AGI means expertise across fields, not a Turing test

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

1. 13:05

   yeah maybe to pull on this thread ray first of all it's wonderful to be chatting with you again always enjoy our conversations the turing test i've argued on this podcast in past that the Turing test went by with a whimper, not a bang. It flew by. The Loebner prize was cancelled before the Turing test was arguably passed, and yet it was passed, and there was no celebration. The Loebner test was not a really good test. He had various practices that were really not in accord with the Turing test.

2. 13:42

   And Turing test is really matching an ordinary person that's talking, not really an expert in the field. AGI thinks actually a better view because we're actually matching the best person in each field. And we have maybe several thousand, maybe several hundred thousand fields that you could be expert in. And AGI means that you can match a human being in any of the fields and then combine the insight into many different fields together, which no human being can do.

3. 14:21

   I mean, Einstein was very good at physics, and he actually was interested in playing a violin, but he was not an expert in playing a violin. He was only an expert in physics. People maybe can master two fields at the most, but there's actually thousands of fields and if you could actually be an expert in all of them and then combine all those insights that's something that's quite unique so that's what AGI represents whereas Turing test is really matching an ordinary person with a lot of mischaracterizations of different things. I agree that AGI and passing the Turing test are, for most common definitions, different standards. The question I was going to ask, though, is, arguably, if you agree with the premise that the Turing test, as reasonably defined, not the original gender presentation-based

4. 15:18

   Turing test, but the subsequent definition, was passed without very much hoopla at all, Do you think the same is going to happen with the singularity? There's, in particular, one of my favorite scenes in Charlie Strauss's novel, Accelerando. You have a bunch of characters who've been all uploaded to a Starwisp traveling to another star system who are all arguing with each other. They're post-human uploads arguing with each other as to whether the singularity has even happened.

5. 15:48

   Do you think that's what's actually going to happen here, where we'll just singularity will zoom by and we'll all be arguing with each other decades later. Did the singularity even happen? Has it happened yet? I mean, these standards are not very clear. Not everybody agrees that we've passed the Turing test. And when we pass AGI, there'll be disagreements. It's disagreements now as to what that means. People say it's basically as good as somebody who's a little bit above average intelligence.

6. 16:20

   I define it as being an expert in every area when there's many different areas that you can be expert in. So that's actually quite an impressive level. And I think we'll get there by 2029. The thing that's, then you can combine your insights into every possible field. Really, I mean, have that large language models can answer questions in lots of different fields. No person can do what a large language model can do today, let alone what will happen by 2029.

7. 16:55

   By the way, we have a moonshots test where you have to fool your spouse for three minutes on a Zoom call.