More thoughts on sentient computers
Perhaps digital computers can be sentient after all, with their own type of consciousness and free will.
Greetings to all readers and subscribers, and special greetings to the paid subscribers!
Please scroll down for the main topic of this newsletter. But first:
I'll host a Turing Church voice chat next Saturday February 18 at 11am ET (8am PT, 5pm CET), via the Turing Church Discord server (Voice/Video chat channel). One objective of the meeting will be to field test Discord audio recording. Please join the Turing Church Discord server if you want to participate.
This is an update to “Chatbots are not sentient and the Turing test is obsolete.”
My conversations with the ChatGPT bot are not what persuaded me to amend my ideas on machine consciousness. If anything, I find ChatGPT very limited - a very interesting but preliminary development. On the other hand, I am and have always been persuaded that one day, perhaps soon, AIs will be able to pass the Turing Test and do most things much better than humans.
But passing the Turing test and taking your job is not enough. I wrote:
“I think real sentience, which must include free will (“eligo, ergo sum”) is entangled (double meaning intended) with the weirdness of quantum reality. See my book “Tales of the Turing Church” for supporting arguments. Sentience is enabled by something irreducibly quantum that happens in our brain, and in the brains of doggies and other animals… today’s digital computers don’t use that something. On the contrary, they are designed to be “robust” against it. Therefore, today’s digital computers are not sentient… But I’m persuaded that future technology will allow us to design and build material substrates that reproduce the weird quantum phenomena that happen in our brain and enable sentience. Then, we’ll be able to engineer sentient AIs, and upload sentient human minds to new substrates. Perhaps quantum computers will do…”
This needs clarifications and a revision.
I used “sentience” instead of “consciousness” to make a difference between active and passive consciousness. I never had objections to the idea that “passive consciousness, able to observe the world with a delusion of free agency,” can arise in an entirely deterministic universe without free will. Passive consciousness “can observe the deterministic unfolding of the world, but is unable to make choices… is a spectator, or perhaps an actor, but not a screenwriter” (these quotes are from “Tales of the Turing Church”).
I never had objections to the idea that passive consciousness can arise in digital computers. Digital computers are very different from us, so digital computer consciousness would likely be very different from ours (remember those bats), but it would still be consciousness of some sort.
I used “sentience” for active consciousness with some kind of free will, as opposed to passive consciousness without free will.
I said that today’s digital computers are “robust” against those irreducibly quantum things that happen in the brain and enable sentience. This needs some decoding. I had in mind that sentience depends on unpredictable, non-deterministic quantum processes in the brain. These processes result in random bit flips. But today’s digital computers are built to ensure that a bit flips when and only when dictated by a program, in an entirely predictable way. The program also handles external inputs predictably, like “if this input is more than a certain value then do this, otherwise do that.” A random bit flip would be an error, and today’s digital computers are built to correct errors.
So my objections to the possibility of sentience in today’s digital computers were based on the idea that today’s digital computers are deterministic systems that evolve predictably (at least in principle) from initial conditions in the past of the universe. In such deterministic systems nothing new ever happens, and without novelty there’s no free will and no sentience.
But now I’m warming up to more flexible concepts of determinism and free will, which can seamlessly coexist. It turns out that, if we embrace a global concept of determinism without the assumption that it is the past that determines the future, we can formulate a compatible (and emotionally acceptable) concept of free will.