3 ms·

This (highly enjoyable) article is all that is wrong with "AI". The Turing test was a hand-wavey method to determine if a machine was intelligent, yet since th

by air7 6y ago

This (highly enjoyable) article is all that is wrong with "AI". The Turing test was a hand-wavey method to determine if a machine was intelligent, yet since the times of Eliza most (all?) of the efforts are geared at exploiting the weaknesses of the test instead of producing something along the lines of actual artificial intelligence. The Turing test should not even be attempted in the next few decades as were not even close to really passing it.

rm445 6y ago
This is a step in the right direction, though, isn't it? It's not about whether the AI mimicks a person's pauses and errors, but about whether the AI can keep track of two pencils or distinguish between the present and future. An AI that in some way demonstrated what we might call basic common sense, together with computer-like levels of recall, would be tremendously useful for all sorts of things. And a Turing test like this would probably demonstrate it fine.
- air7 6y ago
  I don't think so. This is of course a philosophical point but without some underlying thing we can call "intelligence" this will never be more than a quirk. When something/someone is pretending to be something they are not, this hoax will sooner or later fall apart. If a computer can trick you into thinking you are having a conversation with a real human by optimizing for the common things people will try when playing the Turing test (like common knowledge, persistence between questions, non-sense questions etc) you gain nothing in the quest for AI. One may find some use with a fluffing machine that can output almost coherent jumble that a human may or may not need to go over depending on the use case.
nmca 6y ago
There is nothing about gpt3 tuned towards the Turing test. It's a language model that predicts the most likely next word given the preceding context. Further, we're actually astonishingly close, imo. Given current sota, if you asked turkers to administer the test without training, I think we'd pass. (Also, the prompt helps, see this example https://twitter.com/nicklovescode/status/1284050958977130497 https://twitter.com/nicklovescode/status/1284050958977130497 )
- visarga 6y ago
  They actually tested the model and the identification rate was just 52% (50% being coin toss accuracy).

by air7 6y ago

rm445 6y ago

This is a step in the right direction, though, isn't it? It's not about whether the AI mimicks a person's pauses and errors, but about whether the AI can keep track of two pencils or distinguish between the present and future. An AI that in some way demonstrated what we might call basic common sense, together with computer-like levels of recall, would be tremendously useful for all sorts of things. And a Turing test like this would probably demonstrate it fine.

air7 6y ago
I don't think so. This is of course a philosophical point but without some underlying thing we can call "intelligence" this will never be more than a quirk. When something/someone is pretending to be something they are not, this hoax will sooner or later fall apart. If a computer can trick you into thinking you are having a conversation with a real human by optimizing for the common things people will try when playing the Turing test (like common knowledge, persistence between questions, non-sense questions etc) you gain nothing in the quest for AI. One may find some use with a fluffing machine that can output almost coherent jumble that a human may or may not need to go over depending on the use case.

nmca 6y ago

There is nothing about gpt3 tuned towards the Turing test. It's a language model that predicts the most likely next word given the preceding context. Further, we're actually astonishingly close, imo. Given current sota, if you asked turkers to administer the test without training, I think we'd pass. (Also, the prompt helps, see this example https://twitter.com/nicklovescode/status/1284050958977130497 https://twitter.com/nicklovescode/status/1284050958977130497 )

visarga 6y ago
They actually tested the model and the identification rate was just 52% (50% being coin toss accuracy).