thread.

10:00Cognilytica Hosts:But the first thing said, the people building AI earnestly believe that it could kill all of us by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will coach their phrasing in the press to sound sensible. But I hear the same people express fear privately. No other human activity poses this level of threat. So, oh yeah, and I guess the main tweet he said was, I resigned from Anthropoc today. I spent the last three years building or the last three years doing pre-training research at both OpenAI and Anthropoc. Neither company is acting responsibly.

10:28Cognilytica Hosts:They are racing straight to self-improving super intelligence and gambling with our lives. More thoughts below. Anyways, then he kind of goes on the list and he's like, look, like maybe Anthropic's sort of doing something, but there's nothing that they're, you know, they don't really have a plan for recursive super intelligence, uh, self, you know, self-improving. And then everyone's just saying that Anthropic has the absolute worst PR strategy of all time, but maybe it's, maybe it's a genius. I don't know. Then Evan Hubinger, who still currently works on alignment science, he's the alignment science lead at Anthropic.

10:58Cognilytica Hosts:he responded to Jacob's post and he said Jacob is correct here with an m dash I don't know if he's using AI to to respond but I mean I would guess yes but maybe some people are mad at me saying that m dashes are normal anyways he said Jacob is correct here we really do earnestly believe AI could kill all humans exclamation point I personally think it is less than 10 within the next decade I believe anthropic is trying its best but we do not yet have a plan to solve alignment for super intelligences and are not clearly on track to like okay awesome so the guy in charge of this says yeah i think it's you know less there's a there's some percent chance that within the next decade ai could kill all humans and anthropic is doing its best but we don't actually have a problem a plan to solve it so that was a pretty wild response someone below that said is this a sponsored tweet for that artificial film and anyways yeah there's a lot of a lot of drama going on to say the least about all of this.

11:54Cognilytica Hosts:Most of the comments on Jacob's original post are some version of not really taking him seriously, if I'm being honest. And then there's a lot of people that just are sort of AI doomers that are hyping it up. Anyways, a lot of people are screenshotting, I don't know if you remember, but there was a moment where back in 2022, a guy named Blake Lehman said that system has perception of the ability to express thoughts and feelings equivalent to a human child. It's a Google engineer was put on leave after saying AI chatbot had become sentient. So anyways, if you guys all remember that, that was kind of a wild moment from 2022.

12:30Feels like it was longer ago, but some people are saying

12:33Cognilytica Hosts:that kind of reminds them of that guy. And some people are saying Terminator wasn't a movie. It was a prophetic documentary. Honestly, this is pretty overhyped, but he's at 620 ,000 likes right now. So perhaps we are all on track to die in the next couple years. Pretty wild claim. Okay, Suno is launching their V6 model. I love Suno. I've used it a ton. It's a lot of fun. And I've released a few albums with it, some of which make me, you know, hundreds of dollars a month from AI-generated music. My most famous one that I always tell people about is I have relaxing piano music to help your cat fall asleep.

13:12That album makes me$600 a month, and it's just AI-generated. It's actually like AI-generated.

13:17Cognilytica Hosts:It's not slop. It's just relaxing piano music. It sounds great. It's funny how I marketed it. It did really get well, and a lot of people just added it to their playlist to play to their cat. So cats apparently love AI-generated music, and I'm not even apologizing because, you know, if they love it, they love it. Now, one thing that I'll say is the problem with that music and everything else, which Suno takes the full liability for, so it's not my problem, but Suno trained all of their previous models on licensed music. They just trained it on all the music that existed in the world. And all of the record labels sued them.

13:48Cognilytica Hosts:You had Warner, BMG, Believe, Sony, and Universal. And so they decided to train their brand new model. And by the way, here's like my very contrarian take that I think also a lot of people would be mad about. Because this goes for Suna and basically everyone else. They trained it on everything. They got in trouble. They made licensing deals with the distributors. But in the meantime, their company and brand grew a ton. Maybe they wouldn't have had enough money or wouldn't have been able to get permission before. but the brand grew a ton and now they were able to pay right because they've raised over 819 million they're able to pay off all of these fines they're able to pay for licensing and now they're able to train a model with actual licensed music now there's a lot of different takes on this i've been doing i've been using ai music generators for like 10 plus years i think at this point yeah about 10 years and so way before chat gpt ever came out there's been a bunch of music generator models.

14:42Cognilytica Hosts:I don't know what those early ones were using. They weren't that great, but some of them were okay. They weren't horrible. It was pretty fun back in the day. Anyways, with this new model, it's all licensed data. I think that's fantastic. This is what we've seen everyone else do. You get a model, you get it big, you get some revenue, you prove the concept, and then you go pay everybody out. Should you have paid everyone out ahead of time? Maybe, but if the model wasn't proven, it's kind of hard. They wouldn't have had the money to do it. So anyways, that's another argument. This release is going to happen in three tiers.

15:09So they have v6, which is paid.

15:10Cognilytica Hosts:They have v6 wild which is paid and experimental and then they have v6 mini which is free all of the older suno models are going to be retired and the reason is because the data licensing issue suno has raised all of that money and they you know obviously have all these lawsuits going on this launch is on the same day or just one day after suno admitted to training earlier models on youtube videos so beyond just all the music they actually went to youtube videos which is a huge no-no but to be fair, OpenAI never admitted to train Sora, their video model on YouTube. Although I kind of wonder if that might maybe partially why they shut it down.

15:46Cognilytica Hosts:But Miriam Maradi, when she was questioned by the New York Times and asked point blank, did you train the OpenAI Sora video model on YouTube? She kind of just like shied away from the topic. And it was like one of those famously awkward moments where she didn't want to, you know, didn't want to answer the question basically. So we get that OpenAI probably also did that. Okay. Hackers are draining Claude subscriber token allowances with stolen sessions. Anthropic confirmed that InfoSealer malware is hijacking Claude code OAuth tokens. One max user watched usage climb from 45 % to 55 % while doing no work.

16:20And so he's like, what's going on? Why are my tokens disappearing? People are hacking in and just using tokens to do certain projects.

16:27Cognilytica Hosts:Anthropic confirmed this. And there is a separate users that reported their usage jumping from zero to 49 % in 12 minutes after only a couple of prompts and a web search. Anthropics suspended the affected accounts. They invalidated all the sessions and they issued a partial refund. One user received 44 pounds for their$200 a month subscription. Account support tracks total usage, but not itemized usage, meaning token theft could run undetected for months, right? It honestly reminds me of all of the people that were doing like music body and farms where they would steal a whole bunch of people's Spotify passwords.

17:03They would just get them on the dark web because they're usually, you could get a, you know, whatever. So they'd steal them from the dark web. And they would log into a whole bunch of random people's Spotify accounts randomly and just play their own song over and over again. And then they would log out. And it was really hard to track. And Spotify spent a lot of time and money and eventually kind of cracked down on the whole scheme. And I think a lot of people doing it illegally might have went to prison or some, I don't know, different things, depending on how nefarious all of the body and kind of fraud went.

17:31But this reminds me a lot of that. It's the same idea. You get maybe combo lists from the dark web. You steal a whole bunch of people's account tokens and passwords. You log in, just get a couple things done. Or maybe if they have an API, you log in and use a few tokens, and then you log out, and then maybe they'll hardly notice. Now, my last complaint of the day in regards to all of this, which has absolutely nothing to do with that because this was completely legal, no one hacked me, is that today I asked ChatGPT Astra, so their you know gpt6 model i put on the absolute highest usage tier right the highest thinking capability and i asked it to go and review my website and give me some good ideas for some good ideas to you know review and make the branding and the visuals on my website better and over the course of two minutes that it worked for it burnt through 90 or i think actually 80 percent of all my tokens for the entire week so do not touch astra six on the absolute ultra max tier with a nine with a six foot pull that's my advice to the wise if you're worried about your token usage unless you're at the end of the week and you got to use it quick before it resets or you have one really big problem i don't know like i just i'm so blown away that in two minutes it could use 80 of my tokens i'm super annoyed by that i'm in my chat gpt account and they do have

18:53Cognilytica Hosts:a very cool feature, which huge shout out to ChatGPT for adding. And by the way, Claude's got the same problem with Claude though. You basically get locked out of your account with ChatGPT. They say you are now using Luma, a faster model for simpler tasks. Use your reset or buy a reset for$80 to continue using the most advanced models or wait for it to reset after. And then it has my reset date. So basically they switch you onto like one of their most, I don't know, basic models. You can get basic tasks done. But personally, I'm actually very happy about that because at least they're not completely shutting off my account.

19:26Just, yeah, switch me to one of the terrible models and it can sort of do some stuff for me if I'm in a pinch. I actually have a second cloud account that I'll probably be using for a lot of stuff in the meantime. But overall, yeah, good pros and cons to that. Okay, really quickly wrapping up, we got to talk about what meta has launched today. This is Muse.

19:45Cognilytica Hosts:It's a personal AI agent built on a security first pitch. The Super Intelligence Lab product ships on iOS and Android and WhatsApp. It has Stripe checkout and a$300 ,000 bug bounty ceiling. So if you go find bugs, they're actually paying you for it, which I think is genius on their part. The paid tier is for people that are paying$20 a month, and they have a$100 a month tier and a free tier. And of course, all of those is usage-based. They said that you get about$130 ,000 for a single user prompt injections. So like they're just paying you to red team this thing. If you can use it, get it to do something bad, they'll pay you$130 ,000.

20:25Cognilytica Hosts:That's a pretty awesome deal in my opinion. Checkouts, and the reason why they're doing this is because they're adding payments into it. They partnered with Stripe. They have checkouts that run through Stripe's link, single use card numbers, 1Password and ShopPay integrations are all underway. They're working on all that. So what all this means is that you can go to Muse and say, hey, I need to go, you know, I want to eat Taco Bell for lunch. Go and order it for me. And it has access to your Stripe account. It will go get everything ready. It'll have a pop-up that's like, do you confirm spending$14 on this burrito and drink from Taco Bell?

20:56Cognilytica Hosts:You click yes, and it gets shipped straight to your door with DoorDash. I personally love this. I mean, it's a big bet. I know it's a gamble. I know it's risky. I mean, OpenAI and Anthropic are not doing this direct Stripe integration where it can buy stuff for you. But I think number one, it's smart for Stripe because they're gonna get some good volume here. They're gonna prove this concept. It's almost like there's insurance because OpenAI is like offering$130 ,000 for prompt injection. So they're, you know, I mean, that's a massive bounty program. They'll pay you a lot of money if you can hack this thing.