3 ms·
I worked with some people who were actively working on this last year, focusing on CS research. The biggest issue was validation. We could get a system to spit
by agubelu 2y ago
I worked with some people who were actively working on this last year, focusing on CS research.
The biggest issue was validation. We could get a system to spit out possible research directions automatically, but who decides if they're reasonable and/or promising? A human, of course. Moreover, we gave different humans the same set of hypotheses to validate and they came back with wildly different annotations.
- acureau 2y agoThis was exactly what I was thinking. This may be a useful tool for human researchers to drive if it turns out it can generate anything valuable. I can't understand the papers it wrote, much less determine if they make any contributions. I don't think the self-evaluation is going to be too fruitful. Thanks for sharing