Good morning,

This week, I’ve been preparing for a talk on Tuesday, where I will speak about what it might be like to live in a world we share with AI agents.

Will it be like our current relationships with animals? Some are wild, some are industrial, and some are seen as family members.

Or will it be more like our current relationships with other humans, which range from strangers and more transactional work relationships to unconditional friendships and family relationships? (The book The Courage to Be Disliked provides some good structure for thinking about human relationships.)

Maybe it will be something completely different?

My job is to explore the possibilities and open the minds of those who listen. However, the topic feels timely since Mark Zuckerberg has been severely mocked across the internet, partly for being incorrect about the average American's number of friends, but mostly because he believes that AI agents are a great solution to our lack of human relationships.

"There’s this stat that I always think is crazy,” Zuckerberg says in this viral clip. “The average American, I think has, I think it’s fewer than three friends. Three people that they consider friends. And the average person has demand for meaningfully more. I think it’s like 15 friends or something, right?”.

Unfortunately, Mark Zuckerberg leaves out the word “close”. 53% of Americans have between one and four close friends (according to Pew Research Center in 2023). 38% have five or more, while 8% state they have no close friends.

And yes, the lack of close friendships is an issue, not only in the US. I’ve written about that in this newsletter before. But the big picture question here is whether this statistic, misquoted or not, is a good argument when marketing a future of AI agents.

Zuckerberg might be authentically passionate about improving the friendships of US citizens. I don't know. But how many of us are longing for his solution to this problem?

Quite a few of us might even argue that closing down the social media monopoly he is currently running would probably do a lot more to improve our collective friendships than giving us AI friends.

To me, it feels like he burned down the house and is now trying to sell us a tent. “I know you feel lonely and the world falling apart, but look at this 90% human-like agent you can talk to about your feelings.”

Marc Zao-Sanders lists "therapy and companionship" as the top use case in a review for HBR. However, the deep dive into the reports in this category is heavily focused on therapeutic use cases, such as getting help with mental health issues or grief in countries where this is not easily accessible.

For therapy, I can see the point of AI agents complementing humans; it's a transactional relationship. But as a friend, someone to form an emotional attachment with, I wonder if anyone who could choose between an AI friend and a human friend would choose the AI friend. Do we not care about the other party in our relationship having reciprocal feelings?

I’m 100% sure that most humans care.

AI agents as friends are a terrible idea invented by someone with minimal ability to make human friends. And I’m worried because if we pretend that AI friends are as good as real human friends, we might never address the need for authentic physical social connections.

Physical is key. With digital relationships, so much of the information we exchange gets lost. We are biological creatures. We need to be in the same space as other humans. (Yes, even when they are weird and annoying).

And while I wouldn’t claim it’s a causal relationship, the more digitally connected age groups are also the ones that see the most drastic decline in both mental health and sexual activity. Data points that say something about the world we live in.

I’m worried that we as a society will buy the argument that AI relationships will solve loneliness. Suggesting we use it in nursing homes, for children with special needs or just as a solution to the fact that it’s complicated to have relationships with humans, because real people have needs.

Part of me is pleased that Mark Zuckerberg is mocked for his interpretation of the world. Part of me is concerned. Men like him have an awful lot of power over the digital experiences we've integrated into our daily lives.

What will the business model be for friends? Will they sell our data? Will they take a cut of product placements? Will we pay a subscription fee for access?

Time will tell. Maybe I should start charging my human friends for access to my friendship, or sell their data.

I’m very curious to know what you all think about AI friendships. Have you tried them? Are you curious? Do you enjoy them? I would love to know!

More From Me

1. Online Courses

I’m launching two online courses:

• Strategic Foresight for Beginners
• Practical AI Skills for Beginners All newsletter readers will get 50% off the fee for this first batch.

They will be six weeks each, taking about 2-3 hours weekly. They will be a mix of one-on-one sessions, group sessions, and pre-recorded assignments. Read more on Better Odds' new website.

I will be eternally grateful if you could help me spread the word and recommend them to friends and colleagues.

2. More writing

I wrote a Substack post for Better Odds about what people really feel about AI and the environment, their personal lives, and the need for basic math, reading and writing skills. You can read it here. (Some people are on both lists; most people are not.)

Recommendations

• PODCAST: Isabel Allende shares her life story in an episode of The Interview from The New York Times. She talks about having to leave a country you love and how she wrote her first book at 39.
• BOOK: Do you want to do something with your life? Michael Bungay Stanier's book How to Begin by How to Begin—Start Doing Something That Matters is a celebrated guide to getting started.
• REPORT: What do the people who work with the future believe about the world? This first report on "futurists" from Ipsos shares several interesting insights.

Thank you for today!

See you next week, Anna