Working Up

by Dwayne Phillips

Of course software running for the thousandth time works better than some people at some tasks. It has for half-a-dozen decades. Why does this continue to surprise us?

Here is a recent breath-taking story about how AI performs better than doctors at detecting a type of cancer. Of course it does. Put a digital camera on the front end and software running on a computer on the back end. It works better than a human.

The simple reason is the software doesn’t have bad days. The software isn’t tired after a restless night caused by a sick child or worry of a cancer-ridden relative. The software wasn’t in a car accident on the way to work or didn’t have to circle the parking lot for half-an-hour looking for a space.

It has been this way for 50 or 60 years or more. Basic classification algorithms existed long ago and ran on what we would consider to be archaic computers with almost no compute power and memory.

We had these discussions in the 1980s. Expert systems performed better than people. Of course they did. The were not, however, 100% correct. Hospitals stayed away from systems known to be less than 100% because they didn’t want to be sued. Other industries stayed away for the same reasons.

Today, the systems are not 100%, but we accept that now for one reason or another.

Today, we call it “AI” and run supercomputers that slip into our pants pockets. We gasp at the performance. The end of the world is nigh. Well, at least the end of someone’s career is nigh. Old news. Let’s use what we have and move on to some other problem that needs solving.