Apple & Meta Bring Muse to Mac, Anthropic Opens Up & Deezer Filters AI Music

Welcome to the podcast. According to an article from NBC News, cracks are beginning to show in the stock market as doubts grow about AI investment. So there's a lot of talk right now about the AI build out and the economic repercussions of it. And if this is going to pan out, if the ROI is there, as we have two mega IPOs, OpenAI and Anthropic coming up, we're going to dive into all of that. Also, we have Anthropic that is letting METR, famously Jacob Coxton, who was over at, you know, Anthropic for a couple of weeks. And then he did his big whistleblower thing where he went and said that it wasn't safe.

And, you know, over Anthropic, they believe it's going to end humanity. He did his big PR stunt and went to all the big news organizations. And after that, he immediately was hired by METR and Anthropic. immediately handed them a multi million dollar contract. And now Meta researchers are going to be able to look under the hood with employee access at Anthropic. So there's some interesting things happening there. According to CNBC, Apple and Meta are teaming up to make Muse AI models run on Mac computers. I actually got an email from Meta about this that I'm super excited about as far as they're, you know, they're still committed to open source AI models. Deezer is going to let you filter out AI generated music

soon. And if I'm being honest, I actually think this is a phenomenal thing. It's funny for people that don't know, I actually make probably 20 million streams a year on Spotify. And a big chunk of that is from AI generated content that I've produced over the years. And then a bunch of it is with my own voice. So it's a little known fact about me. I don't even publish who the artist is, but I make across a handful of artists that I've published. I think about 20 million streams every single year. But I still think this is a phenomenal feature, and I'll tell you why. And it's not just Deezer, by the way. YouTube Music is also launching a new label that will show you if something's AI -generated

music because there is so much of it on the platform right now. So I will tell you all the pros and cons to all of this. APAT is sending AI agents undercover to get scammed so they can gather fraud intelligence. This is super fascinating to me. And then we have a meta that is launching an open -weight AI model that can run on your Mac, sort of like what I was mentioning earlier.

We also have Amazon's device chief, Panos Pene, who says there is a future where AI devices replace the need to use apps. I want to give my take on all of that because I don't think everything he said is accurate. So we're getting into all of that. Before we do, if you want to get access to over 100 different AI models for just one subscription, so Anthropic, OpenAI, Google Gemini, Grok, all of the image, audio, video, 11 labs, all of that in one platform for $8 .99 a month.

Go check out AIbox .ai. That's my own startup. We've done some massive upgrades last week to it. As far as one of the big things is we have a groups feature where you can make a group. I make a group of four different AI models. So for me, it's Gemini, Claude, ChatGPT, and Grok. I ask one question and all of them at the same time are computing and they give you the same answer. all back at the same time. For me, this is super useful because some models are better at other things. And it's really cool when you do a text. But it's also really cool when you use something like image generation. Sometimes I like open eyes image generation, but sometimes it has this like brown look to it that you might

be familiar with. And anyways, there's a bunch of different image generators, you can put all of them into a group, ask a question, and all of them will generate an image, you can pick your favorite one. So if you want to speed up how fast it and the time and the quality on the image, audio, text, or even video outputs that you get, go check out the new groups feature on AIbox .ai, along with tons of other incredible cool features. You can build tools with it. There's so much more inside of it, but go check it out, AIbox .ai. So the first thing I want to talk about is the... the market, the AI investment market. People are saying there's doubts about the ROI that's coming just one week, according

to NBC News, just one week after hitting fresh all time highs. U .S. stocks are showing signs of strain. And so, you know, this is normal. The stock market goes up and down. But today we are pegging this on. AI. They said that investors are starting to worry about returns on trillions of dollars that have put into AI. This is interesting.

I think after trading closed on Thursday, the S &P 500 was off 1 % from its week ago record. If you look at the tech heavy NASDAQ, that was down more than 3 % from its peak. Okay. The stock market goes up and down. We know this, but I do think this is an interesting moment. And I actually don't think this is unhealthy. Sometimes I think that we get... We get a lot of hype in AI and in technology where we can be very bullish because, oh my gosh, all these incredible things came out and the whole market rises with them because we look at the implications, right? Like OpenAI and Anthropic are surging if their usage metrics and if the revenue of those companies are surging, that means there's a lot of downstream

winners as well. There's, you know, NVIDIA, who's going to be selling more GPUs. There's a lot of compute companies. There's a lot of data center companies. There's companies that build steel that is sent to data centers. There's companies, there's like so many different. people down the supply chain. So anyways, we have these big hype, these big hype moments that do lift the whole stock market. And I think it's okay every once in a while to have these resets where, you know, we're obviously growing if you look at the last year at an insane rate. But I think it's healthy to say like, look, maybe this thing isn't going to. replace everyone in the world next week.

And I think we see some of these drawbacks in stock. So I don't think this is a huge cause for concern or alarm right now. But I do think it's healthy when we see some of these drawbacks. Okay, the big story that I was really excited about this week, or interested in, I guess you could say, is that Anthropic is letting METR researchers look under the hood. They're giving them employee level access to all of the latest models that are coming out. This is I think this is actually an interesting and exciting precedent.

METR, I mean, I'm not, I don't have anything, any sort of special allegiance to that particular group. But I do think it's kind of cool when we have some of these open source or we have some of these AI model companies allowing independent researchers, independent security checkers to come in and look at this. And if I'm being honest, I don't think this is actually that much different than what a lot of these firms, OpenAI has been doing red teaming with third party vendors for a long time. And so anyways, it just feels like anthropic. did their big, you know, spiel where they were like, look, there's a 10 % chance that Claude is going to end all of humanity and it's super dangerous, yada, yada. And then this is

kind of their response. But look, we're letting safety researchers come in and check our models before they go public. I don't think there's anything wrong with that. I don't like the way that they did the PR on all of that. At the end of the day, their model is really great. It is very capable. Letting third party people come and independently verify it. fantastic. And I love this way more than, you know, let's say enforcing some sort of regulatory body in the government where they might not even be competent enough to do some of this research. So I think this this approach is fantastic. We had a recent AI safety accord that a lot of these big firms signed and Anthropic was there. I think Mark

Zuckerberg had a phenomenal point where he was like, look, as far as like getting regulated, like as companies, we're beholden to our shareholders, we're beholden to our board for making some sort of software service that is unsafe. We are liable for that. So third party auditors should absolutely come and look at our stuff and then they should report to the board. And if it's something that's unsafe, then the board decides to push the product forward anyways, then they're liable for it.

And we as a company are liable for it. This is all just kind of like taking care of in regular laws and liability that we as companies take on. So I really love that framing from Zuckerberg on all of this. And I think that it basically will replace the need for heavy handed. government regulation that could slow down the entire field so anthropic letting them in i think this is great in a public statement they said in their blog post they said we want to be publicly accountable for the decisions we're making. We aim to close the gap between what evaluators understand about a frontier AI company systems and what the company itself understands. So under all of this, METR has Anthropic issued company laptops. They have

office space. They have access to internal tools that they said are basically exactly what Anthropic's own risk assessment teams use. And then they said Anthropic employees are encouraged to share information with them just as they would with a fellow employee working on alignment risk. This is cool because I think this is the next level of transparency where they're like, look, there's a third party company, you know, evaluating us. And I think they're putting their money where their mouth is. They're like, our employees can go talk to them about concerns that they might have just like anyone else. And I think having this level of openness and knowing that there's this third party company inside your organization

that's going to, you know, make a PR statement if there's a big issue. We'll just we'll encourage everyone to be very transparent about the safety risks and the guardrails and what they're doing to work on this. I don't think this is bad. You know, they've opted into doing this. And if this may gain some trust with users and with corporations that use their company, I think that's a phenomenal outcome. And so I think that that's that's the best direction. And again, I prefer this much more than heavy handed government bureaucracy determining a lot of these things to regulation.

Apple and Meta are teaming up. This is kind of funny. If you know much about Apple and Meta, they've had a lot of beef. Tim Cook. Geesh, why do I keep wanting to call him Tim Apple, man? X has really warped my brain. Tim Cook and Mark Zuckerberg have had a lot of beef over the years. But and like very famously, you know, a lot of their privacy things, which stopped Facebook being allowed to track you. And then, you know, Facebook lost billions of dollars from their ad network and all sorts of stuff. And there is whatever, all sorts of legal campaigns around this. And probably I would say like. you know, like different campaigns where Meta was like telling small business owners to contact them

and whatever, whatever. So, you know, voice their displeasure with them. So I think it's all interesting. What I will say is after all of that happens, it looks like they are coming together and working together again to help Meta be able to run their Muse AI model on Mac computers. So this is a collaboration between the two of them. We have a, you know, we have new CEO, new leadership over at Meta. apple and it seems like we're now starting to work together what's interesting with this particular collaboration is that it's going to be the meta spark which is honestly this is what's been running their muse model or sorry the muse spark and so i think that this is a this is a great model being able to run

this on your own computer what they've said about this is that you're going to be actually able to run this on two different, I think, Mac Studio PCs, and they're going to both be able to run it. You can run MuseSpark code. You can analyze video. You can do agentic work. It's all going to run locally. It's not in the cloud. So two big perks is that there's more security, and it's also going to save you money. If you can run one of these models, Meta announced, which I kind of mentioned earlier, but Meta has a new open -weight AI model. that they've announced.

That's going to be able to run directly on your MacBook. You don't have to pay Meta. And this is kind of crazy because it's at the time when Meta is really hitting the frontier. Their new Meta wonderful model is, you know, just as good as everything from Anthropic and from OpenAI, allegedly, according to a lot of different benchmarks.

And with that, or, you know, you could say it's 10 % behind, but it's right up there. So this big model is right up there. It'd be like basically if OpenAI was, you know, they have their three different tiers of models or four different tiers of models. It'd be like if they're saying, look, our lowest tier of model, which you can pay us for, is just free to run it on your computer.

So that's what Meta is doing. I'm really impressed with them. It might not even be the lowest tier for Anthropic or for OpenAI. It might be slightly higher than that. So I think this is very... I don't know, charitable might be a good way to say this. They're trying to win trust. They're trying to win users in the best way meta knows how, which is to give away free stuff. And honestly, I'm all here for it. I love the Muse tool. I get it to triage my emails every single day.

It writes draft emails for me and it's just very useful and it's free. So I think that they've made some great products there. Deezer is going to let you filter out AI generated music. If I'm being honest, I actually really like this. Even as someone that makes thousands of dollars a month from AI generated music, I've made this over the years. And by the way, like AI generated music, I started making like probably like seven years ago. So AI music generators are not anything incredibly new. And it's also, there's a weird line to draw because We have two different things.

We have something like Suno or Udio where you give it prompt and it just one shots an entire song with the vocalist and all that. But then we have a lot of tools that have existed for a long time inside of Premiere where you're like adding AI effects on top of your music or you're using quote unquote AI to generate like a piano or to. Or even just like the computers generating a piano or like... Anyway, so it's hard to draw the line on what's fully AI generated versus not. I tend to think just one -shotted songs, right? A song about being in the mountains with rustic sound to it and then Suno spits out a song. I don't tend to find those to be super high quality. And recently I've been listening

to some music and the recommended songs that come after a song that I like and I clicked on is like AI generated and it's kind of annoying and I skip past them. So I appreciate the ability to filter them out if you want, to add a label onto them if you want. We've had disclosures on a lot of the streaming distribution platforms like Deezer or DistroKid or Lander. These are the places where you release your music. We've had disclosures there where you say if you use AI and all of that's in the metadata, so now they're going to be able to filter it. I don't think there's anything wrong with that. What I will say is the next step that... YouTube music is doing. And you know, these are YouTube music,

I could imagine Apple Music and Spotify will do similar things soon in the future. But YouTube music is launching a new thing to label stuff is AI generated music. Again, there's like tricky line of like, how much of it was AI generated, how much of it wasn't, I've released a ton of content where it's my own voice, but like some of the instruments in the background are like AI generated. I'm not an incredible musician, I don't play a lot of instruments. And that's kind of what I've done. So anyways, I think that there's a lot of options. There's a lot of things you can do there. This is good. It's giving people more options. And I think that that is phenomenal.

Guys, thank you so much for tuning into the podcast today. If you enjoyed this episode, make sure to leave a rating and review wherever you get your podcasts and go check out AIbox .ai. And I will see you all in the next episode.

Apple & Meta Bring Muse to Mac, Anthropic Opens Up & Deezer Filters AI Music

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin