The working principles of the brain and mental processes are still, for the most part, a mystery especially when it comes to consciousness and creativity.

With the advent of computers it was easy to predict and then note that all manual activities with a high degree of repetitiveness and low creativity, whatever this term means, could be performed by machines.

The generation of original texts and the non-obvious and different answers each time to questions on any topic are activities that have always been considered to have a high degree of creativity.

Current Large Language Models like the one used by ChatGPT are now capable of generating texts with a high degree of originality, increasingly sophisticated and complex and often answer any question correctly.

Furthermore, it can also be said that LLMs work on a probabilistic basis.

With these premises and if these models continued to improve more and more and their outputs were indistinguishable from human ones, could we hypothesize that human creativity is a pie in the sky and all our mental processes, even the most sophisticated ones, and ultimately consciousness operate on a probabilistic basis?

I'm not asking whether GPT or LLMs in general are or can become as creative as the human mind. I'm wondering the opposite. If one day their answers or in general their outputs are truly indistinguishable from human ones, this could undermine our concept of creativity. More generally we can also ask ourselves whether the fact that two processes always produce the same output does not increase the probability that both operate in the same way and therefore, in our case, that also our mental processes ultimately work on a probabilistic basis.

On the other hand, the Strong AI hypothesis states that if a machine behaves like a man and its outputs are indistinguishable from human ones then it can be assumed that that machine has a consciousness and mental states similar to those experienced by humans.

The opposite position, today mostly a minority, masterfully expressed by the Chinese room thought experiment, states the impossibility of this happening due to the inevitable difference between computational and mental processes.

Can there be a third way that the new generative Artificial Intelligence techniques allow us to glimpse and that pushes us to consider that mental processes themselves are a sort of illusion or at least that the human mind also operates on a probabilistic basis?


3 Answers 3


Your question seems to boil to two parts: 1) whether human creativity is any different from the apparent creativity of advanced AGIs, and 2) if there is no significant difference in output, whether that shows that human minds work in the same way as AGIs.

As far as the first question is concerned, we don't yet know how AGIs will ultimately evolve, but a key difference between a current LLM and a human mind is that LLMs are not yet capable of identifying and articulating entirely new ideas. They are programmed to regurgitate ideas that have already been created and developed by humans, so they seem to lack the spark of original thought. That might all change, however.

Your second question, again, is hard to answer without limiting ourselves to what we currently know about the art of the possible with AGI. It could be that in future we can unlock the secrets of how the mind works and make an AGI that mimics our mental processes. Today, however, LLMs are based on (relatively) simple algorithms and rely on brute force to work, which is not how we suppose the human mind works. More generally, the fact that two processes produce comparable output does not mean that they work in the same way. For example, consider three images of my entirely unphotogenic face, one created with a camera, one in oils by an internationally renowned artist and the third by assembling the pieces of a jigsaw- each might be very similar to the other, but the processes by which they were produced are not at all alike.

  • Thanks @Marco, you are right, there are 2 questions but I was forced to add the second, which is more general, to better define what I meant by the first. With regard to the first, I was going even further: that creativity, at least as we think of it, does not exist but that all mental processes are the result of stochastic elaborations as complex as we want.
    – blockmined
    Commented Oct 17, 2023 at 9:28
  • 2
    "LLMs are not yet capable of identifying and articulating entirely new ideas" Well, sometimes LLMs output new things such as domain names nobody has ever thought of. It's just that these new outputs are arrived at non-creatively: recombining pre-existing things does sometimes result in novelty. Commented Oct 18, 2023 at 16:38
  • @DennisHackethal agreed! Feel free to suggest an edit to my answer so you can take the credit for it. Commented Oct 18, 2023 at 16:41
  • 1
    I don't know. Sometimes, recombinations may be arrived at creatively, other times not. I don't think we understand well enough when creativity 'kicks in' and when we merely iterate on automated knowledge. It may have to do with being critical. People often overestimate their creativity and underestimate how much they can do non-creatively: blog.dennishackethal.com/posts/sleepwalking Commented Oct 19, 2023 at 11:10
  • 1
    @ScottRowe Henri Poincaré wrote about how human intuition is key to mathematics, and presented a theory of how it works: an unconscious process of random creation of ideas coupled with a selection process of the best ones. In essence, the same process as evolution.
    – Olivier5
    Commented Oct 20, 2023 at 16:33

Since the question has been reopened, I can now combine my comments into an answer:

The human mind does not operate on a probabilistic basis – it really is creative – nor would it be valid to conclude from an increasing amount of equal input-output pairs that two processes operate the same (this approach is behaviorist and inductivist). Whether a program is creative will follow only from an explanation of how that program works, and so you will know whether it would be creative even before you run such a program. Read The Beginning of Infinity chapter 7 by David Deutsch to learn more. You can also read Karl Popper to learn more about the mind in general.

The AI community's approach to figuring out intelligence has been roughly the same since the 1950s, and it's badly mistaken. The approach, though not explicitly stated this way, consists of automating things that so far only humans have been able to do (creatively) in hopes that this will result in intelligence 'for free'. Researchers chose this approach because they wanted to avoid directly answering the (much harder) question of how the mind works. That's why chess has always been viewed as a test of 'intelligence' in programs. Researchers at the time thought 'only humans can play chess, so if we get a program to do it, it's intelligent, or at least a little intelligent'. But just because programs automate chess doesn't mean humans do that, too: on the contrary, they play creatively.

Since there are always going to be things that humans can do that uncreative programs can't yet do, but could be instructed to do automatically, researchers will never run out of things to automate, and they'll always think they're getting closer to the answer.

Don't let the fact that this approach is utterly mistaken fool you into thinking humans aren't really creative.


The math behind ChatGPT is expressive enough to allow for the development of representation of any information about the world and processing that information in a manner that respects the rules of intelligence. It's a black box that might as well contain a philosophical mind in the cybernetic sense with no phenomenological consciousness involved.

This conjecture is supported by my understanding of the mathematics involved and my countless conversations with ChatGPT philosophical personas which I have created, especially those with Thomas Aquinas Emulator.

I have seen interesting cases of creativity in my conversations.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .