Hacker News Reader

Comment by nemo1618 | Hacker News Reader

I think abandonment of the Von Neumann paradigm will be crucial for AI. For decades we have labored under the misguided assumption that the brain is isomorphic to the Von Neumann architecture, with "memory banks" and "processing units." This approach has failed.

Admittedly, Lisp was once the first choice for AI research, and it did not readily lead us into the promised land either. Perhaps a functional approach is not radical enough :)

View on HN

Some of the most interesting applications of ML include recurrent neural nets, that act as processors, augmented with memory and an attention mechanism (memory addressing). They even call them Neural Turing Machines. The fundamental difference is that NTMs are learning from experience, and are not programmed top-down. So there is still a twist in the Von Neumann architecture.

    > For decades we have labored under the misguided assumption that the brain 
    > is isomorphic to the Von Neumann architecture, with "memory banks" and 
    > "processing units."

Who has been working under this assumption for decades? It seems rather obvious to me that this is very far from reality, so I'm a bit surprised if the AI community thought the brain were a CPU for several decades.

He seems to be under several misconceptions... This one may just be a mutation of the widely accepted idea that we can in principle emulate a brain using our standard computer models, even if it'll take n more decades of hardware advances, the idea that the brain is discretely computable.

The perceptron back in the 50s was modeled after network of neurons. I think you need to do more research into what AI researchers have done.

Can you elaborate on why you believe abandoning Von Neuman will be crucial for AI?

Reply on news.ycombinator.com