Tech Tuesday: Muse AI Agent

0:00It's Tech Tuesday with ABC News Technology reporter Mike Devosky. We're not sure if he will be a judge in the upcoming AI Beauty Pageant, but he is our guest every Tuesday on Tech Tuesday. The latest and greatest in the world of tech. Test was next product. They call it a Robo Taxi. These are AR glasses. They're called simply the frame. On Wisconsin's Midday News. It's Tuesday. That means Tech Tuesday and our weekly check in with ABC Tech expert Mike Devosky. Hey guys, how are you? Doing great here. Let's talk all things meta here and a shiny new pull the wraps off AI agent called Muse AI. What does Mark Zuckerberg developed here? What is it supposed to do? Yeah, so a few weeks ago Muse AI came out. This is an AI agent from meta. We've heard a lot about this type of technology recently. These AI agents are these programs that can execute on tasks for you. Not just answer questions. So if Chatchy PT is a chatbot, you type out a question.

1:01It gives you an answer. An agent like Muse can go one step further. It can go put something on your calendar. It can go organize your photo albums. It can do some online shopping for you. Well, it has been pretty popular over the last several weeks, regularly topping app store charts. But you can see why this was such a phenomenon for meta. Meta owns huge platforms. Facebook and Instagram used by literally billions of people worldwide. This is in front of a lot of eyeballs. And while open AI and anthropic and others all have similar agentic technology out there, they have to do a lot more work to get it in front of people. Whereas Meta basically has this sort of baked in strategy, this playbook, to get people to try new technologies that they're working on. It seems to have worked at least to some degree with Muse. However, it hasn't been without speed bumps, guys. Meta, as we said, marketed this as something that could potentially do some online shopping for you.

2:02Well, that has now spiraled into a growing standoff between Meta and Amazon. That company over the weekend started blocking Muse AI from browsing on Amazon, seemingly because Amazon never agreed to allow Meta's AI to use the platform in the first place. I actually tried to get Muse AI to go buy some toothpaste off of Amazon just as like a test yesterday. And the way it works is it basically calls up its own little digital desktop that you can see it kind of clicking around and trying to solve the problem that you're giving it. And once it went to Amazon, it kind of gave back this firewall with Amazon's website reading, continued access by an unauthorized AI agent, violates Amazon's conditions of use to which our customers have agreed. Basically, they say that Muse isn't identifying itself as a bot, and that's a violation of their terms of service. Well, that's interesting. It seems like Muse can be really useful, but then at some point it might start to get risky to how much access does Muse need to your personal information, like your email

3:02or your calendar, or your payments to actually buy that toothpaste if you were able to do so. Yeah, so there's sort of two levels of it, right? You can bake in access to things like your Google calendar, which I did yesterday, and I gave it a task. I said, hey, you know, put these items on my calendar, find some time for them in the month of October. And I was able to do that and actually pretty adeptly, I did it on my phone. I was able to watch it actually working on my desktop, almost like a ghost was controlling the computer. It's kind of eerie in some ways. But that is like baked in integration. Google has this partnership with Meta to allow that to happen. If you don't have that partnership, if you're a platform like Amazon or GrubHub or others, Muse can still use it, but it's basically using it like a person would use it. So it's literally going to its digital desktop, calling up a browser, going to Amazon.com, and trying to log in as you and execute on that task. That does require you to hand over some credentials. Meta has had a string of security and privacy breaches over the course of its history, which

4:08should give some people pause about how much you trust this company to handle that data properly, especially when you consider that AI technology adds this extra complexity and is still pretty imperfect, right? Still hallucinate, still goes wrong, still does all sorts of weird stuff unexpectedly. And that creates this extra threat when you're handing over potentially sensitive information. The other big thing to think about here, guys, is that as Amazon blocked Muse AI, it cited this violation of terms of service. Well, the whole reason that Amazon doesn't want these AI bots clicking around on its platform is because it could potentially lose a real revenue stream if these agents become more prominent. Amazon makes its money, yes, by taking a cut of every purchase that happens on that platform. But they also make money on things like sponsored listings. It's a big revenue stream for them.

5:08Advertising on Amazon is huge, right? You see a banner ad, you see different ads across the platform. Amazon makes money on that. If a bot is clicking around the platform, Amazon doesn't make money. Bots don't care about sponsored listings. They are trying to find whatever it is that you gave them the task of going to find. And a human is not looking at those ads. That could potentially reshape how online business is done, not just at Amazon, but on other platforms that sell services to the general consumer. Things like Uber, which also relies on similar business strategies. Grubhub, DoorDash, ZockDock, the potential of AI agents is that they add convenience. But if they remove the human from the process of going out and ordering those services or ordering those items, well, then the business on the back end could potentially be disrupted. And that's kind of the bigger point here, guys. We've spent the last couple of weeks talking about the far-reaching existential threats of artificial intelligence. And whether it's all going to kill us or whatever it happens to be, this is a real concern

6:12that's in front of us right now that's playing out that is only going to grow as these technologies become more prominent. It's kind of a familiar story with AI, right? Here's this shiny new tool. If used properly in an ideal world, here's how it will help your life. And then all of a sudden, you start to learn about data concerns, privacy concerns, maybe I'm enrolled and didn't even know it. And then before you know it, maybe two weeks from now, we'll have a story about how this thing went rogue. And all of a sudden, Mike Debusky bought three Volkswagen beetles that were no actually stuff. Right? I mean, isn't that how this all works? Yeah, potentially. I mean, again, I think that the kind of big takeaway is that, you know, even in an ideal world, Amazon dies here, right? In an ideal world where it's just agents doing this work, you know, there's no business strategy there for Amazon anymore. There's no business strategy for any of these other companies that, that, you know, employ this business. The architecture of the business of the web relies on people looking at advertisements and clicking on sponsored listings in large part. That is, you know, even a threat if AI works 100% of the time, which it doesn't.

7:16So, you know, again, it kind of just goes to just how messy this situation is right now as we continue to reckon with the role of artificial intelligence in our lives. That for its part has regularly found itself kind of behind in the AI race and has been pitching Muse AI as like a way to catch up. Let's not forget that open AI and anthropic and Google and the leaders in this industry, they all have AI agents as well. This is nothing new, but they have to do a lot more work to get people to use it, right? You have to go out of your way and download a different app from open AI or whatever it happens to be. Maybe even subscribe to get the privilege to use it. That AI is free to everybody. And as we said, billions of people use Facebook and Instagram. This is just in front of a lot more eyeballs than it used to be. It's kind of like threads in many ways. If you remember that platform, it was pretty popular there for a couple years, although it has sort of waned in influence, certainly. But in large part, it was popular because if you had an Instagram account, you were like 80% of the way to having a threads account.

8:16And why not? It's free. This AI is kind of running the same playbook here, although with all the added concerns that AI brings with it. It's Tech Tuesday, Mike. Can we look ahead to to Meta Connect? What are the big announcements or developments that you're looking forward to? Yes. So this is Meta's annual tech conference. It's basically like an Apple event they show off new hardware and new software. In the hardware space when we're talking about Meta, we are generally talking about smart glasses, which they've run a pretty robust business on. They have this big partnership with Elselor Luxotica, which makes Ray Bans and Oakles and other popular brands. They integrate Meta's technology into those glasses so you can kind of chat with the various AI's that they have. You can take pictures. You can record video. You can listen to music through these glasses. We might see new generations of the Ray Bans smart glasses. We might see a new generation of the Ray Bans display, which is the version of this technology that actually has small screens integrated or projected onto the lenses. They're pretty cool, interesting, new way to think about how you use your phone or you

9:18interact with technology out in the world. They've also been rumored to be working on smart glasses that don't have cameras in them. This has been the point of a lot of controversy recently, guys, with people using these glasses out in the world to record people without their knowledge. Some critics have even gone so far as to call them pervert glasses. In a post on Instagram recently, Meta CTO Andy Bosworth said the company will have more to say about all of that at this conference. That's something to look out for. Couple other just assorted items here. The company also said that the Muse AI agent that we've been talking about is going to be available to talk to through the glasses as well. So that could be announced here. The Long Roomer Phoenix project, which is a mixed reality headset that they've been working on. That could get its wraps pulled off tomorrow at this event. And a kind of weird one here, guys. C.E.O. Mark Zuckerberg has reportedly created an AI version of himself that can interact with Meta employees, kind of like an internal tool. Go talk to the C.E.O. but the AI version that maybe will show up tomorrow at Meta Connect.

10:21That's fascinating. We'll just have to wait and see. Ooh, could it potentially be Mark Zuckerberg on one shoulder, then like an evil Mark Zuckerberg on the other shoulder? There's our own Mark Zuckerberg. Right. Right. Which one are you going to do? Right? It's a famous loony tunes cartoon. I'm cleared. 2026. For maybe C.E.S. Mike Debusky at A Tech Tunes day, we appreciate it, Mike. Thank you. No problem, guys. Take care.