Deep learning is the wrong path
There’s too much emphasis on deep learning in the field of artificial intelligence, according to NYU professor Gary Marcus. And he believes this skew could lead to its downfall.
Marcus says that while deep learning may be good at mimicking the perceptual tasks of the human brain, it isn’t good at tasks such as understanding conversations or causal relationships.
He argues that machine-learning is pretty good at learning from data, but very poor at abstraction. Classical AI, on the other hand, is good at abstraction, but needs to be hand-coded, and there is too much knowledge in the world to manually input everything.
“So it seems evident that what we want is some kind of synthesis that blends these approaches.”
He adds, “Right now people are trying to use kind of one-size-fits-all technologies to tackle things that are really fundamentally different. Understanding a sentence is fundamentally different from recognising an object. But people are trying to use deep learning to do both.”