The limits of artificial intelligence

Today I was trying to describe to someone my view on the limits of artificial intelligence, compared with the intelligence of the human brain, and I ended up drawing an image on the whiteboard to explain it.

I then went back to my computer and generated the image below, based on that sketch:

Basically, I see human intelligence as a kind of mountainous island in a sea of “how much human intelligence we can emulate with technology”.

As our technology gets better, the water level gradually rises, and we observe that the area of the island is gradually shrinking. This gives us a satisfying feeling that we are “solving” the problem of artificial intelligence.

Alas, only early successes are easy. The more we fill in the shallows, the more challenging things get, as we encounter the ever steeper slopes of human-like intelligence.

The real issue, as I understand it, is that center of the island — actual human intelligence — is vastly high, far taller than anything we are capable of doing with our current level of computation.

It is possible that many years from now we will manage to raise the sea level to the very top of the mountain. But I suspect that if this ever occurs, it will be so far in the future that nobody alive today will still be around when it happens.

Leave a Reply