Hacker News Reader

Comment by j4mie | Hacker News Reader

Why are genetic algorithms included? They are just a tool to explore multidimensional search spaces, and have nothing to do with AI unless they happen to be applied to an AI problem.

It would be nice if this was separated out into "old-fashioned" AI (data mining, statistical inference, machine learning, etc) and bio-inspired technologies (genetic algorithms, cellular automata, neural networks - although the latter may be a grey area).

Alternatively, a better name than "Artificial Intelligence" should have been used for the site. Probably too late now though..

View on HN

"old-fashioned AI" is usually a term reserved for symbolic techniques fashionable in the 60s, 70s, 80s. Learning and data-oriented techniques are mostly the new kids on the block (for Computer Scientists at least-- stats people have been at this game all along).

As for bio-inspired techniques-- why should their point of inspiration be anything but a footnote? Genetic algorithms are just one instance of a stochastic search algorithm (there are many others), cellular automata can be pretty much anything, and so can neural networks (depending on whether you include the dozens of models more esoteric than multilayer perceptrons).

The definition in the Russell and Norvig textbook is nice: "computational rationality"

...but given the widely-accepted computational theory of mind, humans are artificial intelligences too.

Excuse me, but I'd really like to see your evidence of how the computational theory of mind is "widely accepted".

http://en.wikipedia.org/wiki/Computational_theory_of_mind

> This view is common in modern cognitive psychology and is presumed by theorists of evolutionary psychology.

> common in modern cognitive psychology

Only among a small besieged minority-- the majority have moved on to neurocognitive research.

>and is presumed by theorists of evolutionary psychology

That's just straight-up wrong. Perhaps there's some theorist who buys into it, but it's certainly not common.

Actually you can include genetic algorithms into the Artificial Intelligence when your fitness-function applied to individuals gets a little bit "intelligent" (heuristics, neural networks...)

Of course - but one of the on-topic example questions is "How can I avoid premature convergence to a local optima on my genetic algorithm?". That has nothing to do with AI.

I agree, that way they could have included "traditional" local optimization as well, although probably that will come up in machine learning questions anyway.

Just about every Intro to AI course ever teaches heuristic search. GA are a type of heuristic search.

Reply on news.ycombinator.com