The ultimate Turing test is, if it looks like a human and smells like a human and moves like a human and acts like a human, then it’s human.
If your test only has human and machine input, say, chess moves or answers to basic math questions, even a very basic AI will be indistinguishable. If your test has common-knowledge questions and social responses than an AI like ChatGPT may be indistinguishable.
But if your test involves physically looking at the subjects and having them do routine human things like eat, sleep, talk to others…you’ll be able to recognize nearly instantly. Because the best Androids today only reach the uncanny valley; technology isn’t even close to replicating the physical aspect, especially movement, durability, and metabolism, and likely won’t be for a while.
Even if you’re doing the test on FaceTime, current CG models struggle to imitate real actors and AFAIK cannot do so real-time. We have deepfakes, but they are mainly good for typical and prepared positions (e.g. face directly in front of camera), and (like image generation) are still subject to artifacts. Although this may not be true in the future, biology will be a lot harder to replicate.
1 comment
[ 2.4 ms ] story [ 9.6 ms ] threadIf your test only has human and machine input, say, chess moves or answers to basic math questions, even a very basic AI will be indistinguishable. If your test has common-knowledge questions and social responses than an AI like ChatGPT may be indistinguishable.
But if your test involves physically looking at the subjects and having them do routine human things like eat, sleep, talk to others…you’ll be able to recognize nearly instantly. Because the best Androids today only reach the uncanny valley; technology isn’t even close to replicating the physical aspect, especially movement, durability, and metabolism, and likely won’t be for a while.
Even if you’re doing the test on FaceTime, current CG models struggle to imitate real actors and AFAIK cannot do so real-time. We have deepfakes, but they are mainly good for typical and prepared positions (e.g. face directly in front of camera), and (like image generation) are still subject to artifacts. Although this may not be true in the future, biology will be a lot harder to replicate.