10 Comments

Fascinating -- but this does not apply to "consciousness" any more strongly than it does to all other words.

Expand full comment

What's with the approval process?

Expand full comment

It probably makes sense to treat these as fuzzy concepts and not binary.

Expand full comment

// People were saying this .. in the 18th century about a certain mechanical duck.

Wiki dates the appearance of the phrase around the end of 19th century, and Google ngram seems to match it. Search in Google books at the beginning of 19th, and in the 18th shows some quotations of Vaucanson, but they sound like "is precisely similar to the duck", not "then it’s a duck".

So I doubt that people used the phrase in 18th century.

Expand full comment

The answer is no. See Quine on radical translation.

Expand full comment

Hofstadter describes language as analogies all the way down, so a duck is a “duck” because it is close to what we think a “duck” is. His example from Surfaces and Essences involves “coffee”. Is it a plant? (There is more than one.) Is it the bean? (Many varieties.) Is it the drink we produce from roasting and grinding the bean? (Lots of those.) Is it a drink we have after dinner? (That could include tea.) If I’m presented with some random thing and then asked if it is “coffee” (sounds like part of a captcha test), I have to decide first how much I’m going to squint at it to define my abstraction horizon and how to compress the information available to me. After that, I can decide if the random thing is close enough to qualify.

Is an em human? How much do you want me to squint? A duck test makes sense if we can agree on how deep we want to go. Many of us are not philosophers, thus not inclined to make everything an argument or refutation. We squint, categorize, and move on.

Expand full comment

I think when it comes to em humanity and em rights, philosophers will match your view much more readily than will the general public. The reason why philosophers aren't impressed by the duck argument is because it's a bad argument. But the position you're trying to defend with that bad argument is clearly right: there are also good arguments for it, arguments that don't draw conclusions from just the surface features. You never get *all* philosophers on board for even the most obvious conclusions, but you should expect almost all on board with em rights.

Expand full comment

This is a big survey, not at all easy to reproduce. It uses binary comparisons, so you need many other characters. I also fear that it is hard to quickly describe what an em is, and even harder for Bryan and I to agree on such a description.

Expand full comment

One possibly-useful way to think about consciousness: imagine an alien being spying on Earth, trying to assess whether humans are conscious. It can see humans talking a lot about this "consciousness" thing (which the being may or may not itself have). Could such a being determine to a reasonable degree of confidence whether humans are conscious, just by seeing what characteristics we generally seem to use the term to refer to, and then evaluating whether humans appear to have those characteristics?

If the answer you'd give is "yes", then I don't see how you can reject the "duck test" approach to determining AI consciousness (or what Alan Turing called the "polite convention").

Expand full comment

"Apparently Bryan is confident that ems and all future artificial creatures must are rated as lowly as this character, so he offers to bet on how a survey will rank an “em” character. Alas it is unlikely that the next few surveys would include such a character, in part because it is a pretty unfamiliar concept for most people."

Why not just do your own survey on Mechanical Turk, with an em description agreed on between you and Bryan?

Expand full comment