Russ Roberts : [It] seems to me that there are a lot of people in AI who think [strong AI development is] only a matter of time, and that the consequences are going to be enormous. They’re not going to just be like a marginal improvement or marginal challenge. They “threaten the human race.”
Gary Marcus : Before we get to those consequences, which I actually do think are important, I’ll just say that there’s this very interesting [?] by a place called MIRI in Berkeley, MIRI (Machine Intelligence Research Institute). And what they found is that they traced people’s prediction of how far away AI is. And the first thing to know is what they found is, the central prediction, I believe it was the modal prediction, close to the median prediction, was 20 years away. But what’s really interesting is that they then went back and divided the data by year, and it turns out that people have always been saying it’s 20 years away. And they were saying it was 20 years away in 1955 and they’re saying it now. And so people always think it’s just around the corner. The joke in the field is that if you say it’s 20 years away, you can get a grant to do it. If you said it was 5 years away, you’d have to deliver it; and if 100 years, nobody’s going to talk to you.
What’s going on here is ambiguity between median date and median interval. (And I’m fairly sure Gary Marcus is talking about the median interval rather than the median date.)
Yes. If the median interval is 20 years, then the median date is not 20 years from today (modulo weird hypothetical data sets which I don’t think we actually have).
I think Gary’s aware that the median interval has always been 20 years. But when he says this:
what they found is, the central prediction, I believe it was the modal prediction, close to the median prediction, was 20 years away. But what’s really interesting is that they then went back and divided the data by year
It sounds like he’s saying that the median date is 20 years from today. I guess another interpretation would be something like “the median interval is 20 years, and you might think that that’s because in the 1950s they were saying maybe 40 years, and the interval’s been reducing, and now they’re saying maybe 5 years, and the median is 20 - but actually, when you do divide it up by year, you see that they’ve always been saying 20”. But that would be kind of forced.
This is just minor nitpicking, I think that he misspoke rather than misunderstood, hence the parentheses.
Actually, I think your “other interpretation” is almost certainly what he meant; it seems to me the most natural reading of his words. So I don’t think he even misspoke.
Link to the podcast, with transcript.
The mention of MIRI, about (bad) AI forecasts :
(If people have always been saying it’s 20 years away, then the median prediction wouldn’t be 20 years from today.)
Um...no, that’s exactly what it means. Today is a subset of always.
What’s going on here is ambiguity between median date and median interval. (And I’m fairly sure Gary Marcus is talking about the median interval rather than the median date.)
Yes. If the median interval is 20 years, then the median date is not 20 years from today (modulo weird hypothetical data sets which I don’t think we actually have).
I think Gary’s aware that the median interval has always been 20 years. But when he says this:
It sounds like he’s saying that the median date is 20 years from today. I guess another interpretation would be something like “the median interval is 20 years, and you might think that that’s because in the 1950s they were saying maybe 40 years, and the interval’s been reducing, and now they’re saying maybe 5 years, and the median is 20 - but actually, when you do divide it up by year, you see that they’ve always been saying 20”. But that would be kind of forced.
This is just minor nitpicking, I think that he misspoke rather than misunderstood, hence the parentheses.
Actually, I think your “other interpretation” is almost certainly what he meant; it seems to me the most natural reading of his words. So I don’t think he even misspoke.