[00:00:04] Welcome to Tech News Briefing. It's Monday, October 5th. I'm Katie Dayton for The Wall Street Journal. Last month, Metas founder Mark Zuckerberg announced the launch of Muse, a new consumer product that he described as a personal agent that gets things done for you. The bot was designed to help regular people organize their lives, with the ability to complete tasks like buying things online and responding to emails. Muse doesn't just answer questions, it gets things done for you. I use mine to organize projects.
[00:00:34] Muse has proven to be a hit so far. The app has been downloaded millions of times. It also sent ripples across corporate America. Walmart and Sephora signed deals to become Muse retail partners, while Amazon blocked the agent from shopping on its site. But how is it working for everyday users? To answer that question, we're sitting down with Nicole Nguyen, who's been test driving the app. So, Nicole, conceptually, we know Muse is Meta's new AI-powered personal assistant.
[00:01:03] But the company has gone one step further and anthropomorphized this tech. Can you describe what Muse looks like? The Muse mascot, the name is Jolly, and it looks like a blob covered in fleece, kind of like a less evil labubu. It's like a cute, blobby, very happy to see you labubu, is what it looks like. Is it a boy? Is it a girl? Do we know anything about this mascot?
[00:01:33] It's an agent, so it's neither of those. Well, so it, in that case, sounds adorable. Your first move was to try and give Muse a new name. Can you tell me about how that went? The first thing it asks you is name your agent. And I said, I would like your name to be Mark Zuckerberg. And it said, that's not possible because of impersonation reasons. So then I ended up on the Terminator because there's been a lot of, you know, AI-might-kill-us-all talk.
[00:02:02] And I think it's good to keep the worst-case scenario in mind while you're testing out this technology. So you have spent some time experimenting with the Terminator, if that doesn't sound too weird. What was it really good at? I was impressed at the app overall. It looks like a very simple, basic chat app. But what's behind that is this highly technical, very complex virtual machine.
[00:02:30] And what is it good at is a really good question. Because I think when I opened Muse for the first time, I was like, I am not, maybe I'm not a busy enough person, a cool enough person to have an executive assistant. I have no idea, like, what I need this thing to do. So I started thinking of tasks that I needed to accomplish in my own life. Like, I live in San Francisco. The rains are coming eventually. And I bike commute with my kid to work.
[00:02:57] And there's this kind of, like, bike seat rain poncho cover situation that covers both the bike seat and my kid. There is a model I found that's sold only in Europe. And so I asked Muse, like, find me a U.S. retailer that will sell me this or a European retailer that will sell me this without, you know, customs and duties and the whole shebang. And it goes off and it does pretty extensive research across, you know, five or six retail sites.
[00:03:22] I can see it working in the browser, doing research on eBay and other places. And it basically comes to the conclusion this is not available. But there is another Italian manufacturer who makes something very similar. And it's on sale at a local bike shop in California. It's just 200 miles away and they will ship to your address. And so I clicked checkout, checked out, and then handed the task back over to the agent.
[00:03:46] Great. And it's interesting it shows what it's doing on the browser compared to a ChatGPT maybe that just presents you a bunch of results with some citations but doesn't really show you its workings. Yeah, I should make the distinction between a stock chatbot like ChatGPT that you are used to interacting with from launch several years ago. So that large language model is giving you a result based on its training data. So on like a set corpus.
[00:04:15] And there were some features that were added later on like web search. Meta is a company that has perfected the consumer app. Instagram and WhatsApp are some of the most used apps in the world. They know how to make an interface that normal people can pick up right away and like to use and will come back to using. And I think that's Muse's strength. Not that it's a new technology, but it's simplified in a way that a lot more people who aren't software engineers can find it useful.
[00:04:45] Of course, we all know by now that AI can be fallible, to say the least. Where did Muse slip up in your experience? One of the more basic questions I asked early on was, I am a member at this Pilates studio. And do I get a membership referral perk if I refer a friend? And it came back to me and it was like, great news. There is an offer that's really hard to beat. If you refer a friend, you get 50% off and they get 50% off. And this is like a pretty expensive place.
[00:05:15] I was like, count me in. I clicked on the Google form and it was for a CrossFit training gym in Virginia. And I live in San Francisco. So that was one example of it just got its wires crossed. And it was very confident that this was the studio I was referring to. And it wasn't. Did Meta say anything in response to your experiences of the chatbot messing up? Meta did not comment on those mistakes specifically.
[00:05:42] But it is clear in its language in the app that its AI can make mistakes. Coming up. Despite its shortcomings, Muse is a hit so far. But is it safe to use? Welcome back. So we've talked about what it's like to use Muse as a user. Let's turn to what Meta is hoping for from this new product.
[00:06:11] How is the company making money from this first iteration of Muse? So Muse is free to try. And I think that's part of why it's so popular. And if you hit a limit, then you can pay for more tokens, pay for more usage. You know, I've been using Muse nonstop and have never hit a limit. But I think that may change in the future because they do want to make this an experience that people are willing to pay for.
[00:06:35] That's pretty different from what Meta's business strategy has been for a lot of its other products, which has been to take user data and sell that against advertisers. It doesn't sound like Facebook will be doing that with Muse interactions. Is that correct? Meta says that your conversations with Muse are not shared with Meta's ad systems, which I think is markedly different than some of its other products.
[00:07:01] And it's also working on this encrypted, confidential virtual machine, which is an environment that you can use Muse in that is not visible to anyone except for you. If Meta wanted to access those conversations, technically they would not be able to. They've said personnel access to Muse conversations is restricted to certain circumstances.
[00:07:25] Like, for example, if there's a law enforcement request or if you, the user, has reported something went awry, then they can look into your Muse conversations and investigate. Of course, those policies aren't published. And so we don't know what each of those policies are. And so I think that if people want to try Muse, they should share sensitive information really judiciously or maybe not at all.
[00:07:53] And I wondered then, how did you feel about giving a Meta-owned tool control of things like your Gmail and your babysitting schedule? Meta doesn't have a great data use and privacy reputation. And so I was really nervous to use Muse because of that. As I did, though, I realized how long my task list was and how much Muse could actually take off from said bottomless task list.
[00:08:23] And I am someone who just doesn't feel like they have a lot of free time right now. And so I would love for Muse to do low stakes admin that would take up a lot of time out of my day. But I also feel a little creeped out by how much information I'm handing over to Muse and therefore Meta with each passing task. You can tell Muse to forget data that it's collected. You can reset Muse, which clears all of your conversations.
[00:08:53] And so I think that until that confidential virtual machine is available, where all of the information is encrypted and not visible by Meta, I think I need to cut myself off of Muse. So the big question is, will the Terminator continue to be a part of your life going forward? I think the Terminator will be back. But he has to go away for a little bit. That was our personal tech columnist, Nicole Nguyen. And that's it for Tech News Briefing.
[00:09:21] If you're a listener on Spotify, be sure to leave us a comment. Today's show was produced by Julie Chang with supervising producer Katie Ferguson. I'm Katie Dayton for The Wall Street Journal. We'll be back tomorrow with a new episode. Thanks for listening.