Hacker News Reader

Comment by dhs | Hacker News Reader

Yeah, the candidates are not all alike, even though Barsalou cites Spivey, who is really in the analogue camp. In contrast, the "Grounded Cognition" idea revolves around simulation - which may or be not work in an analogue manner -, and only denies the existence of amodal symbols in brains, as opposed to, say, Jeff Hawkins (who didn't do any experiments, either, AFAIK). Bringsjord is still another case; he's an Ex-GOFAI guy who, together with David Ferrucci, built the storytelling system BRUTUS.

http://www.chass.utoronto.ca/~sousa/BRUTUS_rev.html

Bringsjord makes another relevant contribution to the debate in "BRUTUS and the Narrational Case Against Church's Thesis", which used to be available from citeseer, but that site is down ATM, so I cannot provide a link.

But since you're asking a question of your own instead of answering mine, I take it that you don't know of any evidence, either.

View on HN

One cannot present evidence to differentiate between two theories unless they are both coherent and make clear and different predictions. Arguments against theories is a valid way to deal with them; so is asking questions to clarify what they are saying.

I don't understand you there. Spivey can make experiments which lead him to conclude that there may not be any fixed representations of anything in the brain. Bringsjord can manually solve the Busy Beaver for 6-state Turing Machines, while the machines themselves can't. These are examples of what I have; what I'm now looking for are examples of experiments from the results of which the opposite can be concluded.

How is it relevant whether there are any fixed representations in the brain? self-modifying code could achieve that.

there is no reason to believe that bringsjord can solve that problem and a computer can't. divide his method of solving it into very small steps. then answer which step did he do which a computer can't do?

The Busy Beaver is a classic example of a function which is not computable by a Turing Machine.

You haven't said which step he took to compute it which a turing machine can't do.

There are many steps. You have a set of different Turing Machines with alphabet {0,1}, each of which has, say, 4 states. You want to know which of these is the one that, starting from a tape filled with 0's, can write the largest number of consecutive 1's onto the tape, before it halts. If it halts - you don't know that in the beginning. A human can find out, by manually simulating the sequence and counting the steps. It's a lot of work - there are 61.519 possible 4-state machines -, but Bringsjord (or more likely, a group of undergrads available to him) has/have done it. A computer can't do it. For details, please read the paper.

You're telling me that a computer cannot simulate steps of a turing machine, one at a time? it can't store the current state of the turing machine, and the rules, in memory, and use the rules to get from one state to the next?

Are you really saying that a computer with too little memory can't do it, or something like that? because it seems blatantly obvious that a computer can simulate a turing machine.

A Turing Machine is an idealized computer. But no computer/TM can find out which of the possible 61.519 4-state TMs can write the longest string of 1's on a blank tape before halting.

A computer can try each TM one by one in the same way the humans did.

If you're talking about the halting problem now, humans also don't know whether the TM they are manually simulating will halt eventually.

And you still haven't said specifically what the thing is that the humans do and computers can't.

No, a computer cannot do it, due to incompleteness (Once there was a man called Kurt Gödel...) Bringsjords experiment proves that humans can "hypercompute" uncomputable functions. The great majority of functions which exist are uncomputable.

Reply on news.ycombinator.com