It’s a simple heuristic based on an 80 year old party game. You can argue it is scientific as an empirical methodology, but it isn’t objective in analysis.
Or believed that people are rational enough to be swayed by facts and logic.
I don’t know how you get a “fact” out of the imitation game. If anything, the game exposes the subjectivity of the subject being analyzed. You can apply logic based on certain axioms, but what are the axioms upon which the definition of “thinking” (or “gender”) are built?
The former, at the very least, is a complex philosophical snarl that could have a tangible answer. The latter is a muddled interpretation of biological sex and social norms, with the social norms taking much higher precedence.
But a “pass” on either one is ultimately rooted in the savvy of the listener not the objective reality of the speakers. Talking about facts and logic in the imitation game is like talking about facts and logic in a poker game. At some point, you’re just going to have to guess based on incomplete information. That doesn’t mean a bluff is the same as a winning hand.
I wasn’t clear enough. Turing was wondering if machines can think. But there is no sufficiently clear definition of the word “thinking” that could be used to answer the question.
If you want to know if LLMs are AI, you can just look up the definition of AI and check if LLMs meet the criteria. You cannot do that to answer if they are thinking.
So let’s take a task, which we agree takes thinking, and see if a machine can do it as well as a human. If it can, then the machine must be able to think. That’s how you think as a scientist.
The test itself is similar to modern, placebo-controlled medical trials. That was not SOTA at the time, showing how clear thinking he was. Perhaps the WP entry on RCTs helps to understand how logic and reason may be applied in the face of uncertainty.
But of course the test revolves around the definition of a word. Such definitions are fundamentally arbitrary. That means that the test itself is arbitrary. Science is rarely concerned with colloquial definitions. Usually you come up with some sort of operational definition that you use for the purpose of inquiry. The only question is, if that definition is useful.
It’s a simple heuristic based on an 80 year old party game. You can argue it is scientific as an empirical methodology, but it isn’t objective in analysis.
I don’t know how you get a “fact” out of the imitation game. If anything, the game exposes the subjectivity of the subject being analyzed. You can apply logic based on certain axioms, but what are the axioms upon which the definition of “thinking” (or “gender”) are built?
The former, at the very least, is a complex philosophical snarl that could have a tangible answer. The latter is a muddled interpretation of biological sex and social norms, with the social norms taking much higher precedence.
But a “pass” on either one is ultimately rooted in the savvy of the listener not the objective reality of the speakers. Talking about facts and logic in the imitation game is like talking about facts and logic in a poker game. At some point, you’re just going to have to guess based on incomplete information. That doesn’t mean a bluff is the same as a winning hand.
I wasn’t clear enough. Turing was wondering if machines can think. But there is no sufficiently clear definition of the word “thinking” that could be used to answer the question.
If you want to know if LLMs are AI, you can just look up the definition of AI and check if LLMs meet the criteria. You cannot do that to answer if they are thinking.
So let’s take a task, which we agree takes thinking, and see if a machine can do it as well as a human. If it can, then the machine must be able to think. That’s how you think as a scientist.
The test itself is similar to modern, placebo-controlled medical trials. That was not SOTA at the time, showing how clear thinking he was. Perhaps the WP entry on RCTs helps to understand how logic and reason may be applied in the face of uncertainty.
But of course the test revolves around the definition of a word. Such definitions are fundamentally arbitrary. That means that the test itself is arbitrary. Science is rarely concerned with colloquial definitions. Usually you come up with some sort of operational definition that you use for the purpose of inquiry. The only question is, if that definition is useful.