The Rise of (Cute) AI Agents | Fortune Daily

0:00Andrew Nusca: Meta and OpenAI want more access to your data, but don't worry, it's cute and fuzzy. Hello, I'm Kristen Stoller, in for Ellie Austin, who is on the road today. The rise of the personal AI agent is upon us. Your to-do list can get tackled by an agent, but at what cost? Meta's new app Muse books your appointments and does your shopping, and OpenAI answered with its own agent, Dots. Now, invasive things come in cute packaging. These little helpers need access to your email, calendar, and credit card to function. Now, enormous troves of data and dollars are at stake, and it's only the beginning. Senior AI reporter Emily Forlini joins us with the full story.

0:42Andrew Nusca: From New York, this is Fortune Daily.

0:50Andrew Nusca: Emily, thank you so much for joining me. Of course. So it seems like personal AI agents are here, and they are everywhere. They are flooding the New York City subway ads. They are at the top of the app store and they are blocked in some countries. But talk to me about them. Can you explain what a personal agent is and what it can offer people?

1:08Emily Forlini: Yes, this is a huge new wave of AI. And even though everybody's not using them yet, the thought is a lot of us will be. So it's a shift from chatbots and generating text and images to actually performing actions for you. So Meta launched its Muse Assistant, I guess you can call it. They're very cute little animated humanoid little things on your desktop. You name them. OpenAI has dots, which is very similar. So basically, they can rove around your computer for you, browse the web. They can compose emails for you. They can even make purchases. So it's performing actions rather than just that typical chat interface of ChatGPT.

1:48Andrew Nusca: Okay, so very different than what a ChatGPT can do for you. That is so interesting. So I think what is the most interesting to me about this is that this requires a lot of trust from consumers. There's a lot of really personal data you have to give to these companies, your credit card, your logins. Why do you think people are so willing to give this over, especially when companies like Meta, OpenAI have faced a lot of privacy concerns?

2:15Emily Forlini: Yes, both Meta and OpenAI have big consumer trust issues. I mean, meta going back for years and OpenAI more recently with all of the hacking incidents. So you're right. It's a contradiction. We talk so much about how we can't trust these systems and then people are quietly adopting it and really excited about it. So I think it's just productivity. It's also kind of an early adopter set. People want to see what this new tech can do. And so it's an ongoing experiment. But the people so far who are using them say that they're just smashing work for them. Like it's getting through their whole to-do list.

2:49Emily Forlini: I tried to download one, but I was afraid because of the data thing. And so I didn't do it on my personal laptop. Okay. I tried to do it on our work laptops because I was like... I bet IT absolutely loved that, by the way. I got blocked from everything. So yeah, I'm in the camp where I'm hesitant as well. Interesting. Okay.

3:10Andrew Nusca: So would you use it in the future, you think? If we got IT to unblock it, how would you want to use it?

3:15Emily Forlini: Well, it's hard because the biggest thing I have is like we're journalists. So I just talk to so many people per day and sometimes it's overwhelming. I get texts on regular text, signal, WhatsApp. I have emails. I have calls. And I would love a personal assistant to basically streamline all that, make sure I respond to everybody in a timely manner. You can figure out my calendar, book meetings, like share open time slots, all that stuff. But that would require giving email access. And then it's like, what if it sends an email I don't like? or what if it, you know what I mean? It's tricky. So I haven't decided what I think.

3:49Andrew Nusca: Is there a safeguard if it does something on your behalf and, you know, spends X amount of money, buys a sweater or sends an email you don't like? Who's responsible? What happens?

4:00Emily Forlini: You are responsible for sure. So I have heard people saying, you know, I'll ask it to respond to my emails but not send them. So to kind of draft something and serve it up to me and you can kind of approve or edit and things like that. But there have been some fails, like people, the chat bot, or I guess we'll call it a Muse bot. What do we call it? I don't know, an assistant? I like it. Or a dot. Their Muse or their dot, whatever they've named it, will purchase something and they say, oh, no, actually, I don't want that. And they say, oh, well, actually, too late. It's already in shipping status.

4:33Emily Forlini: You know, you have to deal with that. So that's a big headache.

4:36Andrew Nusca: Do we know what Muse and Meta and OpenAI and Dots are doing with all of this personal data that it's collecting?

4:44Emily Forlini: They're using it to make these decisions. So in the world of AI, the term has shifted a little bit from data to context. So context is like data on steroids. So it's not just maybe what you're putting in your email or entering on a web page or the cookies that are collected. It's everything. So you give these assistants access to your PC. So it's like what files are on your desktop, what tab you were just browsing, and how that might be influencing what you're emailing about. In a future world, it's, okay, maybe you're talking about something on the computer, but then your mom calls. And the way you talk to your mom about that subject is very different from what you're doing on the computer.

5:24Emily Forlini: So that might influence your decision. So the AI wants to know absolutely everything that is happening to help make smart decisions that you would make yourself with all of that context. So that's kind of the new word. It's different than data in the sense that it is so all-encompassing.

5:41Andrew Nusca: And what is the difference between Dots and Muse? Do they do the same thing? Are there any differences there?

5:49Emily Forlini: So they appear very similar. Like I said, they're kind of cutesy, human-like, anthropomorphic. I made a dot with OpenAI. Okay, what does yours look like? You have to share with us. It's super weird. It's like an alien-looking thing with sunglasses, and I named it Tina. I don't know. I felt inspired. But Muse, you can also name it and pick the shape and stuff. But the big difference is that Muse has a free tier, so a lot of people are using it. So it's kind of more consumer-oriented in that way. And Dots from OpenAI, you have to have at least$100 per month subscription. And what does that get you, the subscription?

6:27Emily Forlini: More usage, access to things like new features like dots. So I do not pay for that subscription. I did get it for free from OpenAI, full disclosure. I would not pay$100 a month for an AI at this point. But yeah, so it's kind of geared towards different audiences there. But they both have very similar promises of performing smart actions on your behalf.

6:47Andrew Nusca: There's clearly some kind of like marketing play here by making them so cute, so fuzzy, so alien-like like yours. What do you think this design choice says about how they want users to engage?

6:59Emily Forlini: Yeah, I mean, it kind of goes back to Microsoft Clippy. It's very similar. It's like a more powerful version of Clippy, a much more powerful version. So I think it's trying to make people comfortable with all of the hacking and the concerns about AI. People can develop relationships with these assistants. and ultimately that fuels more usage, more comfort, inspires people, I think, to give over more of that context and that data. So they see it almost as a coworker or a person. And then my theory on why OpenAI called it dots is because I think they're going to have multiple dots and then they're going to connect the dots and it's going to be a ring of assistants.

7:40Emily Forlini: So you can maybe refer to them by their names and it's a way to differentiate.

7:44Andrew Nusca: So Emily, I'm curious because I feel like weeks ago We were at the point of like, is AI going to cure cancer or is it going to kill us all? But now I think that's kind of been forgotten. And we've gone to like, let me download this very cute, fuzzy thing. What happened in between?

8:00Emily Forlini: You're right. It's a huge shift. But I do think there are still a lot of concerns about AI safety. So it speaks to everyone's holding all these emotions about AI. So we're scared about it. But we're also interested in it. We don't want to get left behind. You want to try the new technology. This is a fun consumer product, especially in the case of Meta's Muse. So I think they exist together, which is why some of the messages about AI are so mixed and confusing.

8:24Andrew Nusca: And then my other question for you on Amazon, because Amazon blocked Meta's Muse right away. Can you talk to us about that and what's going on there?

8:32Emily Forlini: Yeah, that was huge news. It speaks to some of the conflict that this huge new wave of technology is basically wreaking havoc on the web. and the biggest retailer said, no, we won't allow people to use them. So that's very big. And a lot of reasons said it violated the terms of service, hesitation around people making purchases on Amazon through their assistance. And then Amazon not having control over that shopping experience is really big.

9:01Andrew Nusca: Now, talk me through where this goes over the next, let's say, five, 10 years. Should we expect, you know, Apple, Google, Elon Musk, maybe, to come up with their own form of these personal AI agents?

9:13Emily Forlini: Yeah, I think so. I mean, Elon Musk, you mentioned it. He's been talking about groups of agents collaborating. They've all been talking about this. So this is very much something that has been happening behind the scenes for years in my reporting. And now what's big with news and dots is now it's kind of reaching a mass audience. So this is very much a trend that's been bubbling for a long time. And we'll see if it becomes mainstream. From my perspective, since I report on OpenAI, Anthropic, Meta, Google, like all the big kind of frontier labs, I'm not sure if you could call Meta a frontier lab, but Meta has struggled with AI for several years now.

9:51Emily Forlini: This is their first big hit. So really what this means for me in my reporting is, is Meta something to watch? Are they going to become a bigger player, more serious player in AI? And that's kind of the conversation my sources are having now.

10:04Andrew Nusca: Excellent. Well, I look forward to seeing what's next. Thank you so much, Emily. And if you could stick around, I actually have one more question for you, kind of a selfish one about AI hiring and OpenAI in this process. So you had a scoop that I thought was really interesting about OpenAI adding something called the mission interview. Talk to us about what that is.

10:26Emily Forlini: Yeah, so it's a small scoop, but it's a big update in OpenAI's recruiting process. and I have a feeling it will become a fixture in that process for many thousands of candidates to come. So basically it's one interview loop with somebody who is outside the hiring team, so another employee, and they have a program that employees can do to get certified to lead these interviews. And with the name Mission Interview, we assume that it's kind of testing if the candidate is on board with the mission and seeing how they'll contribute to the mission. So what is the mission? OpenAI says it's that AGI or artificial general intelligence will benefit all of humanity.

11:04Emily Forlini: So very lofty. But how do you answer that without sounding scripted? I don't know. I mean, I got the sense. I'm not sure if there will be some of the same questions that they ask. I'm not sure if every interview will be the same. We do know that Anthropix's version of this is very, I'll say, wacky. They ask questions like, you know, if the stock price went to zero, would you still work here? And they, you know, tell me about a moral quandary you had and how you handled it. So it's very, it's different. I mean, this is kind of a common practice in tech, though, in this concept where you have someone from outside the hiring team who comes in and does basically like a cultural fit.

11:45Andrew Nusca: It reminds me of Amazon because, and I know you put this in your story too, but in my Amazon story I wrote earlier this year, the leadership principles were such a focus. And it's a little bit of a twist on that, but it feels very similar.

11:58Emily Forlini: Yeah. So, I mean, I worked at Amazon, so I went through that process. And I had that interview. It was called the bar raiser. So someone from outside the hiring team who's trained to be a bar raiser comes in, sees if the new employee is going to raise the bar, like raise the quality of other people in that position. And they also ask questions about the leadership principles. You don't know that they're asking about that, but they have a bank of questions for each principle and they will give that to the interviewer. And so that's a little more structured than what I'm hearing about this mission interview, especially based on what Anthropic is doing.

12:31Emily Forlini: I think OpenAI's will be different, but I don't know if it's backed by structure like leadership principles, which there are a fixed number that kind of inspire the questions. So I'm not sure. I think OpenAI is very much figuring out what it'll look like. It's a pilot now. It's only been rolled out to two teams. So we'll see.

12:49Andrew Nusca: And you wrote that they are welcoming, you know, disagreement, debate. Does that seem genuine to you, given all that they're going through right now with is AI good for humanity?

13:01Emily Forlini: So with this, I mean, OpenAI is going through a lot right now, which is in terms of its public perception. And so I think they're trying to scrutinize candidates and make sure that they're there for the right reasons and that they'll contribute in responsible ways. So they did talk to me about how they want it to be. There's intellectual diversity and debate. But they also use an interesting word, which is how candidates evaluate tradeoffs. And that's a big thing with AI. So again, it could harm humanity like we talked about, or it could be very helpful, just like with Muse and Dots. It's like both those things.

13:32Emily Forlini: So how do you trade off the impact of the technology? And so they used that word, which I thought was very telling.

13:39Andrew Nusca: Yeah. I also think it's interesting that you have all these people competing for pretty much the highest salaries you can get right now. You reported that they're offering a mid-career engineer a job that was$935 ,000 salary. Yeah. Crazy. Yeah. How, and you mentioned the word trade-off, which makes me think, like, if I'm trying to get a nearly million dollar salary, you know, how genuine am I going to be in my response? I just want that job. I want that money.

14:06Emily Forlini: Right. And there are more people. I mean, I had somebody who told me that researchers, AI researchers out of college are making$2 to$4 million. Like a former OpenAI employee told me that. We're in the wrong business. We're in the wrong business. I mean, they work all day and all night and they have really high stakes. I mean, the experiments that they run at work cost millions or billions of dollars when they're using the compute infrastructure. So the whole thing is huge, high dollar value. This is just my personal opinion. I don't think you can separate that. You know, of course, it's still a job.

14:38Emily Forlini: So you're at a job to make money, and that's a huge factor.

14:41Andrew Nusca: And as our, you know, future of work reporter here, I do have to ask you, do you think that this mission interview is coming for the rest of corporate America? Are other people outside of tech going to be seeing this?

14:51Emily Forlini: I don't know about the mission part because AI companies are very into that word mission and big words like humanity and AGI. But I think OpenAI is actually a little bit late to this concept. Like we talked about Amazon's had this for years. Anthropic already has one. A lot of companies have a kind of cultural fit interview. So OpenAI here is late and they're actually doing it, I think, as a function of the company getting bigger. So they're not a small lab anymore where everyone knows each other. They have 830 open roles right now. They're growing like crazy. And so I think what these interviews do is provide a control mechanism and a consistency with how they evaluate candidates as they get bigger.

15:33Andrew Nusca: Thanks, Emily. For more stories about hiring and the future of work, check out my newsletter, the Fortune Workplace Innovation Newsletter, every Monday. Up next, Brainstorm Editorial Director Andrew Neska takes us through Norway's ban on smart glasses. There's a new AI policy change happening right before your eyes. Literally. The government in Norway is proposing a temporary ban on AI glasses. AI glasses look just like regular glasses. Meta and Essilor Luxottica, the company that owns Ray-Ban and Oakley, they've done a big deal and now they have cameras inside regular looking frames. Where the AI comes in is that it adds that layer of intelligence on what it can see and hear.

16:14Andrew Nusca: So instead of just recording and processing these things, it's analyzing them. It can tell you who you're looking at or what the temperature might be. It's tough to find a person who isn't concerned about AI glasses. Glasses go with you everywhere. They go with you to the gym. They go with you to school. They go with you to the doctor's office. And these are places that perhaps we are not used to free the recording. And there absolutely have been problems with this. I mean, look, you can see a person holding a smartphone. Glasses are quite a different story. If someone is wearing a functional pair of AI glasses and recording you, there is a little light next to the lens, just like on your smartphone or your computer monitor.

16:53Andrew Nusca: But of course, the big concern is people tamper with those. Now, Norway's proposed ban is not a blanket ban. It only is to take place in certain environments, schools, doctor's offices, sensitive areas where we really, really don't want recording without consent. Norway's government says it plans to propose the bill as soon as possible. If that comes to pass, just like many other countries, Norway's parliament will vote on it and then the king will sign it into law. That's it for Fortune Daily. I'm Kristen Stoller. Remember to like and subscribe and I'll see you next time.

17:32Emily Forlini: Great.

17:32Andrew Nusca: Woo. Guys, this is so fun. Thanks for the best Wednesday ever!