Re: Qualia and the Galactic Loony Bin

Eliezer S. Yudkowsky (
Wed, 23 Jun 1999 02:03:50 -0500

Harvey Newstrom wrote:
> The replay is not conscious. To get the replay to work, the Turing Test
> administrator has to do the exact same test. If the question is delayed a
> second, the "brain" will answer a question that was not asked. If the
> speaker to the "brain" burned out, it will answer without hearing the
> question. If the questions are not identical, the wrong answers will be
> given. The really-conscious brain would correctly react to these
> situations. The fake-conscious brain will fail to react.

Incorrect. The Giant Lookup Table is a tree, not a sequence. Given any sequence of inputs, ones and zeroes over an Ethernet line - which includes delays - produce a given output over the same Ethernet line at certain times. It's child's play to produce a lookup table that gives the illusion of being temporally sensitive.

> Your example proves nothing. A tape-recording of the answers would also


> pass your test and would appear to give the proper answers.
> I think the first run is conscious because it can adapt to any changes in
> the test. No matter which test is given, the real brain will pass.

I believe I covered this territory in my hal-referenced post on "subjective counterfactuals". When we say "If A then B", we mean "If not A then not B." The problem is that this is an entirely subjective distinction, depending on our estimate of what was most likely to vary, or even on the variable that just happens to be most salient - we can say the glass dropped because you relaxed your hand, because I startled you, or because of the law of gravity.

Oh, just read it.

           Eliezer S. Yudkowsky
Running on BeOS           Typing in Dvorak          Programming with Patterns
Voting for Libertarians   Heading for Singularity   There Is A Better Way