Why Marcus Is Wrong About AI

My friend Marcus Hutchins put out a long, well-written, and entertaining piece about all the reasons he thinks AI is hype.

I think it was very good, and I think he's very wrong.

I will probably do another video on this because I consider this to be such an important issue, and I worry his quality writing will sway many people to be complacent on the issue. But really it all comes down to one thing.

He's using a shitty definition of intelligence.

His definition is basically novel discovery. Like, net-new invention. Unless I misunderstood him, he doesn't believe everyday cardiology or marriage counseling counts as intelligence. So like two of the most skill-heavy and intellectual jobs in the world. As in, an f-ing Cardiologist. An M.D. Or a Psychiatrist. Etc. In the course of their normal patient work.

Why? Because it's just patterns. It's just using their training.

You meet people with the same problems, in slightly different configurations, and you consult your training, and you give out pretty much the same advice as you gave the last 120 people with similar issues. Now extend that to the rest of everyday knowledge work. It's not real intelligence because they didn't do anything new. Anything novel.My paraphrasing of his arguments in our debate

Cool story, except that definitionally devalues 99% of all knowledge work done on the planet everyday.

And we can intuitively see he's wrong here because of one glaring fact: the work hasn't already been automated for decades. If it were so easy to just pattern match, none of these knowledge workers would even have jobs. The work would have been replaced by automation decades ago.

It's not automated because it requires intelligence.

What's intelligence? Hard to pin down, but my definition is something like the ability to take a new, everyday problem and apply your knowledge and understanding of the world to come up with a useful solution.

Scripts can't do that. Programs can't do that. Only humans can, and that's why the entire field of knowledge work exists. You might think doing customer support is easy—or being an administrative assistant—but millions of people are paid real money to do these jobs every day.

It's real-world, dynamic problem solving using human brains and human intelligence. And it's valuable. If it weren't, companies wouldn't be paying billions of dollars a year to have people do it.

This is also a dead-simple explanation for why AI is so disruptive—it's the first tech ever invented that can do something like what we do.

Marcus is wrong about this because he's defining intelligence in a way that doesn't matter in everyday life.

And as a result, he's convincing large numbers of people to ignore something that they should absolutely be paying attention to.