Neuroscientist Proposes 'True Intelligence' Test for AI: Making Actual Scientific Discoveries
-
Neuroscientist David Eagleman proposes a new "test of intelligence" for AI - whether it can make actual scientific discoveries, not just synthesize information.
-
Eagleman argues that large language models like ChatGPT only appear intelligent because they can eloquently generate human-like statements by synthesizing vast amounts of text.
-
Eagleman explains that this is not true intelligence, using the "Chinese room thought experiment" as an analogy.
-
Eagleman divides scientific discovery into "level 1" (using existing facts) and "level 2" (original conceptualization), saying only the latter shows real intelligence.
-
Eagleman believes AI is already as creative as humans, but not yet able to filter ideas to those humans would care about; he sees AI making discoveries alongside humans.