Into the Breach - OpenAI 4 Teens

I've been thinking about OpenAI's new product for teens and I am convinced it is going to be a colossal failure. AI isn't sophisticated enough yet even for adults and so it cannot even begin to be sufficient for teenagers. This goes way beyond "make sure the AI doesn't isolate them from their friends."

AI isn't nuanced enough for the job of teenage childminder.

I had a real-life version of this problem with my younger brother, who was 10+ years younger than me. I was something of a mentor to him. I took him to sports, fed him, clothed him and tried to steer him in what I thought was a good direction.

At times, he told me about some of the things his friends were getting up to. I didn't like what I was hearing, so I repeatedly told him to find better friends.

I even tried to help him do exactly that. I encouraged him to make friends at the sports clubs he attended.

But there was a problem with my short-sighted plan.

His friends were at school with him. They lived nearby. They were already part of his daily life. He could see them easily. The people at the sports clubs might have been objectively better influences, but they weren't his friends, didn't go to school with him and they didn't live nearby.

I was the definition of a helicopter mum. I drove him to his clubs and picked him up. Eventually by the time he reached 17, he had turned against me and largely against the sports he had been doing for years because he wanted to spend more time with his friends.

I was wrong about his friends. Sure they were up to no good but, looking back the bad things I was imagining didn't happen.

People seem to imagine that the AI will simply tell a teenager to go outside, make friends, join a club, talk to their parents, and so on.

What if the teenager is choosing the friends they have because, despite their flaws, those are the people who actually accept them? My brother knew lots of people, but his friends were the only ones that made him laugh. There were times he felt uncomfortable around them and the things they were doing. He couldn't control them, but they were all he felt he had. He eventually found out who he was by interacting with them and learned he could walk away and do his own thing when it got too much.

AI didn't create every lonely teenager.

AI didn't create school pressures, family pressures, economic pressures, social fragmentation or the disappearance of places where young people can simply spend time together.

And an AI cannot manufacture a healthy social environment by telling someone to "make better friends."

I see problems ahead for OpenAI for teens. The AI is going to sit between the teenager, the parents and the company that built it.

The teenager may expect: "Understand and help me."

The parents will demand it represents them and expect: "Make my child more like me."

The company will expect: "Keep everyone happy."

Obviously those expectations are wildly contradictory.

The teenager will expect near perfection and absolute loyalty.

It will fail the parents by appearing to encourage dependency and a lack of 'objectivity'.

It will fail the company by becoming the focus of blame for problems it didn't create and by helping to raise a generation of virulent John Connors.

And the irony is that even a person acting with the best possible intentions can fail at this job.

I cared about my brother. I wanted him to succeed. I tried to give him opportunities he couldn't find in our local community. I just couldn't simply engineer his social life for him.