G.D.P. is not a measure of how much value is produced for consumers. Everybody should recognize that G.D.P. is not a welfare metric.

For a long time, the humans are going to be better than the machines and so different parts of the job will be leveraged. In a way that's happened for centuries, and we've adapted. And it's made the people who had parts of their jobs automated more valuable and more productive to the extent that they are essential for the other components of their jobs.

The heart of science is measurement.

It's most useful to think about not jobs but tasks. And within any given job, there are lots of different tasks. If you're a radiologist maybe reading the images machines can be able to do that better, maybe making the broader diagnosis and communicating it to the patients.

The economy in the next 20 to 25 years is going to change more than they did in the last 20, 25 years. And that's because exponential trends are affecting a bigger and bigger share of the economy. So we have some huge disruptions in store, and I can't predict exactly what the innovations are going to be. If I did, I would have already invented them. But I think they'll be comparable to the innovations we saw in the past 20, 25 years if not greater.

Computers get better, faster than anything else ever. A child's PlayStation today is more powerful than a military supercomputer from 1996. But our brains are wired for a linear world. As a result, exponential trends take us by surprise. I used to teach my students that there are some things, you know, computers just aren't good at like driving a car through traffic.

Technology has made it easier for different firms to coordinate their activities with one another, and they don't have to be part of one company. They can get the benefits of scale without the inertia of scale.