The idea of taking inspiration from how the brain works for designing algorithms has been a fruitful endeavor across domains like cybernetic, artificial intelligence, and cognitive science. However, recent achievements in deep learning has provided some surprising counterevidence, where adopting strategies that are different from those adopted by the brain is successful. We review here the cases of learning rules and vision processing. We suggest two possible justifications of these evidences. It might be that our knowledge of how a problem is solved by the brain is incomplete or lackluster. Therefore, we are not able to translate the genuine brain solution to this problems into the proper algorithm. Or, it might be that the algorithmic solution applied by the brain to a problem is not the most effective for digital computers. Note that the two possibilities are not necessarily mutually exclusive.
File in questo prodotto:
Non ci sono file associati a questo prodotto.