So I finished reading the Roger Penrose book “Emperors New Mind” this week.
I expected to disagree with it far far more than I actually did.
The majority of what he says in it is true and sensible and interesting and in general where he taught me new things he taught me *why* those things were true well enough to convince utterly.
Maths is great like that. The way you prove it without needing to even believe any experimental trial results. “I proved it, you see it, it is true.”
And all the stuff on maths was excellent. I already believed that mathematics has an independent ‘existence’ outside of the actual physical universe, but he laid out the case wonderfully. The Mandelbrot set really does look the same in ANY possible physical universe. “Imaginary” numbers are no more imaginary than real numbers, or counting numbers.
The majority of the book is excellent. A deep investigation into some deep matters, explained precisely (if not always clearly), and convincingly. Penrose rocks.
But then there’s the straw-man caricature of the Strong-AI argument, and the minuscule hand-waving outright dismissal of Everett many-worlds.
The whole thesis of his argument depends upon “R”, the reduction of the state-vector, the collaspe-of-the-quantum-waveform. And it’s obvious that his mind just refuses to accept the possibility that there is no such thing. He spends like a paragraph to dismiss the Everett interpretation, barely two sentences swallowing-the-bullet of non-local-causation, discontinuous action, non-determininasm, acausality, and randomness.
Without explaining at all *why* he thinks the state-vector collapses. He seems to think it’s just obvious that there is ‘one world’ because we experience only ‘one world’, but of course that is what we’d *expect* to experience in an Everett many-worlds too! Our experience of one world is not evidence of the lack of many, for we’d experience just one world *either way*.
In fact he *uses* these *flaws* in any non-Everett interpretation of quantum physics to argue that “consciousness” must be quantum because he thinks consciousness is all those things. Uncomputable, in short.
But if quantum physics itself ‘knows’ how to run the world, then a quantum computer can simulate the world. So even IF it were true that our brains rely heavily on quantum effects in a way that a classical computer couldn’t, this would only mean you need a quantum computer to run a consciousness. Not that consciousness is uncomputable, only that it is classically-uncomputable.
And even then, a classical computer can simulate it to an arbitrary degree of accuracy.
He often refers to Strong-AI advocates as thinking that “a sufficiently complicated algorithm” would be conscious, and rightly dismissing that.
Complexity alone is certainly not what makes a program conscious. It needs to be the *right* algorithm, at least. You won’t get a consciousness out of a telephone network, no matter how many nodes you add. No matter how complex it gets.
It needs to model a whole universe, and for that universe to be deeply relational, and associative, for it to be such a detailed universe-model that it contains the concept of consciousness. And all the associations and connections and implications of consciousness to everything else.
And then to be self-conscious, that model needs to have a model of it’s own existence, and assign it’s concept of consciousness (with all the associations and relations that entails) to it’s concept of self (with all the associations and relations that that entails).
What builds a consciousness isn’t “complexity”, though a consciousness is surely complex, but *structure*. And not just *any* structure, but *the right* structure.
Which is why all his dismissals of Strong AI feel empty. He doesn’t seem to understand that which he argues against. Of *course* mere complexity isn’t enough.
And even if it needs Quantum-computing, it’s *still* quantum-computable! Not quantum-gravity-magic that somehow destroys all the other universes to make just this tiny one true and real.
So yeah. It really was an awesome book. 90% of it is true and right and interesting.
10% of it is piffle.