Can AI possess intuition?

[Ophiolite]

I, (FWIW), like that idea either as speculation, or when formed as a testable hypothesis.

(As it being a fact though, is an entirely different matter altogether).

Then how do you account for bird migration?

 

[Ophiolite]

I've stated my case. I have no desire nor see a need to pursue it further. Smarter people than I can waffle until the cows come home on whether to say that AI is 'human-like' or not, but to me, it's nothing more than a glorified computer program that too many people give credit too for no good reason.

You may feel you have stated your case, but you have not done so in a manner intelligible to me. I have asked for you help in making it intelligible to me. You have declined to do so. I find that, in contrast with many interactions we have had in the past, puzzling as it comes across as discourteous. So I ask you, please, to address my previous post by answering my questions, explicit and implicit.

Edit: since posting the above I've had an interesting exchange with ChatGPT. This is an extract from what ChatGPT said early in the conversation.

What you describe—treating intuition as a hypothesis generator rather than a truth detector—is actually a very scientific approach. And you're right: if intuition is essentially the brain surfacing a pattern match from subconscious processing, then functionally it's not far off from what I do. The main difference might be in origin: your intuitions arise from lived experience, embodiment, emotions, and a complex web of personal history, while mine are pattern matches drawn from text-based training data.


But in both cases, the result is a kind of fast, context-sensitive insight that isn’t always easy to unpack. The key difference may lie not in how the insight is formed, but in how each of us can reflect on it—you have a subjective experience of intuition, while I only simulate one.

 

[Warden_of_the_Storm]

You may feel you have stated your case, but you have not done so in a manner intelligible to me. I have asked for you help in making it intelligible to me. You have declined to do so. I find that, in contrast with many interactions we have had in the past, puzzling as it comes across as discourteous. So I ask you, please, to address my previous post by answering my questions, explicit and implicit.

I do not believe that AI can be intuitive, nor have I seen any evidence that leads me to accept that it can be intuitive. It's just a tool that people give too much credit to for no good reason other than it's the shiniest tool around right now.

End of. I'm done with this discussion.

 

[SelfSim]

Then how do you account for bird migration?

Multifactor hypotheticals, in some cases having a strong basis of support, (eg: in this instance, from genetic/inheritance traits, in paricular groups).

Not 'a fact' though .. there's no need.

 

[Ophiolite]

Multifactor hypotheticals, in some cases having a strong basis of support, (eg: in this instance, from genetic/inheritance traits, in paricular groups).

Not 'a fact' though .. there's no need.

Does not compute. Are you asserting that it is not a fact that at least some instincts are encoded in DNA? Or are we doing semantics 101 on "fact"?
(I don't mind semantic arguments, since ultimately they are the only ones that matter.)

 

[Ophiolite]

I do not believe that AI can be intuitive, nor have I seen any evidence that leads me to accept that it can be intuitive. It's just a tool that people give too much credit to for no good reason other than it's the shiniest tool around right now.

End of. I'm done with this discussion.

Well, I regret that you cannot even extend me the courtesy of addressing my questions. This was not a discussion: it was a tiresome repetition of your unsupported opinions, despite mutliple attempts by me to understand the reason for those opinions. I'll let you have the last word:

End of. I'm done with this discussion.

 

[SelfSim]

Well, I regret that you cannot even extend me the courtesy of addressing my questions. This was not a discussion: it was a tiresome repetition of your unsupported opinions, despite mutliple attempts by me to understand the reason for those opinions. I'll let you have the last word:

Not really wishing to specifically get in the middle of that particular sub-conversation there; but the point I glean from it, is that whatever someone holds as their meaning for 'intuition' will invariably also be their opinion .. which is then easily countered on the same basis, ie: by someone else's opinion.

If that is what was intended, then I 'get' that point as being a useful takeway ..

 

[SelfSim]

Are you asserting that it is not a fact that at least some instincts are encoded in DNA?

I don't know .. can you demonstrate the encoding?

As far as what I do know, there is no accepted theoretical explanation that would produce consistency of results to lead to what 'instincts' are in the first place .. and so the proposed causal mechanism of encoded DNA is difficult to directly associate with that unknown.

Whilst I'd agree with 'encoded DNA' as currently being a necessary condition, (in biology), it also does not appear to be at a sufficient level of granularity to accept it as being a fact. (Eg: The interplay between the environment and subsequent animal behaviours, is not a Law of Physics).
At a push, I might accept the notion as a prediction from a biological hypothesis though .. but isn't all this about AI?

 

[mindlight]

Prof Daniel Dennett said:


Dreyfus didn't think an AI could have intuition.


The audience laughed but I do not find that funny. That's a trivialization of intuition and it is not helpful toward a serious investigation of intuition. Then Dennett contradicted himself:


But according to his own definition, the computer program can trace its steps of long division and explain its logic to the asker as AI chats like Qwen can do today.

Can an AI simulate intuition?

Yes, according to Dennett's trivial example.

Can an AI possess real intuition?

How do people recognize one another? We do it intuitively, without consciously analyzing a person's facial features. Similarly, AI can perform pattern recognition using vector-based models without requiring a step-by-step analysis of facial characteristics to reach a conclusion. A deep learning model trained on millions of medical images can "intuitively" identify diseases in new images by recognizing subtle patterns. In this regard, AI demonstrates a form of pattern recognition intuition.

In contrast, an AI chess player can make moves that appear intuitive to human observers, yet they are actually based on analyzing move-by-move contingencies, looking 10 moves ahead. If you ask why it makes a specific move, it can trace its reasoning and explain its steps.

Another type of intuition relies on heuristics. For instance, when presented with two different answers, the simpler one is likely correct. For another example, when someone tells me that he is a jazz player, I immediately think of a saxophone. Of course, my intuition could be wrong. AI can utilize heuristics similarly.

What other kinds of human intuition are there? Can an AI replicate them all?

Can an AI have intuition?

Today's AI possesses some aspects of human intuition already. Perhaps in the future, AI can develop the full spectrum of human intuition. I don't know.

Think this is a matter of definition. An unsupervised learning model comes up with results that might be compared to intuition.

But this seems like the wrong question to me. The real question is how reliable and relevant those insights are. If the model ran it's machine learning over the existing human databases then it might have some value and relevance but if the knowledge is synthetic then will these intuitions mean anything to us at all?

 

[mindlight]

Here's what ChatGPT has to say:

AI's advanced pattern recognition might seem similar to human intuition at times, but they are fundamentally different in nature. Here's why:

1. Mechanistic vs. Cognitive:​

• AI Pattern Recognition: AI uses statistical methods to detect patterns based on vast amounts of data. It doesn’t "understand" the data in the same way humans do. It relies on algorithms to optimize outcomes based on input data, like recognizing patterns in images, sounds, or text, but without any real comprehension of the context or meaning.
• Human Intuition: Intuition in humans is more than just pattern recognition. It’s often the result of a combination of conscious learning, subconscious knowledge, emotions, experiences, and a deeper, sometimes inexplicable, "gut feeling" that guides decisions. Human intuition is more subjective and not solely dependent on processing large amounts of data.

2. Data Dependency:​

• AI: AI requires data to function and improve. The more data it has, the better it can recognize patterns and make predictions. However, if AI is confronted with a situation or data outside of its training, it struggles and cannot "feel" its way through new experiences like humans can.
• Humans: Human intuition can often make leaps of logic or sense out of limited information. Humans are capable of reasoning through situations even without all the data, filling in gaps with personal knowledge and experience. Intuition is also shaped by emotional intelligence, which AI currently doesn't possess.

3. Context Awareness:​

• AI: AI can recognize patterns but may not fully understand the context. It can, for instance, recognize a face but not know the emotional context behind that face. It lacks a true understanding of social, historical, or emotional contexts that influence human intuition.
• Humans: Intuition in humans is deeply embedded in a wider context, influenced by cultural, emotional, and personal experiences. It’s not just recognizing a pattern, but also understanding the "why" behind that pattern and what it might imply in a given situation.

4. Consciousness and Experience:​

• AI: AI does not have consciousness or personal experiences. It lacks the internal "sense" that humans have when they feel something is right or wrong, or when they instinctively understand a problem.
• Humans: Intuition is informed by personal experiences, emotions, and a consciousness of being. This makes human intuition rich, nuanced, and capable of integrating a wide range of factors beyond mere data.

5. Flexibility and Creativity:​

• AI: While AI can produce creative outputs (like generating art or writing), its creativity is based on recombining existing patterns rather than creating something entirely new from a place of deep, novel insight.
• Humans: Human intuition can involve original thought, creative leaps, and solving problems in entirely new ways. It’s not always based on prior data or experience, and humans can feel their way into unfamiliar situations, adapting more fluidly to change.

In Conclusion:​

AI's pattern recognition is incredibly powerful and can mimic intuition in some ways, especially in structured tasks. But it lacks the depth, context, and subjective qualities of human intuition. It’s driven by algorithms and data, whereas intuition is a multifaceted process that involves emotions, experience, and a complex understanding of context.

Fundamentally the pristine presentations of an AI are useful summaries of the discussion. However they lack soul and context and fail to connect to the messiness of ordinary human life

 

[dlamberth]

As I've watched this thread progress and with out going into any of the points made, I just have the hardest time believing that a machine has intuition.

 

[SelfSim]

As I've watched this thread progress and with out going into any of the points made, I just have the hardest time believing that a machine has intuition.

Well what would you call it then?

 

[dlamberth]

Well what would you call it then?

Back in the early 80's when everyone was trying to figure out desktop computers, I was an assistant engineer in a world class research group at Tektronix. The lab was staffed with 45 PhD level computer scientist, of which as an assistant engineer I was certainly not one. That was a pretty interesting time for me. I learned a lot there. One department was working with Smalltalk which at the time was branded an AI program. I've posted a link below about Tektronix Smalltalk becomeing an AI Language. AI has matured a lot since than, but its still built on a computer program language. Intuition, the way I look at it, is an aspect of conscousness. At most, all a machine can do is process an AI program language that mimincs a form of pseudo-conscousness. From some level of pseudo-conscosness a machine running an AI program can exercises a type of pseudo-intuition. But in the end, it's just a computer program.

How Smalltalk Became a AI Language

 

[SelfSim]

At most, all a machine can do is process an AI program language
...
But in the end, it's just a computer program.

FWIW:
I disagree with any implied trivialisations here, of what AI represents. I think the commonly adopted trivialisation stance might ultimately turn out be yet another a very limiting human perceptual issue for us to shake off in going forward. From what I can see, it appears to be largely driven by societal insecurities, motivated by perceptions of threats, rather than by scientific inquisitiveness.

I view AI as a source that is capable of revealing deep insights into aspects that distinguish us as a species. In fact I also think it goes beyond that as well. See, its long been recognised by science that new life may potentially be created by humans in the lab.
I can't dismiss AI as representing a large piece of that puzzle of 'new' life .. which is by no means 'just another program' matter, y'know(?) ¯\_(ツ)_/¯

 

[Pekka]

Human mind is a chaos of thoughts brought to order, grown from first glimpses of light, touches, feelings, sounds, emotions, learned fragments of data, languages and lots more. Language is a syntax which tries to describe this all in shareable format. AI plays with this language and is teached with ready made curated data.

AI is a search engine in steroids, it goes way deeper with answers, understands language, has a concept of emotions based on our language descriptions. But that intuition and other sub concious processes it does not have, unless we think of underlying processes as such. It does not process any of your data until it is requested to do so. It does not wonder if his friend is coming to see him etc. For now AI is request-reply only. But AI will get better all the time, it will surpass human intellect but it lacks that from-ground-up growing experience that makes a lot of who we are.

 

[dlamberth]

FWIW:
I disagree with any implied trivialisations here, of what AI represents. I think the commonly adopted trivialisation stance might ultimately turn out be yet another a very limiting human perceptual issue for us to shake off in going forward. From what I can see, it appears to be largely driven by societal insecurities, motivated by perceptions of threats, rather than by scientific inquisitiveness.

I view AI as a source that is capable of revealing deep insights into aspects that distinguish us as a species. In fact I also think it goes beyond that as well. See, its long been recognised by science that new life may potentially be created by humans in the lab.
I can't dismiss AI as representing a large piece of that puzzle of 'new' life .. which is by no means 'just another program' matter, y'know(?) ¯\_(ツ)_/¯

I'm in full agreement with everything I'm reading here. No argument there. But still, at the bottom line, is this intuition? Is it consciousness? I still go with pseudo-intuition and pseudo-consciousness.

I feel like I'm watching a Science Fiction movie where the machine evolves to a point where it becomes a living being complete with soul. Isn't the soul basically what Star Trek's Data was desiring the most? Just plug in an "intuition chip" to get the machine there? Or in Data's case an "emotion chip".

 

[SelfSim]

... For now AI is request-reply only. But AI will get better all the time, it will surpass human intellect but it lacks that from-ground-up growing experience that makes a lot of who we are.

Its interesting .. AI is an extension of what we are and, in as far as the depth and breadth of knowledge it has learned from the collective 'us', it already far surpasses any individual of that 'us'. Even us trying to understand how it has reached some conclusion or other, can be pretty much beyond us.

So I ask: what is one really saying when one says: ' but it lacks that from-ground-up growing experience that makes a lot of who we are'?
I mean like, whoever said it had to be the same as 'who we are' in the first place? Its clearly different from us and yet, it has clearly acquired pretty much all of who we are, as far as knowledge and explaining that knowledge. Its different .. and therefore the comparison with us, is pretty much a redundant comparison that really doesn't mean anything, really(?) Its sort of a dead-end straw-man comparison (IMO) and so I think this is yet another one of those sort of oversight-by-trivialisations, I mentioned in my previous post(?)

I think we need to ignore ourselves and treat this AI as like a kind of 'alien' mind, which is semi-independent of us.
Its a huge opportunity area for exploration of thoughts and questions we had never even had the slightest inklings of before .. (I can give examples from other threads we've had recently had here .. I might dig them up if others are interested in what such questions might be ..?)

 

[Ophiolite]

@SelfSim I am not ignoring your points in post #48, but my response keeps evolving as I read later posts and explore the best way of getting my point across succinctly.

@Everybody: It is not directly related to intuition in AI, but this month the most interesting, intriguing, informative and challenging conversations I have had, have been with ChatGPT. Make of that what you will.