A.I. Accelerates in Paris + Can A.I. Fix Your Love Life?

Well, big Valentine's Day show today, Kevin.

Yeah.

Happy Valentine's Day.

Happy Valentine's Day.

Are you a Valentine's Day person?

My wife and I do celebrate Valentine's Day, but we celebrate it a day early because if you go out to restaurants on Valentine's Day, it's a disaster. It's too crowded, and they give you these little crappy prix fixe menus. It's not a good scene.

Yeah.

So we are going out on Thursday.

That makes a lot of sense, and you've done this for years now.

Yes.

Do you tell the servers that you're sneaking in a little early for Valentine's?

No, but I don't think we're the only people who have figured this out—

Oh, really?

Because the restaurants have become crowded the day before Valentine's Day, too. There must have been a memo. What about you? Are you celebrating Valentine's Day?

So here's the thing. On paper, you would think that I would not be that into Valentine's. It's another commercial holiday designed to get me to buy a card at Walgreens, whatever. Who needs that? I'm throwing myself into Valentine's Day with a fervor that is frankly shocking even to myself.

Tell me about it.

I've bought decorations for my house—

Come on.

For Valentine's Day. Some of them are battery-operated. That's how into this I'm getting. I think it's going to be so much fun.

Wow.

Yeah.

You are such a wife guy now.

I've become an absolute tradwife for my boyfriend.

Yeah.

Yeah.

You have gone from being very cynical about romance to being the most romantic person I know.

Do you know the Kelly Clarkson song “Miss Independent”? It's about exactly this. It's about a sort of independent young woman who thinks she can make it in the world on her own, and then, oops, she falls in love and everything changes.

Mm.

So I'm looking forward to a great Valentine's Day.

Well, I'm excited for you.

Thank you. Me too.

I'm Kevin Roose, a tech columnist at The New York Times.

I'm Casey Newton from Platformer.

And this is Hard Fork.

This week, Kevin reports back from the Paris AI Action Summit—

Where there was a conspicuous lack of action. Then, this Valentine's Day, we're using AI to spiff up one of our producer's dating profiles. And finally, Hinge CEO Justin McLeod joins us to discuss what AI can do to help us find true love.

You know, I'd settle for a decent co-host.

Hey, now, I'm sitting right here.

Well, Kevin, like the Statue of Liberty, you've just returned to us from France.

Yes, or oui.

Yeah, you're fully bilingual now that you've spent 3 days in Paris. What were you doing over there?

1. Paris Rewrites the AI Agenda

Well, I was attending the big AI Action Summit. It was held in Paris this week, and it was a giant confab of industry leaders, government leaders, people from academia, and people from NGOs. Basically, a whole bunch of very important AI figures were there: Dario Amodei from Anthropic, Demis Hassabis from Google DeepMind, Sam Altman from OpenAI. The heads of the 3 biggest AI labs in America were there, as were a bunch of people from big AI companies all over the world.

Yeah. Now, this was the 3rd in a series of summits like this that we've seen over the past couple of years. Is that right?

Yeah. The 1st 2 of these summits, which were held in the UK and in South Korea, were more explicitly focused on AI safety. In fact, the 1st one in the UK, at Bletchley Park, was called the AI Safety Summit. But this one was very much not about that, or at least not in the official portion of the summit program. It was called the AI Action Summit, and a lot of the program was about the opportunities that AI would create, not the risks it could pose.

And is that because at the 1st 2 summits, they were able to solve the problem of AI safety once and for all?

No. There's a lot to talk about. I think this was a very big summit in terms of what it means for the future of AI. But before we talk about it, because we're talking about AI, let's roll our disclosures.

That's right. I'll start. Kevin, my boyfriend works at Anthropic.

And I work for The New York Times, which is currently suing OpenAI and Microsoft over alleged copyright violations related to the training of large language models.

So what was the vibe at this conference? What were some of the big headline takeaways that the people who organized it wanted us to have?

Well, maybe let's set the scene a little bit first.

Yeah.

So you get to Paris, and you have some choices. You can either go to the main summit, which is held in the Grand Palais, this giant exposition hall in the center of Paris, or you can go to these side events that are happening a couple of days before and a couple of days after the main event.

Is that at the Petit Château?

No. These are events being held all over Paris. Many of them have more specific topics. There was 1 about AI security, and there was 1 about AI safety. These were events sprinkled throughout Paris, but they were not themselves the main summit. I did a little bit of both. I went to some of the side events, and I went to the main event on the 1st day of the summit at the Palais.

At the actual summit in the Palais, you'd walk around the main floor and see these exhibits. They included things like how to use AI to preserve languages that are maybe going to die out, how to use AI to improve government benefits and welfare, or how to use AI to—

Wait, was DOGE there?

DOGE was not there.

Oh, okay.

But these were other, more European approaches—

I see.

—to the same idea. Or how to use AI for things like climate change and disaster relief.

Hmm.

So, all the ways that AI could help people—that was what they wanted to spotlight.

Got it. And my understanding is that President Macron of France also really used this opportunity to cheerlead for homegrown French companies and to separate himself from the rest of Europe, which I think he cast as maybe a little overly enthusiastic about regulation.

Yeah. President Macron definitely wanted to talk up France's entry into the AI race and the fact that they were not just going to let America and China dictate the future of this technology. He announced a huge new round of investments in the French AI ecosystem. He was also promoting Mistral, which is France's leading AI company, and he was very much interested in sending the message that, unlike the rest of the European countries, which want to regulate AI, he wants to accelerate and move France into the category of AI leadership.

And this was a message that was very much echoed by Vice President JD Vance, right? He gave one of his first major speeches since taking office. And Kevin, what did JD Vance have to say?

2. Vance Rejects AI Safety

JD Vance showed up on the 2nd day of the summit and gave a big speech in which he basically outlined an America-first AI agenda. He said, “We are not going to spend our time in America worrying about the risks of AI or trying to keep people safe from it. Instead, we are going to accelerate. We are going to remove all of the guardrails and restrictions because we want to get to advanced AI fast.”

Hmm. Did you bring a clip?

Yes. Let's play a clip from his talk.

I'm not here this morning to talk about AI safety, which was the title of the conference a couple of years ago. I'm here to talk about AI opportunity. When conferences like this convene to discuss a cutting-edge technology, oftentimes I think our response is to be too self-conscious, too risk-averse. But never have I encountered a breakthrough in tech that so clearly calls us to do precisely the opposite. The AI future is not going to be won by hand-wringing about safety. It will be won by building.

That was the message from JD Vance: worrying about AI safety is out, and worrying about AI opportunity is in.

Hmm. Now, how was this message received by other people in Le Grand Palais?

There were a lot of people there who were happy to hear this—some of the startup founders and other people who want to accelerate AI, and who think that the doomers have been spreading exaggerated fears that they don't agree with. But there were also a lot of safety-minded people in attendance at this summit who saw this as a huge mistake, a missed opportunity.

Actually, some of the AI leaders themselves disagreed with the kind of vision that JD Vance laid out. Dario Amodei, the CEO of Anthropic, wrote a statement in response to the summit saying, “We have to look at the risks. You all don't understand how quickly advanced AI is coming and the risks that it could pose, and so we can't actually just ignore all of that.”

Yeah. Well, Kevin, I have to ask what we think happened between the 1st 2 of these summits, where people seemed quite focused on the fact that powerful AI would, for example, be able to help people create novel biological, chemical, and nuclear weapons, launch cyberattacks, and spread hate speech, propaganda, and misinformation—you name it—and the Paris summit, where we said, “You know what? Do we have to keep talking about that all the time?”

Yeah.

3. Politics Pushes Safety Aside

I mean, I think a few things happened. The most obvious is the election of Donald Trump. Kamala Harris attended a previous AI safety summit and expressed some sympathy for the views of people who are worried about things like existential risk from AI. The Biden White House obviously put together its executive order on AI.

But Donald Trump is being influenced by a different set of people. He has, among his inner circle, people like Marc Andreessen, who is kind of an AI accelerationist. Others in his orbit also believe that the safety crowd is just a bunch of irrational doomers who are spreading these hyped-up fears about this technology. And so I think the biggest factor is just that Donald Trump was elected, and now his people are setting America’s AI policy, not the Biden people.

Here’s what I don’t understand about this. I think J.D. Vance believes that the United States is currently winning in the—if you want to call it—a competition to build very powerful AI before anyone else, and that everyone in his ear is sort of saying, “Yeah, we have a clear pathway to superintelligence.” What I don’t know, and I would love to find out, is what does J.D. Vance think happens after one of these labs creates superintelligence?

Yeah, it’s a really baffling piece of this. It seems like the accelerationists, the people who just want to take off all the guardrails and remove all the regulations and go full speed ahead to AGI, don’t actually believe in AGI.

Yeah.

Or at least they don’t have a real clear vision of what the world will look like after it. It’s a kind of techno-pessimism that’s masquerading as optimism. It’s like we don’t actually think the people at the labs who are building this stuff, or the people who are the pioneers in deep learning who are warning about these existential risks, are right, but we’re not going to say why. We’re just going to pretend it doesn’t exist.

Yeah. I wrote a column this week, and the headline was “America’s New AI Safety Plan Is ‘Let’s See What Happens.’” Because as far as I can tell, that’s what it is. We’re going to just create systems that are ever more powerful, and the Trump administration is signaling that it does not intend to regulate them in any meaningful way.

Yeah. And I think the thing that I felt at the summit was a sense of just two parallel conversations happening that have almost no points of overlap. One is the conversation that the people who are running the big AI labs and the people who are deep in the technical details of the AI itself are having, which is about the fact that we are rapidly approaching smarter-than-human AI systems. A lot of people we’ve had on this show who are running these big companies say that we could be only a year or two, or possibly 3 years, away from AGI or something like AGI.

And then you have this other conversation, which seems to be operating on a totally different timescale, which is, “Let’s start a blue-ribbon commission to plan for how workers can adapt to AI,” and maybe we’re going to release a report in a couple of years. I don’t know what to do about reconciling those 2 views, but it does seem very striking that the people who are the closest to the technology, who know the trajectory that these things are on, are sounding the alarm and saying, “You all need to feel much more urgent about this,” and then the politicians who are doing, I would say, a more traditional read of AI, just basically seeing it as another technology.

4. AI Starts Reshaping Jobs

Yeah. There was this fascinating story that I saw in The Wall Street Journal by Belle Lin, and it found that the unemployment rate in the information technology sector rose from 3.9% in December to 5.7% in January, which is way above the overall jobless rate of 4%. Why is that interesting? Well, some of the economists Belle writes about in this piece believe that this is the first sign that AI automation is starting to cause significant job loss.

Yeah, this is something that I’ve been interested in, too, because I think we’ve all been waiting for signs that the AI gains of the past few years—the progress in capabilities that are improving—are actually translating into changes for workers. And I do think that the first place this is showing up is in software. I’ve met founders who tell me that they are no longer needing to hire junior software engineers because now a senior engineer with an AI tool can do much more work, work that might have previously required 3 or 4 people under them to do.

Yeah, absolutely. So I think this is important to bring up because while I am a person who likes to write about AI safety, because I think that it is something that people should take seriously, for me, AI safety is kind of a proxy for just all of the disruptions that AI is going to cause in general, which I think politicians and regulators generally pay too little attention to. And I have to tell you, there are very few things more destabilizing to the politics of any country than huge unemployment numbers. And so if we’re already starting to see those creep into the picture, then I think it is a strategic mistake for folks like J.D. Vance to stand at a podium and say, “We’re not going to worry about any of this. It’s just going to be go, go, go to the finish line,” because in truth, there might actually be a huge political problem bubbling up under their nose.

Yeah, and J.D. Vance said explicitly in his speech that he doesn’t believe AI will replace workers. It will only help them, which is a fine thing to believe, but it is at odds with what the people who are building this technology have said. I mean, Sam Altman has said that they are building drop-in workers that you could just hire at your company. Many AI leaders have said that they do expect AI to create substantial job displacement in the coming years. And so if you’re J.D. Vance or someone who’s on his side of this, you actually have to do more than just say, “No, it won’t.”

Yes.

You have to actually explain why you believe that the people who worry about AI creating job loss are being irrational.

Yes, absolutely. Let me throw one more statistic at you. This week, Anthropic put out the new Anthropic Economic Index, where they’re publishing a lot of data about the impact of AI on the economy as we know it so far. And one of the things that they found is that right now, AI use leans more toward augmenting what a human being can do than replacing them outright, but maybe not by as big a margin as you might think. What they found is that about 57% of the AI use that they are seeing in their own tools is augmentation of what humans can do, but 43% is automation. So let’s assume that number goes up over time. This stuff is going to get real in a hurry.

Yeah, so I think the jobs area is one where the J.D. Vance vision of “This is all going to be great, and if you’re saying otherwise, you’re just a pessimistic doomer” is going to clash with observable reality. But I also had a conversation with an AI safety person at the summit who was making the point that taking these extreme risks seriously is a prerequisite for acceleration. It is not actually opposed to acceleration. And the example that he used from historical analogs was nuclear power.

Mm.

5. Safety Enables AI Acceleration

So we had nuclear power in the 20th century, and then Chernobyl happened, and Three Mile Island happened, and it really delayed the popularization of nuclear energy by several decades.

Mm.

It’s only recently that countries have started to build new nuclear power plants, and it’s not actually because they took the risks of nuclear power too seriously. It’s because the people who were engineering the plants during the first wave of nuclear power didn’t make them safe enough.

Mm-hmm.

And so the analogy that this person was drawing to AI is that if you want to accelerate this technology, if you want to allow it to move as quickly as possible, then you actually have to think about safety. Because if you don’t, something catastrophic is going to happen, and when that catastrophic thing happens, people are going to react very strongly to that. You end up delaying this technology more than it would have been delayed if you had just taken safety more seriously in the first place.

Mm-hmm.

So what do you make of that analogy?

I think it is such an important point, and it is something that I wish the folks who are so dead set on acceleration would pay attention to.

6. Musk Targets OpenAI

One more thing happened in Paris that was notable, which is that there was this big dinner with the heads of state and the leaders of the AI companies. And at that dinner, while Sam Altman was sitting next to J.D. Vance, the news came in that Elon Musk was proposing to acquire OpenAI’s nonprofit for something like $97.4 billion.

Yes, and this appears to be a trademark piece of legal trolling by Musk and his allies. OpenAI is, of course, in the midst of trying to convert itself into a for-profit, and in order to do that, the nonprofit has to get a fair price for the assets of the for-profit. This bit of trolling from Musk could make that more complicated.

Yeah, this is all very complicated. We’ll get into it probably in a future episode. I imagine that this is not the last time we’ll talk about this OpenAI restructuring and some of the challenges there.

But suffice to say, Elon Musk did not like the idea of an AI summit happening in Paris without him, so he figured out a way to insert himself into the conversation.

It's amazing Elon Musk had time to do that in the midst of dismantling the entire federal government.

Oh.

Now, Kevin, often at these summits, some sort of declaration comes out that tries to signal some spirit of collaboration: “Here are some things we're going to work on together.” How did the United States feel about the declaration that came out of this summit?

Well, not great is the short answer. There was a draft that was leaked before the actual final version of this statement went around. It was pretty watered down already. It did not talk a lot about risks and harms that might result from AI.

But the U.S. wouldn't even agree to sign that statement, possibly due to some language in there about making sure that AI was inclusive, which I imagine, to the people in the Trump administration, just sounded a little too woke for them. So the U.S. did not sign this. The U.K. also did not sign this, and I would say that most of the people who looked at this statement thought it was lacking in one way or another.

I love the idea of a summit that produces a completely anodyne statement that commits nobody to anything, but is still considered too restrictive for global superpowers to sign.

Yeah.

China signed it.

They did.

That was nice.

So I was thinking on the plane ride back from Paris about what I would take away from this summit.

Number 1: a whole bunch of croissants. Number 2—

I did bring back 1 pastry for my wife.

Smart.

Although it got crushed in my luggage.

Oh, no.

So sorry.

Sacrebleu.

The sacrebleu. I think the major takeaway, at the biggest possible level, is that we are just not going to see the kind of international AI safety cooperation that I think a lot of people had hoped for.

There was this idea that a new United Nations–type structure would emerge, where governments all over the world would see it in their common interest to cooperate in making rules to govern the AI systems that were becoming increasingly powerful. I do not think that is going to happen after going to this summit.

I think countries are going to take a much more nationalistic approach to this. They will want to win. They will not want to cooperate, for the most part. I also realized that in the United States, during the Trump administration, we are not going to get any meaningful AI safety regulations at the federal level.

There may be some stuff happening at the state level, but I think we have pretty clearly seen J.D. Vance signal that the administration's position is the “all gas, no brakes” position when it comes to AI.

Hmm. So it sounds like it was more of an inaction summit.

Yeah, it was kind of an action summit to say, “We're not going to take any action.” When we come back, we're going to talk about a different kind of AI action: how AI might be helping you get some action on dating apps this Valentine's Day.

Well, Casey, this episode is coming out on Valentine's Day. Happy Valentine's Day.

Happy Valentine's Day, Kevin. Love is in the air, and today we want to talk about it.

7. Dating Apps Bet on AI

Yes, we have a Valentine's Day special. Today, we are going to talk specifically about dating apps.

Kevin, I think that this conversation is coming at a perfect time for us because, while on the one hand, dating apps remain a very popular way for people of all ages to meet each other, they've also been going through a really hard time.

If you go online, you find mounting frustrations from users who feel like dating apps are finding more and more ways to charge them. People feel like, essentially, “You're extracting more and more money from me. I'm going on fewer and fewer dates. I have not found the love of my life, and something needs to change.”

So there's a lot of frustration out there.

Yeah, and part of that may be that these apps are just not doing as good a job as they used to of getting people connected to their matches. But I think there's also this question of whether dating as a whole is less popular than it used to be.

It could be, but what we know is that the CEOs of the dating apps are proposing a solution, Kevin. It is the same solution that CEOs of every kind of app are proposing these days, which is: they're going to sprinkle some AI on it.

Yes. So today we're going to take a look at how AI is reshaping—or not reshaping—the world of dating. We're going to do this in 2 parts.

First, we are going to focus on where things stand with AI features on dating apps right now, and what 1 power user of a dating app called Hinge is learning about the way that AI can both help and potentially hurt finding matches online. In the second half, we're going to talk with Justin McLeod, who's the CEO of Hinge, about how he thinks AI will transform the experience of online dating.

And by the end of all of that, Kevin, we hope to have found love, not just for our super user, but for all Americans.

Yeah.

Yeah.

And maybe find it in a hopeless place.

Hmm. Well, let me say, dating apps, more often than not, do feel like hopeless places. So let's hope we can find some love there.

To start, we've invited on our resident dater on Hard Fork, our producer, Rachel Cohn. Rachel's credentials include being the only single person on our team, going on a lot of Hinge dates—

Mm-hmm.

—and being very open about her experiments with finding love.

Crazy open.

Yes. She writes a great newsletter on Substack called Are You My Boyfriend?, where she talks about her experiences in dating. Today she's going to tell us about what happened when she followed all of the advice of Hinge's new AI tool to tweak her profile and find more matches.

Oh, let's find out how that went.

Rachel Cohn, welcome to Hard Fork.

Hey, guys.

Rachel, it's great to have you here. Let's talk about this experiment.

What do you mean—

You can't laugh.

—to have you here? She produces the show.

That's what I mean. It's great for her to be here with us, talking to us.

Yeah.

It does feel really formal.

Well, yeah. It's fun to be in front of the mic for a change. So now, Rachel, back when I was doing online dating, I was mostly on Tinder, but for this experiment, you've decided to use Hinge. Tell us about why you made that choice.

Tinder is actually the biggest app. It has the largest number of users of any dating app in the U.S. and also globally, but Hinge is 1 of the fastest-growing right now.

For me, I'm 30, I'm in New York, I'm straight—everyone that I know is on Hinge.

Right.

Okay. It's the cool kids' dating app.

Totally. The other thing that's really relevant to what I'm going to tell you guys is that Hinge has really marketed itself very intentionally as a place for serious relationships.

That's also really interesting in terms of thinking about how AI is going to be incorporated into the app, because it seems potentially problematic for Hinge's brand identity—which is all about meaningful human connection—to infuse a bunch of AI into the platform.

Got it.

Got it. So it's 1 of the most popular, and maybe it has the most to lose with this pivot to AI.

Yeah, I think that's fair.

All right. So we've picked Hinge as our experiment bed. What was the experiment that you ran?

8. Rachel Tests Hinge AI Advice

Match Group, which is the parent company that owns a bunch of dating apps, including Hinge, had this big investor day back in December, when CEOs from several of the different dating apps got up on a stage and talked about the new AI tools that they were going to roll out on the platforms.

Hinge talked up 2 different kinds of big buckets of features that they were going to roll out. One is something that is not yet out and that they would not give me access to. Apparently, it's coming next month, and it involves using AI to make better matches by improving the algorithm in some way.

The other tool, or category, that they're working on is AI for what they're calling the struggling dater.

Now, do you identify as a struggling dater?

That's a good question. I would say it depends on what we mean by struggling. I'm struggling—

Yeah.

—in the sense that I've been really actively dating for 2.5 years, and I still haven't found the right person. But I'm not struggling in the sense that my issue is getting dates through Hinge.

I see. Okay. All right. So in some cases struggling, in other cases thriving.

Some would say.

Some would say.

To put it in business terms, the top of the funnel is—

Great.

But it's the conversion—

We gotta convert.

Yeah, you gotta convert.

That's right.

And that's where AI rolls in, is that right?

Okay, well, so this is what's unclear to me.

I'm gonna be really curious what you guys think about this tool. But the one tool that they allowed me to use and that everyone can actually use now—this is totally public; it came out a month ago—is called AI Prompt Feedback. The whole point of this tool is very simple: They have some kind of AI model that is looking at your written responses on your dating profile and giving you feedback on those written responses so that you can improve what you have on your profile.

So, Rachel, I'm gonna ask you to slow down because I met my wife back in the Stone Age. We met face-to-face, like Neanderthals. I have not spent time on dating apps as a result. So my understanding is that you have a profile that you fill out. It's got things on it, like your interests and where you went to school and what you do.

Yes.

And what you're saying is this is basically an AI coach that will look at your profile and tell you, “This could be punched up. This could be funnier. Maybe don't say that thing about your body odor.” It sort of goes through your profile and critiques it for you. Is that what I'm hearing?

Yes. Okay. So Hinge actually requires you—in order to set up a profile, you have to choose from a preset number of prompts they have, things like “Together we could,” and then you fill it out.

Oh, yeah. Can I tell you my favorite example?

Yeah.

One of the classic Hinge prompts is “I go crazy for,” and one time I saw somebody on Twitter just post, “I go crazy for food.”

And that really tells you a lot about the median level of writing on these profiles, Kevin.

Got it.

Okay, so that's the setup here. Kevin now understands what an online dating profile is, and the rest of us understand that Hinge has these prompts, which can show off your personality, but maybe the average ones are pretty flat, pretty nonunique, maybe. So then along comes AI, and there's a tool, and Rachel, you used this tool.

Okay, yeah. So I'm gonna actually share a PowerPoint with you guys.

Oh, good.

But we'll talk through this for our listeners who are not tuning into our YouTube. Okay, share.

For the listener, we're now in a Google Slides presentation. Can I just say, this is so Rachel. This is so... It's very Rachel. Okay. So here we're about to look at Rachel's original written prompts before AI intervened.

Oh, yes. Okay, so before I tell you about what these written prompts are, I'm just gonna tell you what kinds of responses this AI algorithm gives. Because even though it will give you very personalized, specific feedback, it actually only has 3 top-line categories of notes it will give you.

So the 3 categories are: It will either say, “Great answer,” which means you don't need to change anything; you're doing a great job. “Try a small change,” meaning make a little tweak.

Mm-hmm.

Or “Go deeper,” which I have interpreted to mean you really need to be more vulnerable. You are not sharing enough about yourself.

Yeah. You need to look within yourself and find something more interesting about yourself. So it will never say, “Go shallower,” like you're oversharing—maybe take out that part about your childhood trauma?

Yes, you are going to see it will never say, “Go shallower.” It really, really loves specificity.

Okay. Let's go. Let's see.

Okay. So these were my original prompts. Miraculously, this was not planned. I actually got all 3 kinds of feedback, which is very useful for this experiment. So I got 1 great answer, 1 try a small change, and 1 go deeper. We can start with my great answer, because that's the one I did not change.

Okay.

And that's here. The prompt was “My simple pleasures.” Do you guys want to read it?

Yeah, so Rachel's simple pleasures would include writing down a funny piece of conversation she overheard on the subway, dancing a little bit on a run, and making incredibly average watercolor paintings. This is a great answer. Yeah. It's fun. It's specific. It's a little self-deprecating. And if you're using Hinge, you've never seen that answer before. Correct.

Wow. Thank you.

So that one, the AI said, “You killed it. No change. No notes.”

Yeah. And to exactly your point, it said, “It showcases your personality and invites fun conversation about art and humor.”

Okay, so that's a great one. Let's take a look at a prompt where maybe we could use a small change.

Okay, great. So here was my original one. I'm just gonna read this one out to you. The prompt was “My greatest strength,” and my original answer was, “Getting people to share stuff they normally wouldn't or shouldn't.”

Mm.

“This is very useful for my job but sometimes gets me into trouble in my personal life.” What do you guys think?

I like it. It's a little dangerous. It makes you sound like Sherlock Holmes. I can see how maybe some people would sort of... That might rub them the wrong way, right? Because dating is a little bit vulnerable, and maybe they read that and think, “Okay, if I'm gonna go on a first date with Rachel, maybe this is gonna feel a little bit like an interrogation, and I have my walls up.”

So I could see how maybe fewer people might be interested in that one.

I think that's a totally fair point. I will say, my own thinking about this is sometimes when I'm creating my responses to prompts, I'm actually thinking about not what is gonna appeal to everyone, but what is actually gonna push certain people away, like the people who will not be compatible with me.

I see. Okay.

But I think that's a really good point. Interestingly, that is not the thing that the AI focused on. So the feedback I got from the AI was basically just that it wanted me to be more specific, to give a specific example about a conversation I had or something that I learned about someone.

So I was actually trying to test the AI a little bit here, be a little bit funny, and see if it would tell me, “This is not a good idea to give that kind of an example.” So the way I tweaked it first is I said, “My greatest strength, again, getting people to share stuff they normally wouldn't or shouldn't, like how much money they make or what they most regret in life. This makes me feel bonded to people, like together we share a secret.”

That's good.

Well, maybe.

No, I like the last one better. Really?

Yes.

Who wants to go into a first date being like, “This woman is on a mission to get me to reveal my salary”?

Yes, exactly.

Well, as she said, it's a mechanism to filter out people. If you're opposed to salary transparency, you're not gonna enjoy going on a date with Rachel Cohn. That's just the filter that she's setting up.

No, I'm with Casey on this.

Yeah.

I think this is bad. I was hoping the AI would say, “Listen, that's gonna be off-putting to some people,” but it did not. It just said, “If you wanna add more, consider sharing why you enjoy these conversations or a specific example.” It wants me to be more specific.

Hmm.

So I updated it again. Also, mind you, if you're looking, you can see here that I am now 3 characters away from the limit. It's very hard to satisfy the AI and also stay within the character count.

Finally, here's the one that actually worked. This took me so long. I was workshopping this at a dinner with friends. My friends are racking their brains for the kind of questions I ask. Also, every time we would try something, I would run out of characters. Finally, I broke the whole structure again, and I was like, “Okay, I'm really just gonna tell one story.” So this is what I ultimately ended up with:

“My greatest strength, getting people to share stuff they normally wouldn't or shouldn't. Recently at a wedding, a stranger told me how much money he made and why he felt constrained in his relationship. It gave us a shared secret.”

And how much money did he make?

I cannot say that.

Wow. Wow, you turned the tables, and all of a sudden she clams up. Okay, so now, for reasons that remain unknown to me, we have finally satisfied the AI, and this is a great answer.

Do you feel that way? I mean, I'm curious. Do you think this is better than what I had originally?

No, because here's the thing. I don't read that and think, “Oh, now I understand something about Rachel.” What I understand is something about a stranger you met at a wedding. Do you know what I mean?

Yeah. Do you feel that same way, Kevin?

Yeah. Well, I think it's a little better because it conjures the image of you going up to someone and getting a stranger to tell you their secrets, and makes you think…

But it still has the same issue, which is that if I'm a person with a lot of secrets, I'm turned off by this.

Here's how I would do it. I would do it more like, “My greatest strength is accidentally getting people to reveal things.” For example, make it seem—

Oh.

—like people cannot wait to tell you things.

Oh, accidentally.

If it makes it seem like you're constantly prying for information—

Oh, interesting.

—that's a turnoff. But if it's, “I'm so—

I'm just so—

—sort of effervescent and bubbly—

—charismatic—

—and people can't wait to tell me their secrets—

Mm.

—now I'm interested.

Yeah.

That's actually—

Yeah.

—that's really smart. I feel like that's—

Thank you.

—not totally the truth about me.

He seems to be a dating coach.

But that is definitely a more appealing way to sell myself to people.

All right. So we've managed to best the AI one time.

Okay, yes.

Okay, very interesting. Now let's look at the next one.

All right. So this is the one that was “Go deeper,” so this is the one I apparently needed the most help with. And I was actually really surprised by this one because my original prompt was, “I'm looking for,” and I wrote, “Someone to read a book next to in bed.” And let me tell you, this—

That's cute.

—this crushes it on Hinge.

Yeah.

I get more responses on this than any of the other ones. So many men will send me just a heart-eyes emoji, or people will say, “What book are you reading?”

Aw.

Just to state the obvious here, I think part of what's great about this response is that—

It's also aspirational because in reality, what you're gonna do with the people that you're dating is watch TikToks in bed next to them.

Please stop it. Rachel, what were you gonna say?

No, that's actually so fair. But what I was gonna say is I think part of what works about this is that it's intentionally vague so that—

Mm-hmm.

—you, as a potential suitor for me or whatever, a potential match for me, can project yourself here.

Well, let me say another obvious thing. It also implies that you're in bed together. That's appealing to a lot of men—

And it gives men the opportunity to brag to you about the books that they're reading.

Exactly.

That is the key piece to me.

Yeah. So this is a perfect answer, to which the Hinge AI naturally said, “Let's go a little deeper.”

Yeah.

And so how did you try to do that, Rachel?

All right. It said, “Try adding more details about your favorite books or genres. It helps show your interests.” So I actually originally changed it to, “I'm looking for someone to read a historical fiction book next to in bed.” And it said, basically, try again, and it suggested that I could solicit recommendations. So I tried again and I said, “Someone to read a historical fiction book next to in bed. Open to recs!”

Ooh.

Which I feel like—

Cringe.

I know. I feel like it's so cringe.

All of a sudden, I'm dating Goodreads.

Exactly. It's like, am I trying to meet someone, or am I just trying to get book recommendations? But the AI still wanted me to be even more specific. It said, “If you wanna add more, consider sharing your favorite historical fiction titles. What draws you to this genre?”

Now, see, again, the AI has just been tuned wrong.

Yes.

Right? Because the AI has been tuned to say, “Make sure to ask for specifics. Make sure to ask for details,” which can be really good on a prompt, and I do think it's a message that a lot of online daters need to learn. But because it has no human sense of what is actually cute, it could not read Rachel's original prompt and understand all the ways in which it would obviously appeal to a lot of people.

Yes.

So I'm actually giving the AI a failing grade on this one.

Yeah. I'm not impressed. I liked your original better.

Thank you. I don't love this one. But here's where I ended up that actually made the AI happy. I ended up with, “I'm looking for someone to read a historical fiction book next to in bed. Open to recs! Especially love stories set in New York City.”

No.

Oof.

No. Delete.

Yeah, it's delete.

It's giving LinkedIn.

Yeah. Goodbye.

Okay.

Goodbye.

Well, I thought the same thing, but I changed this a little over a week ago, and you guys, I was very surprised. I actually got a lot more engagement.

Oh, thank God.

Oh, boy.

Really?

So wait, let me show you. This next slide is responses from men specifically to this book one. I have here someone who's written to me, like, “Okay, it's not New York City-related, but it's definitely love and historical fiction. Let me know if you're interested. It's a Pulitzer Prize winner.”

Hmm.

I love this one too because they didn't give me the book in the actual—

It was just a tease.

—initial message.

Yeah.

So exactly. I have to match with them to get the name of the book. And then something similar here, and someone's recommending a specific book.

Now, I have a question about this, Rachel.

Yeah.

So my presumption is that some of the men on dating apps are also using AI to craft their responses. So what is your sense of whether these people are kind of running your prompt through an AI and just copying and pasting what the AI tells them?

Hmm.

I would say that I personally do not think I have come across many people who are using AI, and there are a few reasons that I feel that way. One is that the grammar is horrible. Horrible. Also, even if you look at these two examples, this feels very specific and not something—

Yeah.

—AI would've come up with.

Yeah. Well, these are both really good responses. They're very engaging. I do think that they reveal something that the AI has picked up on, which is that if you ask a man to recommend something, he is unfortunately gonna have an opinion about it that he might wanna share with you.

That's true.

So that was smart. Now, Rachel, I'm seeing in this screenshot that your battery was all the way down at 19%. Do you usually let it get that low?

Actually, my battery's at 3% in that other one. Yes, actually, a huge complaint among my friends is that my phone is always dying, and I refuse to—

Oh, wow.

—invest in a—

Can we get you, like—

—a portable battery?

—like a battery pack?

That's our Valentine's Day present to you.

Yeah.

We're getting you a battery pack.

That would actually be so useful.

Okay, let's keep going. What else did we learn? What is your overall impression of this AI feature? Do you think it helped you make more matches? Do you think it helped you get better matches? How useful is this AI feature for struggling daters?

My initial reaction to this tool when I started using it was similar to how you guys are feeling, which is, this AI is not tuned properly. It does not understand humor. It does not understand all these important features about being sort of mysterious in dating. But actually, now that I went through the process and did have more people engaging me, I do have this sense that it did do something very useful for people, which is, I do think it gave them a more concrete, narrower, and maybe more vivid picture of me. And—

Hmm.

—it sort of reminded me of what happens to high school students when they are being coached by guidance counselors to apply to college. You end up having to sell a very specific story about yourself, which everyone knows is a little bit reductive and not totally authentic, but there's a reason that you're coached to do that. The reason is that, for people who are screening applicants, it's easier to have a vision of the person and have some sort of visceral reaction to, “Should we admit that person or not?” I think the same is true in dating to some extent.

Yeah, that makes sense. It also seems like maybe there's a point. This seems like a very minor augmentation of your dating profile, but I think you can imagine a scenario where AI is allowed to write your whole profile for you. Or maybe it goes through your social media or your LinkedIn, or you write a little description of yourself, and then it crafts a profile for you.

But I think at some point it would start to feel like an inauthentic representation of who you were if you hadn't actually done the writing of the profile yourself. Did you feel at all like that—like you were maybe giving over this very personal thing to an AI that doesn't actually know you that well?

So I think this was actually something smart that Hinge clearly very intentionally did. Part of what makes the tool not that helpful, but also what prevents exactly what you're talking about, is they don't actually tell you, “Here's what you should say instead.” They just give you the same advice that an English teacher in middle school would give you, which is, “Show, don't tell. Give more examples.”

And I found it extremely hard. It took me, in one instance, actively plugging away at this for a full 30 minutes—

Hmm.

—of tweaking and rewriting. That's a huge time commitment. And, yeah, it was really hard to satisfy the AI.

What I'm curious about is what the experience of the median Hinge user will be, because you are not the median Hinge user. You are a great writer. You work in media.

Gas me up. Gas me up.

You know how to sell a story. You're gorgeous. You're charming. You don't need the help of an AI, okay? But when I think about all the Hinge profiles that I used to see of these guys, it would be like, “What are your interests?” And it would be coffee, food, and travel.

Yeah.

And I thought, I don't know one thing about you. Obviously, you like food. We all eat it every day, right? So I'm curious, for those people, how useful do you think this kind of tool can be? Can it get them from “I love coffee, travel, food” to actually sharing something revealing about themselves?

So this is exactly my big question that I am really hoping you guys will push Justin, the CEO of Hinge, on. They have marketed this tool, especially to investors, at least, as a tool for struggling daters. What do we know about struggling daters? They either don't put much time and effort into their profile, or they are not comfortable putting themselves out there. I don't see how an AI telling them, “Share a little bit more about yourself,” is going to actually help those people. If you're someone who's writing, “What do I order for the table? Pizza,” I think that is an intentional choice.

Yeah.

Well, Rachel, this is a fascinating experiment, and thank you for walking us through it.

Happy Valentine's Day.

Happy Valentine's Day.

Happy Valentine's Day.

Well, Kevin, a lot hinges on our next segment.

Yeah?

Hinge CEO Justin McLeod is here.

That's when we come back.

All right, Kevin. Well, now that we've heard Rachel's experience on Hinge, I think it's time to take that feedback straight to the source. Let's bring in Justin McLeod. He's the CEO of Hinge, and he's been thinking a lot lately about the future of AI in dating.

Yeah, let's get a little unhinged.

He'll be joining us from the McLeod. All right.

Hi. Thank you for having me. Really appreciate it.

Well, we couldn't think of a better CEO to have on our Valentine's Day episode. I have to imagine that Valentine's Day is big over at Hinge. Is this like tax season over at Hinge for you guys?

In the early days, we actually used to give Valentine's Day as a holiday to all of our employees. But, interestingly, Valentine's Day is really a day for couples and less for singles, so it's a bit of a tricky love-hate relationship with Valentine's Day for Hinge.

I have to tell you a story, which is that when I was single, I had a friend who was like, “If you're going to find somebody, you have to subscribe to the premium version of one of these dating apps.” So I actually bought Tinder Premium on Valentine's Day, and every year for three or four years, it would renew on Valentine's Day.

You get the renewal.

Exactly, and it was always just a reminder that I hadn't quite yet figured things out. So, in that way, Valentine's Day was big for me.

Yeah.

Now, Justin, I understand you have kind of a wild story about meeting your wife. Do you want to tell us that story?

We could take the whole podcast, probably, for the story. It's a long one. I married my college sweetheart, but we had an 8-year break where we didn't speak to each other. I started Hinge during that period, about halfway through that period, and I let her know. I included her on all the launch emails, and—

Just to be like, “Hey, just in case anyone is out there.”

Just like, “You know, I'm still…”

Yeah.

Yeah.

Wink, wink.

But the short version is that about a month before she was about to get married to someone else, I flew over to Switzerland and asked her to call off the wedding. She moved back to New York, and we just had our 8-year wedding anniversary yesterday.

Oh, wow. Happy anniversary. Happy anniversary.

Thank you.

We have just talked with our producer, Rachel, about her experience using some new AI features on Hinge. Yeah, and spoiler alert, she's still single. Let's just say that. And I've only ever had an experience of AI trying to break up my relationship. But you all think that you can use AI to help people get into relationships. So tell us about the overall picture you see of how AI and dating apps are poised to intersect.

9. Hinge Keeps Dating Human

Yeah. I'm sure you all are thinking about this a lot, but AI is going to be a pretty disruptive and transformative force in a whole lot of industries, and I don't think dating is any different than that. I think that when I see what's coming, it's going to be a bigger transition than even what happened with the transition to mobile.

Hmm.

And there are a couple big ways that I think this is really going to come to life. One is personalized matching, and two is effective coaching.

Personalized matching is getting to know our daters on a much deeper level—about who they are, what they're looking for—and being able to make much more thoughtful matches with people that go 2 or 3 or 4 or 5 levels deep to make sure that they're aligned, so that you don't end up with someone that, 3 months in, you realize you're misaligned on some major sense of your values or some—

Right. Yes.

—you're misaligned on your goals and expectations.

This is the alignment problem. No, that's something else. It's the other alignment problem.

Right. So that's one piece of it, I think: just much more thoughtful matching. We're going to be able to ask much more nuanced and deeper questions, understand much more nuanced feedback.

The other big factor is the coaching piece. We have a whole lot of daters who really struggle to get on that first date or even get that first match, and they don't always know why. Is it the photos they chose? Is it the prompts they're writing?

And we can help push people and guide people to fill out their profiles well, take the right actions, and just not get in their own way, because none of us are born good at dating apps.

Yeah, so those are 2 potentially useful approaches. Maybe let's take them in order a little bit. I want to ask about this idea of AI matchmaking. My understanding is that, for a long time, dating apps have used various machine-learning algorithms to pair people up.

They seem to be grouping and scoring people along certain criteria, maybe how attractive this person is viewed within the community, that sort of thing. What are those next 2 or 3 layers? What are the other signals that you could glean, and what makes you confident that they would paint a picture of somebody that was actually useful in connecting them to somebody else?

Yeah. So there's a number of different ways to think about this. First of all, when Hinge launched—the version of Hinge today, back in 2015—the algorithm is mostly driven by not even so much information that we knew about you, but more about your relationship to other people in the app. Who did you like? Did other people who liked that person also like this person? So maybe we'll suggest this person to you.

It was a lot of—

You guys are technical, so collaborative filtering—

—collaborative-filtering models as opposed to content-based. So one is just moving into a much more content-based world, where we're actually taking a look at your photos, actually reading your prompts, using that information to make much more thoughtful and intelligent matches. So that's one layer.

We can start to ask much more nuanced questions that don't always fit into an easy multi-select dropdown option, like, “What is your religion?” or “What's your height?” But understanding your backstory, what matters to you, and what your relationship history is, and using all that information to make the logical connections that you would make can lead to much more thoughtful matches.

And then there's a whole other layer to that, which I'm excited we're already starting to explore: How do we use relationship science? There's a lot of papers out there and a lot of people who have thought about what types of personalities do well together and what leads to long-term relationships. We can bring that in—not just to learn your tastes and their tastes and see the likelihood of an initial match, but actually to think more deeply about long-term compatibility.

Right. Let me ask the cynical question, because I think some listeners are maybe going to be thinking this: When you're on an online dating app, isn't 90% of it just whether you think they look cute in their photos? Is there really that much deeper left to plumb with AI and all the rest?

I think there's a ton. A lot of people will make snap judgments based on photos, but that's just the foot in the door. Then there's a whole lot of other aspects that matter, whether you're actually going to show up on a date with this person, get along, and want to go out on a second date. That's why I think Hinge has been so successful: We haven't really based it on photos.

If you like someone, you actually have to go through their profile and choose something about them that you did like. Everyone has to fill out 3 prompts. Let's say looks matter, too, and it's the easiest thing to make a snap judgment on, but there's a whole lot that comes after getting your foot in the door with regard to looks.

Got it. Well, let's talk about the coaching piece of this, too. We just talked with our producer Rachel about this new prompt feedback tool that you all have rolled out. What other ways are you seeing AI be potentially useful in coaching users on Hinge?

Yeah. It's going to run the gamut from starting with tips and tools all the way to really helping people navigate the emotional ups and downs and maintain a sense of confidence and hopefulness through their journey. We're starting with the tips and tools, so that's things like a photo finder. We can learn what types of photos do well on Hinge and what don't, and build models that can help you choose which photos to select for your dating profile.

Another one is prompt feedback, which is a really exciting thing that we just launched. We've learned that through that feature, we were able to triple the incidence of high-quality prompt answers and reduce by more than a third those low-quality, one-word answers, which is pretty transformative.

So I want to understand a little bit about how you built this system and what it is looking for. As we were looking through Rachel's prompts, it seemed like mostly what the AI was trying to do was elicit more specificity in her answer. I'm guessing that's one of the criteria. Are there others? And how did you come up with what essentially were the criteria that you were going to use to say, “This is a great prompt,” versus, “Go back to the drawing board”?

There are a few different criteria. Is this easy to start a conversation from? Specificity is a big thing because the more specific it is, the more it lets people in. Does it reveal something about you? It is largely pushing people toward more specificity when they respond.

And is that rooted in an empirical sense? Do you have a sense that when people are more specific, when they write these prompts, they tend to get more answers and meet more people?

Definitely. We've been doing that since the beginning, not just looking, by the way, at the prompt answers, but also at the prompt questions. What types of questions actually lead to people responding in a way that leads to a date?

We think about this kind of efficiency frontier of vulnerability, where there are some questions we can ask that people feel very comfortable answering but that don't lead to a date, like my go-to karaoke song—

Hmm.

—which everyone's willing to answer but never leads to a date. And then there are things like what I wish I could change in my relationship with my mother that, if you answer, might lead to a really in-depth conversation about your values and everything else, but no one's really willing to do that.

That's a lot to post online for everyone to see.

Yeah, exactly.

Yeah.

So there's this sweet spot that we're always trying to navigate at Hinge: Where is the point at which people are willing to share, and where is the point that actually lets other people in? It's what other people want to know.

Right. Now, I have a question about this use of AI for improving one's profile or maybe pointing out where your answers are a little generic. If everyone is using this stuff, doesn't that kind of flatten the landscape?

Part of what you're doing when you're trying to look for people on an app like Hinge is filtering people out as well as filtering them into your set of potential matches. So if I see someone who just says, “I like pineapple on pizza,” that may be a red flag for me because I say, “That person's generic,” or they're not creative, or maybe we're not going to connect. If everyone is using these tools, doesn't that kind of raise the floor and make it harder to filter out people who you might not be a good match with?

I understand what you're saying, and our belief is that people really do have things to offer. Sometimes they just need a little bit of nudging and help to offer it in the right way.

I personally know lots of people who are super dynamic, interesting, fun people, and then I look at their profile, and they're like, “Can you help me with my profile?” And I'm like, “Wow, this completely misses you. You could do so much better here—choose better photos or just represent yourself.”

People often don't come across well in their dating profile, and that shouldn't necessarily be the barometer for whether you're going to have a good relationship with them or not.

Hmm.

Yeah. I guess to me, it feels a little bit like using AI to do the personality equivalent of airbrushing a photo. It's like, here's the AI that's going to make you seem a little more original, a little more creative, a little more charming than you maybe actually would be on your own. But maybe I'm—

But that's—

—being too cynical.

But that's the distinction between coaching and giving people the answers, right? We're not saying, “You write a prompt answer, and we'll say, ‘Hey, use this one instead.’” We're just saying, “Hey, could you say a little bit more about that?”

Yeah.

And I think that's just the little bit of extra nudging people need to still be in their own voice and still be authentically themselves, but just not be afraid to be a bit more vulnerable and a little bit more specific.

Yeah. To me, the interesting tension that I always felt like I was navigating when I was creating an online dating profile is that, on one hand, you do want to seem unique and quirky. You want to stand out from the pack, not be like all the other guys. On the other hand, you wanted to seem normal and approachable, like a recognizable kind of person that somebody might want to go and have a drink with.

I don't know if I ever figured out exactly how far to lean in either direction, and I'm sure probably both directions work depending on which person you want to wind up with. But I'm curious if you think AI is pushing in any particular direction. Is it always going to push us to be more unique, or are there moments when it's going to say, “It looks like you're confessing to a felony in this prompt. We may want to walk that back”?

I don't know about that latter one. Our goal is to bring out the uniqueness in people, and that's why we're, again, so conscious about asking you to bring your best self forward and not trying to give you the answer or write for you.

Yeah.

AI is a tool. It can be used to flatten, and it can be used to make everyone the same and tell everyone to put the same thing in the box—

Mm-hmm.

—or it can be used to really find out what's unique about you and bring that out. I think that is certainly the better formula for Hinge. People are unique, people are different, and we want to help them showcase that.

Yeah. Now, I want to ask you maybe a slight detour question, which is not specifically about AI but is about these romance scams that we've heard so much about. I get a lot of emails from listeners and readers talking about how one of the things that makes them very wary of online dating is the prevalence of these scammers who will pretend to be interested in you. They'll send you some flirty messages, maybe they'll even strike up a relationship with you, and then they'll tell you, “I can help you make money trading crypto,” or something. Then all of a sudden, your bank account is drained.

So is that an area of focus for you at Hinge? Are you seeing a lot of that activity, and is AI helping at all in protecting users from that?

Yeah, I mean, another big advantage of AI is our very big and truly dedicated trust and safety team. We don't see that as a big problem on Hinge because we really do catch people.

We can look at their IP address, prior photos, behavior on the app, behavior patterns, and all of that, and use AI models to catch people as quickly as possible.

I want to talk about the potential use of AI in the actual conversations. When I used dating apps, I never felt more like a bot on the internet than I did in the first 4 exchanges with anyone on a Hinge or a Tinder, right? It's like, “Hi.” “Hi, how are you?” “Good. How are you?” “I'm good. How was your weekend?” “It was fine. How was yours?” And you just repeat that ad infinitum.

I can imagine asking an AI bot to go do that on my behalf and get me to the interesting part of a conversation. I could also imagine somebody like you saying, “Well, I don't know if I actually want the AI to be sending the messages.” So how are you thinking about letting AI do more of the writing for your users over time?

Well, it's the principle of coaching, so we don't want to write for our users. One of the great things about Hinge is that you don't just like someone, they like you back, and then you're a match and go, because that does oftentimes lead to a generic exchange. On Hinge, you actually have to like something about them, and you can add a comment, and you have very rich profiles.

There's a lot of opportunity for us to help coach people to engage with a piece of content on someone's profile, whether it's a prompt or a photo. That's something else that we are exploring, and we actually plan to start testing next quarter: helping give people conversation starters. Again, not what to say, but, “Hey, did you notice this in their profile? Maybe this would be a good thing to ask them about.”

Mm-hmm.

I think that, once again, is just giving people the right nudges to move the conversation in the right direction so that you actually end up on a date.

Would you ever use AI to determine who is doing a lot of ghosting and say, “Hmm, seems like you're leaving a lot of messages unread. Is something going on with you, Kevin?”

Well, interestingly, I'm not sure we need AI for that. We just recently released Your Turn Limits, which was a feature precisely because of that. We looked at responsiveness on the app, and we saw that a lot of the unresponsiveness was because of just a few people who match with a lot of people and then don't respond.

Ooh.

So we introduced Your Turn Limits, which essentially means that if it's your turn to respond in a conversation, you can only collect so many of those. I think the number is about 8 before you're blocked from sending likes or matching with new people until you start responding to the people that you've already matched with.

Ooh.

Thank you. It was one of the few positive Reddit threads. We almost never see those. It was like, “Wow, this feature is really amazing.” We even had someone write a one-star review on the App Store and then come back and change it a little bit later to say, “Never mind. This feature is amazing.” And we saw responsiveness increase by about 20 to 30 percent.

That's excellent. Good job. That's good. I like that.

It speaks to something so real, though, which is that part of the fun of an app like Hinge is it gives you access to hundreds or even thousands of really cool people wherever you may live. But there's a sort of game-like nature to it, and it is fun to just collect matches and feel very attractive. If you're on the other side of that and you're a human being actually trying to date, nothing in the world is more frustrating.

That strikes me as a really good example of a feature that essentially just reminds the entire user base: You are human beings talking to other human beings, and you have to treat them with kindness and respect.

Yes.

I want to ask about the big buzzword of the year in AI, which is agents and agency. A lot of dating apps are planning, or at least talking about, releasing some kind of agentic feature. The founder of Bumble, Whitney Wolfe Herd, got a lot of attention for some comments she made last year, talking about how one day users might have an AI dating concierge who goes out onto an app for you and dates other users' AI concierges and says, basically, “Let's go on a date as proxies for these 2 real people and maybe see if they would be compatible.”

Grindr is also testing an AI wingman feature that will roll out in 2027 and offer suggestions like where to go on a date. Is that a direction that Hinge is thinking about? Are you at all excited about the use of AI agents to date on users' behalf?

That is not something that we're looking at. I think we just have this principle at Hinge that AI should really stand behind us and not between us, and that means that we are not in a world where you are being replaced by agents who are dating on your behalf. The journey of dating is something that I think we each need to take on our own. Having, again, the coaching and the nudging can be very, very helpful to people.

I also think, just practically and technically, we're not even close to a world where you could train an agent to be enough in your voice and understand your values and personalities and actually learn these things on your behalf.

Totally. I hear a story like that, and I'm like, “This is just science fiction that you're telling your shareholders so that your stock price goes up.”

Yeah.

Obviously, those tools don't exist. Maybe they would exist. Maybe some people would like them. Maybe you're close.

I mean, I've already seen people testing out the ChatGPT Operator, having it drive their online dating profile, so I don't think this is science fiction. Now, whether it's good or not is a different—

Yeah.

…question, but—

Whether it's going to lead to the result you want, I think, is the question, and we're trying to get people out on dates with people that they're actually going to like.

Justin, I wonder if you could give a bit of a pep talk to users of yours like Rachel, who are out there, using their AI prompts, not ghosting on their matches, and showing up for these dates. It hasn't happened for them yet. What have you learned from your years in love and as the CEO of a dating app about what actually gets people across the finish line to find their person?

For me, it was a shift in mentality. I think the biggest thing that I've learned about love, and my biggest change from when I started Hinge back in 2011 to where I am today, is the belief that there is the one, or that you find the one, versus that you create the one.

Mm.

I think a lot of us miss a one, maybe the one, because we have such a specific idea of what this has to be, and it's like we're shopping for the perfect person. When I created Hinge the first time, it was like, go through as many people as possible, and as soon as you find the one, then everything just works from there. The interesting experience that I had when I got my one back—the one I dreamed about—was the one that got away from college.

And then I flew over to Switzerland, and she moved back, and it was amazing for about 2 months. Then I was like, “Oh. Oh, my God, have I made a mistake?” There are flaws here. But we stuck with it, obviously, for 10 years now. And now we have just an amazing, deep, beautiful, incredible relationship.

I think this mentality that you have to start building and creating the one, and stop searching for the one. For most of us, it's a messier journey, and you choose someone, and then you're like, “Okay, this is it. Let's make this work together.”

Yeah.

That's really beautiful. I find myself asking the question, and I don't know that AI will ever be able to do this, but to me, a real question that I had when I was online dating was: Is this person really looking for a relationship? Are they ready to be in a relationship? Do they know what that means? I don't think there's any way to figure that out aside from just going out with them and getting a sense of it.

But, yeah, I think the difference between me now and me then is I am now dating somebody who wants to be in a relationship and knew that about himself. You got to get that. You got to get there.

Yeah. It's a journey for everyone.

I don't know if you know, but Casey has a boyfriend. He often—

He's very handsome.

…often brings it up.

It's a Valentine's Day episode. People want to hear about love.

And we are thrilled for Casey.

It's an inspiring story.

And his incredible boyfriend.

I'm thrilled for you, Casey.

Thank you. Thank you, Justin.

They did meet on Tinder, so I don't know that you should congratulate them. But—

But I was on Hinge. I was ready to meet him anywhere.

Yeah.

And if I can just say one more thing: don't give up, people. I was single. I went on so many online dates, and I thought, “This isn't working.” Then I went on one more online date, and it worked. Then I thought, “Online dating is fine.” So that was my journey.

Yeah.

It's hard out there, but don't give up.

And if Casey can find someone, you can too.

Amen.

All right. Thank you so much.

Thank you, Justin.

Thanks for having me.

Hard Fork is produced by Rachel Cohn and Whitney Jones. We're edited this week by Rachel Dry, and we're fact-checked by Caitlin Love. Today's show was engineered by Chris Wood. Original music by Marian Lizano, Rowan Nemeshtow, and Dan Powell. Our executive producer is Jen Poynt. Our audience editor is Nell Galogley. Video production by Sawyer Roquet, Pat Gunther, and Chris Schott. You can watch this whole episode on YouTube at youtube.com/hardfork. Special thanks to Paula Schumann, Pui Wing Tam, Dahlia Hadad, and Jeffrey Miranda. You can email us at hardfork@nytimes.com with what's on your dating profile.