The Mutual Friend

An AI companion app; paid subscribers can give me feedback Monday evening.

As you know, I have been trying Pi (Personal Intelligence). It is a chatbot that tries to be your companion in conversation.

I also saw Anders L’s essay that ends with a claim that AI would be useful in a dating app. So, here is what I came up with. It’s not a dating app, but sort of like one.

The hypothesis is that getting introduced by a mutual friend is more relaxed and more likely to prove successful than trying to meet someone on a dating app. The proposed app is a social network where the AI (The Mutual Friend) introduces people who TMF thinks could be good friends.

It does not have to lead to sex. Or even to romance. You might be introduced to someone of the gender that you’re not attracted to, with a goal of friendship. Or you might be introduced to someone you eventually decide to marry. The idea is to meet someone who you will enjoy being around for a long time.

You don’t get to fill out a profile. Profiles tend to be pretty shallow in terms of content. Even if the person is not lying, it’s hard to get an idea of what someone is really like from a profile.

Instead, you have a conversation with TMF. The conversation would be open-ended. TMF might start by asking you to describe some of your friends and why you like them. It would keep the conversation rolling, getting an idea of what you are like. Your interests and values would be revealed in the conversations.

TMF would have to be trained on a set of happy couples. It would interview the individuals in those couples, in order to acquire a sense of what makes for compatibility. Once people start using TMF, it would also learn from experience.

TMF would need a large network of subscribers. I suppose an incumbent like Facebook would have an advantage in launching TMF, but let’s think of it as a start-up.

As a start-up, TMF would have to begin the way that Facebook did, with a specific niche. When I discussed the idea with Pi, it suggested targeting young professionals in Austin, Texas. If TMF works in a pilot mode, then word of mouth would enable it to spread.

Would people try to game TMF, engaging it in deceptive conversations? I’m sure that would happen eventually, but at the outset no one would really know how it was matching people. And by the time people figure out how it’s working, TMF may be able to spot deception.

Comments welcome. This can be a topic for the paid subscriber discussion next Monday evening.