ProHuman: Governing the Global AI Revolution
Jul 15, 2026 · 42 min · 10 segments
For years, artificial intelligence safety research has focused on accidental risks and unintended consequences. But what if AI models were demonstrating risks that weren’t accidental at all? What if…
When you were first writing that book in 1995, is this what you imagined the future would look like?
So I was certainly expecting that AI capabilities would improve drastically from where we were in the 1990s.
I didn't expect them to take the direction that they have, where we're essentially training gigantic multi-trillion parameter black boxes from vast amounts of text data, and that seems to produce something that's quite intelligent.
The general direction we were going was gradually understanding the components of intelligence, how one reasons, how one makes decisions, how one learns, and then trying to figure out how to fit those components together to make a system.
And that's a very traditional engineering approach, the same way that we go about building airplanes and nuclear power stations and cars and all kinds of other things.
And we've abandoned that almost entirely in favor of the giant black box approach.
The timeline, however, is, I would say, somewhat consistent with what I was expecting and to some extent what Alan Turing predicted back in 1950 when he said...
I think he was maybe a little bit optimistic saying that by the year 2000, uh, we might expect a system to be able to pass the Turing test for five minutes, meaning to imitate a human well enough that you couldn't tell the difference.
And so tho- those are things that I think we were expecting on roughly this timeline.
When you were first writing that book in 1995, is this what you imagined the future would look like?
So I was certainly expecting that AI capabilities would improve drastically from where we were in the 1990s.
I didn't expect them to take the direction that they have, where we're essentially training gigantic multi-trillion parameter black boxes from vast amounts of text data, and that seems to produce something that's quite intelligent.
The general direction we were going was gradually understanding the components of intelligence, how one reasons, how one makes decisions, how one learns, and then trying to figure out how to fit those components together to make a system.
And that's a very traditional engineering approach, the same way that we go about building airplanes and nuclear power stations and cars and all kinds of other things.
And we've abandoned that almost entirely in favor of the giant black box approach.
The timeline, however, is, I would say, somewhat consistent with what I was expecting and to some extent what Alan Turing predicted back in 1950 when he said...
I think he was maybe a little bit optimistic saying that by the year 2000, uh, we might expect a system to be able to pass the Turing test for five minutes, meaning to imitate a human well enough that you couldn't tell the difference.
And so tho- those are things that I think we were expecting on roughly this timeline.
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
Search every transcript — by keyword, by phrase, or by meaning, across every show Radar indexes
Trends — what is surging across podcasts, measured against its own baseline
Alerts — when a name you follow appears in a newly indexed episode
No account is needed to search Radar.