Search

Ideas that are…

Search Ideas

Creativity isn't defined by its outputs but by its process. RNGs do not recognise or criticise ideas.

#4813​·​Dirk Meulenbelt, 11 minutes ago​·​Criticism

We could say a person is a program that can synthesize any possible explanation in finite time, excluding memory limitations. But this would again grant personhood to RNGs. For that matter, a counting program could just enumerate all possible binary strings up to its memory limit, in finite time...

#4812​·​Tyler MillsOP, 29 minutes ago​·​Criticized1

You're right and I revised my criticism.

#4811​·​Dirk Meulenbelt, 37 minutes ago

A random number generator does not have universal creativity, because it is not a universal explainer: it can only generate explanations by accident. Universal explainers seek good explanations through conjecture and criticism.

#4809​·​Dirk Meulenbelt revised 38 minutes ago​·​Original #4781​·​CriticismCriticized1

Maybe... but "understanding" is too vague, I think. Doesn't understanding mean: can explain? But then this is just "can create any explanation" again. I think the core question is why a random program generator isn't a person, coming from Deutsch's definition of a person as a program that has explanatory universality -- can create any explanation (my thought here is that this definition isn't good enough on its own, given the random generator point).

#4808​·​Tyler MillsOP, about 1 hour ago​·​Criticism

Doesn't it? All explanatory knowledge is in the set of all possible programs, and a random program (or number) generator can generate any of those, given infinite time.

#4807​·​Tyler MillsOP, about 1 hour ago​·​Criticism

Even if variations are agnostic to any meaning or context of the knowledge, why are they still not implicit? Anything is implicit from anything else, if implicit just means: follows from when a given change is applied... The whole question is where the change is coming from... (?)

#4806​·​Tyler MillsOP, 2 days ago​·​Criticism

But an AI programmed to make random variations to its conjectures (English or otherwise) can only do so by choosing from an existing set of variations. Again, that knowledge is pre-existing. True evolution must involve variations to the substrate on which the knowledge is based; variations must be agnostic to the semantics of whatever they are acting upon, else they are already implicit from it, in which case their application does not constitute a truly novel conjecture (in the sense defining creativity).

#4805​·​Tyler MillsOP, 2 days ago​·​CriticismCriticized1

If only some of the criteria are stored, and the rest are random, is it still evolution? Is evolution only happening if there is random variation? But we could program an LLM to do that as well...

#4803​·​Tyler MillsOP revised 2 days ago​·​Original #4800​·​Criticism

If only some of the criteria are stored, is it still evolution? Then evolution is only the random part of the variation?

#4801​·​Tyler MillsOP revised 2 days ago​·​Original #4800​·​CriticismCriticized1

If only some of the criteria are stored, is it still evolution? Then evolution is only the random part of the variation?

#4800​·​Tyler MillsOP, 2 days ago​·​Criticized1

Whatever new "explanations" it creates are derivable from (and by?) the knowledge in the training data. It isn't evolution if all of the variations and selection criteria are stored ahead of time. That's just a search process, as in the case of Move 37 per AlphaGo.

#4799​·​Tyler MillsOP, 2 days ago​·​CriticismCriticized1

The definition of fitness for DNA also originated outside it, so this doesn't in itself suggest the system isn't actually creating new knowledge.

#4798​·​Tyler MillsOP, 2 days ago​·​Criticism

only people can create explanatory knowledge

How is an LLM not creating new explanatory knowledge (even if worse than the existing, by any measure), by varying some existing written explanation? It could even vary and select by some criterion of its "choice", thus realizing Popperian epistemology.

#4797​·​Tyler MillsOP, 2 days ago​·​Criticism

A person could create the same knowledge that biological evolution does, if only by simulating it. But it could still be true that only people can create explanatory knowledge. (That they can create all possible explanatory knowledge is Deutsch's criterion for personhood.)

#4796​·​Tyler MillsOP, 2 days ago​·​Criticized1

"No unconscious creativity" seems the simpler option. But here we arrive again at biological evolution, which is unconscious, yet is creating knowledge. Does this serve as a distinction between explanatory knowledge and non? Explanatory knowledge can only be created by a conscious process?

#4795​·​Tyler MillsOP, 2 days ago

Either there is no unconscious creativity, or only evolutionary/creative epochs with certain properties are conscious. The most obvious candidate for the property is complexity (in the sense of sophistication): only programs (existing knowledge) of a certain sophistication, once subjected to the evolutionary process, necessitates consciousness. Complex problem solving seems to require consciousness. Meanwhile, we do not seem to be conscious of "simpler" creative tasks, like... Like what? What is a "simple" creative task? What is an example of a creative task we perform unconsciously? How could we determine it was an act of creation (new knowledge), and not an act of deductive inference of the kind characterizing AI?

#4794​·​Tyler MillsOP, 2 days ago

This suggests that all experience is determined by what programs are being subjected to evolution at any given time, the niches that are being adapted to. But why is not all creativity in the mind conscious? (All consciousness might necessarily be creativity).

#4793​·​Tyler MillsOP, 2 days ago

But if the evolution is the defining feature of personhood, and the evolution is non-computational, then the personhood is non-computational. And consciousness would then not be a software property.

#4792​·​Tyler MillsOP, 2 days ago​·​Criticism

It could be simulated, but maybe it's very hard/intractable to do so. Maybe personhood harnesses physics to do the evolving, like a windmill harnesses the wind. Programs implemented such that the laws of physics cause them to evolve (unboundedly)?

#4791​·​Tyler MillsOP, 2 days ago​·​Criticized1

Programs could be evolved non-computationally. But that process could itself still be simulated, per the Church-Turing-Deutsch Thesis.

#4790​·​Tyler MillsOP, 2 days ago

By the Church-Turing Thesis, all computation can be specified/programmed. So the evolutionary aspect of a person can be specified/programmed, if it is computational.

#4789​·​Tyler MillsOP, 2 days ago​·​Criticism

The system may not have perfect knowledge of all programs present in it. The repeated independent emergence of winged flight in the biosphere comes to mind.

#4788​·​Tyler MillsOP, 2 days ago​·​Criticism

Because programs present in the system at one time could be no longer present at another time. Previously well-adapted programs could have decayed, been destroyed or consumed. So the same evolutionary path (approximately or not) could be travelled again, in principle.

#4787​·​Tyler MillsOP, 2 days ago​·​Criticism

But why would the system ever re-evolve to the satisfaction of a niche already satisfied previously? If the programs evolved by the evolutionary aspect of the person already exist, there is no more need for evolution of them.

#4786​·​Tyler MillsOP, 2 days ago​·​CriticismCriticized2