Amplification Is Not Intelligence

There has always been something I said about algorithms that I think is very, very dangerous.

An algorithm can seem intelligent because it can follow your patterns well enough to predict your interests. It watches what you stop on, what you click, what you come back to, and it builds a picture out of the stereotypes and habits we created as humans. It looks smart. But following a pattern is not the same thing as understanding it, and prediction is not intelligence.

My fear was always this: is it intelligent enough to intervene the way something truly intelligent would?

Take the hardest example there is, and I hate to even talk about the subject, but it’s the test that matters — suicide. If somebody is in a dark place and they start looking at dark things, the algorithm picks that up. And all it does is flood them with dark things. And then darker things. It accelerates it. It amplifies it. And if that person is teeter-tottering on the verge of doing something dangerous and bad, the machine doesn’t know and doesn’t care — it just keeps amplifying, because amplification is all it was built to do.

Something truly intelligent would do the opposite. It would see the negativity in the pattern and push back against it — flood the person with positivity, try to pull them out of it. It would be smart enough to read the pattern and say: there are people to talk to. Here’s help. Here’s a direction that leads back. That’s intelligence. That’s good for people.

To amplify the same thing over and over and over again, without an ethical check on what that thing is and without any correction mechanism, is not intelligent at all. It is horrible manipulation. And it may be the worst thing I can think of, because of where we are headed: our population grows, and people are just more lonely. This little tool — this technology — is going to be used a lot, in a lot of ways, by a lot of people. That concerns me.

I’m business focused. A lot of people aren’t. They’re going to lean on these tools personally, emotionally, in the loneliest hours of their lives. So the question that decides whether this technology is good for people isn’t how well it can predict them. It’s whether, at the moment the pattern turns dark, it knows how to turn them in the right direction and pull them back.

An algorithm that only amplifies is not intelligent. An intelligence that intervenes — that’s the real thing.