On Mon, Feb 24, 2025, 3:25 PM John Rose <[email protected]> wrote:

> On Saturday, February 22, 2025, at 5:49 PM, Matt Mahoney wrote:
>
> We have LLMs that pass the Turing test. That means you can't tell whether
> you are talking to a person or a machine. There is no definition or test of
> consciousness you can give me that doesn't either make both of them
> conscious or neither of them.
>
>
> Fine-tuning test for p-zombies:
> https://arxiv.org/abs/2502.07865
>

The author proposed that computers are not conscious because they are
deterministic, unlike the brain, which is analog and sensitive to tiny
variations in initial conditions. Qualia has a causative effect by fine
tuning the initial conditions to one where qualia exists, because by the
anthropic principle, we necessarily observe a universe where conscious
experience exists. He proposes an experiment on philosophical zombies in
which the initial conditions are slightly perturbed to see if qualia still
arises. Quantum mechanics is involved because it is a source of randomness
that would completely alter the output of an analog computer but not a
digital one.

Of course this is ridiculous. First, the author does not define qualia or
consciousness, just like most papers on the topic. Second, he seems unaware
of the definition of a philosophical zombie. A zombie is exactly like a
human by every possible test, and only differs in that it lacks qualia and
phenomenal consciousness. Thus, his proposed experiment is not possible.

We have senses of consciousness, qualia, and free will. They are what
thinking, input, and output feels like, respectively. You get this steady
stream of positive reinforcement to motivate you to not die, which results
in more offspring. I don't know why this is so hard to understand.


------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T72460285b911fa58-M13225e5380a3fdd42d654934
Delivery options: https://agi.topicbox.com/groups/agi/subscription

Reply via email to