I’ve been surprised by how bad the majority of scholarship is around the “inspired-by” or “metaphorical” genre of algorithms
Heh. Ultimately, all of AI is in this genre. A particularly bad aspect of this is that it usually means people choose their research problems based on what they think their approach can solve.
I’ll briefly plug my proposal for AI-related inquiries, that explicitly rejects the “metaphorical” approach by actually defining the question before finding the solution. At the same time it doesn’t rule out NNs, GAs, etc, but requires hard proof, in the form of an ungameable compression score, of their quality.
Modern AI is an odd combination of statistics, applied math, discrete math/combinatorics and logic. My theory is the only reason AI is a subfield of Computer Science at all is founder bias (Turing).
Totally agree. My slogan is that for AI to succeed it has to become an empirical science: it should use math, but only to the extent that the math is useful to describe reality. And it should be curiousity-driven and not application-driven like almost all modern research in computer vision and natural language processing.
Heh. Ultimately, all of AI is in this genre. A particularly bad aspect of this is that it usually means people choose their research problems based on what they think their approach can solve.
I’ll briefly plug my proposal for AI-related inquiries, that explicitly rejects the “metaphorical” approach by actually defining the question before finding the solution. At the same time it doesn’t rule out NNs, GAs, etc, but requires hard proof, in the form of an ungameable compression score, of their quality.
Modern AI is an odd combination of statistics, applied math, discrete math/combinatorics and logic. My theory is the only reason AI is a subfield of Computer Science at all is founder bias (Turing).
Totally agree. My slogan is that for AI to succeed it has to become an empirical science: it should use math, but only to the extent that the math is useful to describe reality. And it should be curiousity-driven and not application-driven like almost all modern research in computer vision and natural language processing.
I’m curious about how to not die (as an individual and a species). Would that count as curiosity or application drive? :P