4 ms·

Or we could dispense with the myth that the Turing Test has ever been any kind of "goal post" to begin with. It's like Schrodinger's Cat, it was a pretty minor

by gipp 3y ago

Or we could dispense with the myth that the Turing Test has ever been any kind of "goal post" to begin with. It's like Schrodinger's Cat, it was a pretty minor and unprincipled thought experiment, that a major thinker in the field made in a fairly informal manner. The public perception of its importance and its level of prominence in pop culture around the topic has always far exceeded its actual importance.

mcphage 3y ago
> Or we could dispense with the myth that the Turing Test has ever been any kind of "goal post" to begin with. Well, it definitely was a goal post for many years—pretending it wasn't doesn't seem like a good idea. On the other hand, now that we have chatbots that can pass it, its flaws as a test have become apparent. And as you say, it's informal, and underspecified in a way that wasn't clear when it was defined. But I also don't think it's much of a surprise that "passing" it doesn't actually mean what Turing thought it would mean back in 1950. We didn't have a good idea of what parts of intelligence are "easy" versus what parts are "hard".
- gipp 3y ago
  > Well, it definitely was a goal post for many years—pretending it wasn't doesn't seem like a good idea. In the sense that people treated it as one, sure. Not in the sense that it would ever have actually been any kind of important threshold for "reaching AGI". A goalpost that's based mainly on popular perceptions from bad sci-fi probably should be moved. "Moving the goalposts" just doesn't work as an accusation here, is my point.
- hnfong 3y ago
  The Turing Test was one of the goal posts for establishing intelligence, I don't think it was one for establishing humanity. I mean, at least some murderers are intelligent. But without an acceptable moral compass, does intelligence alone suffice to qualify them as "fellow humans"? I would say no.

by gipp 3y ago

mcphage 3y ago

> Or we could dispense with the myth that the Turing Test has ever been any kind of "goal post" to begin with. Well, it definitely was a goal post for many years—pretending it wasn't doesn't seem like a good idea. On the other hand, now that we have chatbots that can pass it, its flaws as a test have become apparent. And as you say, it's informal, and underspecified in a way that wasn't clear when it was defined. But I also don't think it's much of a surprise that "passing" it doesn't actually mean what Turing thought it would mean back in 1950. We didn't have a good idea of what parts of intelligence are "easy" versus what parts are "hard".

gipp 3y ago
> Well, it definitely was a goal post for many years—pretending it wasn't doesn't seem like a good idea. In the sense that people treated it as one, sure. Not in the sense that it would ever have actually been any kind of important threshold for "reaching AGI". A goalpost that's based mainly on popular perceptions from bad sci-fi probably should be moved. "Moving the goalposts" just doesn't work as an accusation here, is my point.
hnfong 3y ago
The Turing Test was one of the goal posts for establishing intelligence, I don't think it was one for establishing humanity. I mean, at least some murderers are intelligent. But without an acceptable moral compass, does intelligence alone suffice to qualify them as "fellow humans"? I would say no.