Ask YC: What should I ask John McCarthy?

Lisp is not a functional language except in the trivial sense that it supports HOFs. Common Lisp is not even functional in the sense of Scheme, where purely-functional programming is not enforced but seems to be encouraged (disclaimer: IANAS). Common Lisp is a pragmatic, multiparadigm language in which functional, imperative and other styles can and should be used as appropriate. In a very real sense Lisp is that language "in the middle."

A philosophical AI class? Please elaborate. I took an AI class for my CS degree and it was technical.

My view may sound rather radical, but I think not everything that's called "A.I." has to do with intelligence. In other words, if a progress is reported by an AI researcher, it doesn't mean anything.

There are many forms of intelligence and for example Turing machines are just one, probably. If we are talking, however, about intelligent survival machines like ourselves, then we are not even close to understanding how it works.

Well, we have vehicles that drive themselves across deserts, but not machines that will argue with you over whether or not Will Ferrel is the greatest comedian ever. I say that we are making progress, though not as much as some had hoped.

I'd say that the algorithms are key in AI, not necessarily the language. You can program neural networks, genetic algorithms, search in Blub.

The early AI programs tended to solve toy problems. No one had thought much then about the implications of the curse of dimensionality: that each entity you add to the problem (say, throw another object into your object-manipulating robot's environment) increases the problem space exponentially (or even factorially).

But you're right, there has been a ton of progress in AI. The initial bias toward ontologies, expert systems, and top-down algorithms has given way to bottom-up systems that are data-driven rather than abstraction-driven.

One major example is using neural networks, SVMs, RBFs to discover implicit features in a data set rather than depending on an expert to code up those features explicitly. Experts don't scale, but data will always be with us. Thus we've seen increasing interest in information retrieval as opposed to ontological knowledge engineering.

But a lot is going on in the field even today. I found this talk very interesting:

http://video.google.com/videoplay?docid=-2469649805161172416

honestly, i think hard AI (as propounded in various SF genres) will always be next 20 years away. a borg like future with man-machine symbiosis seems more likely i.e use machine to _augument_ human capacities, rather than supplant them.

If humans can perform simple arithmetic operations, does that mean calculators are a solution to another "AI problem"? So are voice/face recognition systems (which, in fact, never gave satisfactory results so you can't even call them "solved").

""Lisp was meant to be a language for AI research originally": false premise": false premise.

A programming system called LISP (for LISt Processor) has been developed for the IBM 704 computer by the Artificial Intelligence group at M.I.T. The system was designed to facilitate experiments with a proposed system called the Advice Taker, whereby a machine could be instructed to handle declarative as well as imperative sentences and could exhibit ``common sense'' in carrying out its instructions.

http://www-formal.stanford.edu/jmc/recursive/node1.html