3 ms·
Ask HN: Why is most GPT evaluation done by AI researchers and not psychologists?
Isn't intelligence testing the domain of psychology?
- AdieuToLogic 3y ago> Why is most GPT evaluation done by AI researchers and not psychologists? Because GPT is an algorithm which implements a predictive statistical model and does not "think".
- gregjor 3y agoPsychologists work with people, not with computer programs. Intelligence testing came from the field of psychology, but now mostly gets used by employers, the military, schools. AI researchers make claims about intelligence of their products so they can sell AI products.
- amichail 3y agoMaybe many psychologists will change their research focus soon? Also, maybe sociologists will soon consider social interactions between AIs as well as AIs and humans?
- gregjor 3y agoI don't expect AI experts will start offering anger management or prescribing Paxil. Different fields, no reason I can think of to start treating so-called AIs as conscious entities.
- JohnFen 3y ago> maybe sociologists will soon consider social interactions between AIs as well as AIs and humans? Wouldn't that be more appropriate for human-machine interface experts than sociologists?
- JohnFen 3y agoBecause psychologists study minds, and GPTthings aren't minds. And even if GPTthings are or become minds, you run against the next problem: psychologists study human minds. If fields analogous to psychology become useful for AI, they will have to be specialist fields because they certainly wouldn't be human and so would be outside the field of psychologist expertise. Much like psychologists don't study animal minds.
- cratermoon 3y ago> Isn't intelligence testing the domain of psychology? Yes and no. Binet and his colleagues who created the first intelligence tests were psychologists. But in the 120+ years that have passed since, psychologists have determined that intelligence testing is full of problems, and have mostly abandoned them in favor of targeted assessments.