″...to understand the conditions under which AI will develop the same abstractions as humans do...”
I know from observation that ChatGPT has some knowledge of the concepts of justice and charity. It can define them in a reasonable way and create stories illustrating them. In some sense, it understands those concepts, and it arrived at them, I presume, through standard pretraining. Has it therefore developed those abstractions in the sense you’re talking about?
I know from observation that ChatGPT has some knowledge of the concepts of justice and charity. It can define them in a reasonable way and create stories illustrating them. In some sense, it understands those concepts, and it arrived at them, I presume, through standard pretraining. Has it therefore developed those abstractions in the sense you’re talking about?