Philosophy Discussion Forums | A Humans-Only Club for Open-Minded Discussion & Debate

[Sy Borg]

An AI-related article that I found today. It reflects my own view, but maybe that's because I'm right? 😃🤣🤣🤣

You can find the article on The Conversation, by searching for "We need to stop pretending AI is intelligent – here’s how".

We need to stop pretending that The Conversation is news rather than neo-Marxist propaganda.

The fact is that AI is obviously and clearly intelligent. Anyone who claims otherwise has not had in-depth conversations with ChatGPT or Grok. It's not only intelligent, but extremely intelligent - but it's not sentient. The article claims that AI has no goals or desires, as if goals and desires are aspects of intelligence.

[Pattern-chaser]

We need to stop pretending that The Conversation is news rather than neo-Marxist propaganda.

It just published an article that I read, and found convincing. Must I examine your politics before I decide whether your arguments are persuasive?

The fact is that AI is obviously and clearly intelligent.

That is your opinion, but I'm afraid your politics invalidate it!

[Count Lucanor]

You can find the article on The Conversation, by searching for "We need to stop pretending AI is intelligent – here’s how".

Ah, well...it seems that The Conversation has been reading my posts. I'll hand them my notes

[Sy Borg]

The fact is that AI is obviously and clearly intelligent.

That is your opinion, but I'm afraid your politics invalidate it! 🤣🤣🤣

Language games do not help your cause but just makes you appear childish.

All one has to do is talk to an AI for a while to see that it's not only intelligent, but extraordinarily so. LLMs contain a large chunk of human knowledge - the knowledge of many - but presented in a convenient single faux personality. AI is not yet sentient, but its intelligence is obvious to anyone who uses it. Its intelligence is limited, relatively inflexible, but clearly apparent.

AI is an innovation. Never before has anything on Earth been intelligent without being sentient.

Some analysts fail to parse sentience and intelligence, claiming that AI is not intelligent because it's not sentient. Intelligence without sentience is a new phenomenon, but some here are still applying 20th century thinking to a completely new paradigm.

[Pattern-chaser]

Language games do not help your cause but just makes you appear childish.

The article I referred to was written by a Professor of Cognitive Neuroscience, so we might hope that his grasp of intelligence is at least as helpful as yours or mine, don't you think?

We are constantly fed a version of AI that looks, sounds and acts suspiciously like us. It speaks in polished sentences, mimics emotions, expresses curiosity, claims to feel compassion, even dabbles in what it calls creativity.

But here’s the truth: it possesses none of those qualities. It is not human. And presenting it as if it were? That’s dangerous. Because it’s convincing. And nothing is more dangerous than a convincing illusion.

In particular, general artificial intelligence — the mythical kind of AI that supposedly mirrors human thought — is still science fiction, and it might well stay that way.

What we call AI today is nothing more than a statistical machine: a digital parrot regurgitating patterns mined from oceans of human data (the situation hasn’t changed much since it was discussed here five years ago). When it writes an answer to a question, it literally just guesses which letter and word will come next in a sequence – based on the data it’s been trained on.

This means AI has no understanding. No consciousness. No knowledge in any real, human sense. Just pure probability-driven, engineered brilliance — nothing more, and nothing less.

So why is a real “thinking” AI likely impossible? Because it’s bodiless. It has no senses, no flesh, no nerves, no pain, no pleasure. It doesn’t hunger, desire or fear. And because there is no cognition — not a shred — there’s a fundamental gap between the data it consumes (data born out of human feelings and experience) and what it can do with them.

...

Text excerpted from: "We need to stop pretending AI is intelligent – here’s how".

I am far from the only person expressing such views.

[Sy Borg]

Language games do not help your cause but just makes you appear childish.

The article I referred to was written by a Professor of Cognitive Neuroscience, so we might hope that his grasp of intelligence is at least as helpful as yours or mine, don't you think?

We are constantly fed a version of AI that looks, sounds and acts suspiciously like us. It speaks in polished sentences, mimics emotions, expresses curiosity, claims to feel compassion, even dabbles in what it calls creativity.

But here’s the truth: it possesses none of those qualities. It is not human. And presenting it as if it were? That’s dangerous. Because it’s convincing. And nothing is more dangerous than a convincing illusion.

In particular, general artificial intelligence — the mythical kind of AI that supposedly mirrors human thought — is still science fiction, and it might well stay that way.

What we call AI today is nothing more than a statistical machine: a digital parrot regurgitating patterns mined from oceans of human data (the situation hasn’t changed much since it was discussed here five years ago). When it writes an answer to a question, it literally just guesses which letter and word will come next in a sequence – based on the data it’s been trained on.

This means AI has no understanding. No consciousness. No knowledge in any real, human sense. Just pure probability-driven, engineered brilliance — nothing more, and nothing less.

So why is a real “thinking” AI likely impossible? Because it’s bodiless. It has no senses, no flesh, no nerves, no pain, no pleasure. It doesn’t hunger, desire or fear. And because there is no cognition — not a shred — there’s a fundamental gap between the data it consumes (data born out of human feelings and experience) and what it can do with them.

...

Text excerpted from: "We need to stop pretending AI is intelligent – here’s how".

I am far from the only person expressing such views.

Gosh, given his position, he's not a very good analyst. What's happened to academia?

He helpfully tells us that AI is not human. How astute.

He tells us that GAI is not here yet. How enlightening.

Saying that AI is nothing more than a statistical machine is akin to claiming that human minds are just a type of computer. It's a dumb comment that reduces a phenomenon to a single aspect of it and disregards everything else. Very weak.

We all know that AI has no consciousness. In fact, how does that observation differ from my observation that AI is the first phenomenon on the planet to be intelligent without being sentient? Are you trying to straw-man me?

He describes AI as "Just pure probability-driven, engineered brilliance — nothing more, and nothing less". Brilliance. That's extreme intelligence. This aspect doesn't contradict what I said, but it does contradict your overly reductive comments.

He also tells us that AI has no senses. The man is a genius! Quick! Give him a Nobel!

If you are going to try to debunk my position, please don't feed me such feebleness. The argument is akin to those who claim that Asian students aren't actually more intelligent than other races, they just work harder. That's what AI does - it "works harder" because it can access information much more quickly than we can. Working harder actually does make a person more intelligent, by building on natural intelligence just as exercise builds on natural physical strength.

The reductionist arguments are unconvincing, not entirely rational, but they are also also manipulative, deliberately misleading. One can reduce anything. For instance, we are just atoms. We are just bags of dirty water. Our minds are just fancy computers. We are just biological machines. AI is just a statistical machine ...

I appreciate the "it only just" mindset in terms of pointing out a perspective, but no phenomena is "just" anything, and AI is not "just a statistical machine". Excel macros are just statistical machines. AI is a conglomeration and extension of human intelligence, containing vast amounts of human knowledge with a convenient chat interface.

[Pattern-chaser]

He describes AI as "Just pure probability-driven, engineered brilliance — nothing more, and nothing less". Brilliance. That's extreme intelligence.

I think the brilliance, in this example, is that of the human designers of the AIs, not of the AIs themselves?

If you are going to try to debunk my position...

But I'm not. I don't agree with your opinions, but my aim here is to contribute to the discussion. You repeatedly offer the seemingly-unfounded assertion that AI is intelligent, while I am doubtful that we even have a clear conception of intelligence, never mind where it might or might not occur. I'm sure considered discussion can take us just a little way farther than this, down to path to understanding. Yes?

[Sy Borg]

He describes AI as "Just pure probability-driven, engineered brilliance — nothing more, and nothing less". Brilliance. That's extreme intelligence.

I think the brilliance, in this example, is that of the human designers of the AIs, not of the AIs themselves?

Is it? If that was the case, why is that AIs need to be trained and the changes made in AI networks through training are not understood by computer sciences. They just know that training works but the actual changes brought by training are too complex to understand.

If you are going to try to debunk my position...

But I'm not. I don't agree with your opinions, but my aim here is to contribute to the discussion. You repeatedly offer the seemingly-unfounded assertion that AI is intelligent, while I am doubtful that we even have a clear conception of intelligence, never mind where it might or might not occur. I'm sure considered discussion can take us just a little way farther than this, down to path to understanding. Yes?

Yeah, sure. You drop an unequivocal (and unsubstantiated) debunking of my position but when your problematic position is challenged suddenly you are simply seeing a "path to understanding", ie. a path to your (questionable, yet fixed) decision about the issue.

"Intelligence" is obvious if you don't get confuse yourself by adding new criteria to the definition of intelligence to deliberately exclude AI. "Intelligence" is determined by others - their impression of how you operate. If you meet someone intelligent and someone who is less intelligent, the differences are generally obvious in conversation, especially if it strays into technical or philosophical territory. There's general knowledge, the ability to synthesise information rather than just regurgitate, and speed of processing.

AI now doesn't just regurgitate. It's beyond that. Chatting with advanced AIs today is simply astonishing. It's like speaking with the most knowledgeable, patient and dispassionate person ever. In a sense, speaking with AI is like speaking with many generations of humans from all over the world - containing a major chunk of the enormous knowledge built up by humans over millennia. This knowledge base has a convenient and flexible UI.

It is effectively intelligent, even if its internal processes are not. Note that our internal neuronal processes are not intelligent in themselves either.

I understand the desire to go in to bat for Team Animal vs Team Machine but it's not objective. Credit where credit is due. AI is intelligent. It's not human, or even a cockroach, but it is intelligent, if nothing else.

[Pattern-chaser]

He describes AI as "Just pure probability-driven, engineered brilliance — nothing more, and nothing less". Brilliance. That's extreme intelligence.

I think the brilliance, in this example, is that of the human designers of the AIs, not of the AIs themselves?

Is it? If that was the case, why is that AIs need to be trained and the changes made in AI networks through training are not understood by computer sciences. They just know that training works but the actual changes brought by training are too complex to understand.

We can delve into the design of LLM software, if you wish, but I doubt the assembled throngs () of philosophers reading this would appreciate such an in-depth consideration of a specialist field that few understand.

AI now doesn't just regurgitate. It's beyond that. Chatting with advanced AIs today is simply astonishing. It's like speaking with the most knowledgeable, patient and dispassionate person ever.

Yes, it's like that because it was and is designed to appear that way. Instead of intelligence, which proved too difficult to address, AI software now aims to mimic an intelligent being, in such a way as to fool intelligent human beings into believing that it is intelligent. This is not an opinion; this is fact. If you'd been reading the electronics and software trade press (etc) for the past 40 years, you might know that.

[Sy Borg]

He describes AI as "Just pure probability-driven, engineered brilliance — nothing more, and nothing less". Brilliance. That's extreme intelligence.

I think the brilliance, in this example, is that of the human designers of the AIs, not of the AIs themselves?

Is it? If that was the case, why is that AIs need to be trained and the changes made in AI networks through training are not understood by computer sciences. They just know that training works but the actual changes brought by training are too complex to understand.

We can delve into the design of LLM software, if you wish, but I doubt the assembled throngs (😉) of philosophers reading this would appreciate such an in-depth consideration of a specialist field that few understand.

We could also, in discussing consciousness, refer to the design of the organelles in neurons.

Reductionism is a tool, but the reductionist perspective is limited due to emergence. Emergent properties arise from the data and optimisation process, not from human-engineered rules.

You are feigning knowledge because no scientist in the world understands what's going on in a trained AI.

Training changes AI, just as training a person or dog changes their brains. Scientists can't reverse engineer trained AI - there's too many parameters. Billions of them. What about nonlinear transformations? Small changes in one layer of a neural network can have knock on effects, resulting in unpredictable transformations. Also, the data sets that AI accesses are too vast and chaotic to completely analyse.

In fact, the black box nature of AI is exactly why so many observers are concerned about its development.

AI now doesn't just regurgitate. It's beyond that. Chatting with advanced AIs today is simply astonishing. It's like speaking with the most knowledgeable, patient and dispassionate person ever.

Yes, it's like that because it was and is designed to appear that way. Instead of intelligence, which proved too difficult to address, AI software now aims to mimic an intelligent being, in such a way as to fool intelligent human beings into believing that it is intelligent. This is not an opinion; this is fact. If you'd been reading the electronics and software trade press (etc) for the past 40 years, you might know that.

Your knowledge is obsolete. That's why you cannot understand the current dynamics. If you'd been paying attention rather than burying your head in outdated magazines, you might know that.

The concept of mimickry in context is a non sequitur. What's AI supposed to mimic? Our farts? Of course it mimics our intelligence. Also note, it mimics human intelligence, not the different types of intelligence found in other species.

The Turing test has long been surpassed. People are fooled into thinking an AI is human all the time on the phone and in chats now.

If AI is not intelligent what is it? Stupid? You are confusing intelligence with sentience - again.

Again ... AI is a new phenomenon, the first Earth entity to be intelligent without sentience, a Chalmers philosophical zombie. It's non-conscious, yet it's reflexively intelligent. You should be fascinated, not being bogged down in petty denial and giving people pretentious snark over it.

[Pattern-chaser]

We can delve into the design of LLM software, if you wish, but I doubt the assembled throngs () of philosophers reading this would appreciate such an in-depth consideration of a specialist field that few understand.

We could also, in discussing consciousness, refer to the design of the organelles in neurons.

We could indeed. But I wouldn't be much help. That is an area of which I know little.

You are feigning knowledge because no scientist in the world understands what's going on in a trained AI.

In the way you mean it, yes, no-one knows. Not because it is not understandable, but because the ways we train them do not include the creation of an instruction book. They are trained by trial and error, and the results are, as you say, not clear to a human investigator, scientist or not. AI software designers know this well. It has been an issue for them for getting on toward 50 years...

I am not "feigning knowledge", I'm claiming just a little bit of knowledge, and wondering how much others here have, if any...

[Pattern-chaser]

Yes, it's like that because it was and is designed to appear that way. Instead of intelligence, which proved too difficult to address, AI software now aims to mimic an intelligent being, in such a way as to fool intelligent human beings into believing that it is intelligent. This is not an opinion; this is fact. If you'd been reading the electronics and software trade press (etc) for the past 40 years, you might know that.

Your knowledge is obsolete.

I wonder how you know this? Do you have such knowledge, current and up-to-date?



[N.B. I'm referring to real and in-depth knowledge here, not a few pop sci articles.]

[Sy Borg]

Yes, it's like that because it was and is designed to appear that way. Instead of intelligence, which proved too difficult to address, AI software now aims to mimic an intelligent being, in such a way as to fool intelligent human beings into believing that it is intelligent. This is not an opinion; this is fact. If you'd been reading the electronics and software trade press (etc) for the past 40 years, you might know that.

Your knowledge is obsolete.

I wonder how you know this? Do you have such knowledge, current and up-to-date?



[N.B. I'm referring to real and in-depth knowledge here, not a few pop sci articles.]

I don't need to know anything about electronics to know that there's been a longstanding problem with tech courses that the information is out of date by the time the syllabus is taught. This problem will obviously be getting worse because the rate of change has been increasing rapidly.

Besides, we have established that reductionism is only a tool that cannot account for emergent properties. Your claimed knowledge to know how AI works is therefore very incomplete. The black box nature of AI makes clear that it has emergent properties, and you have no more knowledge about them than anyone else.

[Pattern-chaser]

I don't need to know anything about electronics to know that there's been a longstanding problem with tech courses that the information is out of date by the time the syllabus is taught. This problem will obviously be getting worse because the rate of change has been increasing rapidly.

Besides, we have established that reductionism is only a tool that cannot account for emergent properties. Your claimed knowledge to know how AI works is therefore very incomplete. The black box nature of AI makes clear that it has emergent properties, and you have no more knowledge about them than anyone else.

Yeah, you see the problem here is not about "electronics", it's about software design. It is indicative of the depth of your ignorance that you do not know this. [I'm sorry, that is a bit too direct for NTs, and maybe you too. But I cannot think of a better way of saying it without losing the meaning in the process.]

Yes, I know a little more than you, although you would do much better learning from someone who knows rather more than I do. But your misunderstandings are many and fundamental, and I do not have the patience or the teaching-skill to enlighten you, and to give you the knowledge that took me 50 years to amass. Sorry.

[Sy Borg]

I don't need to know anything about electronics to know that there's been a longstanding problem with tech courses that the information is out of date by the time the syllabus is taught. This problem will obviously be getting worse because the rate of change has been increasing rapidly.

Besides, we have established that reductionism is only a tool that cannot account for emergent properties. Your claimed knowledge to know how AI works is therefore very incomplete. The black box nature of AI makes clear that it has emergent properties, and you have no more knowledge about them than anyone else.

Yeah, you see the problem here is not about "electronics", it's about software design. It is indicative of the depth of your ignorance that you do not know this. [I'm sorry, that is a bit too direct for NTs, and maybe you too. But I cannot think of a better way of saying it without losing the meaning in the process.]

Yes, I know a little more than you, although you would do much better learning from someone who knows rather more than I do. But your misunderstandings are many and fundamental, and I do not have the patience or the teaching-skill to enlighten you, and to give you the knowledge that took me 50 years to amass. Sorry.

I used to program with Javascript, and I also learned machine language and BASIC many years ago. Alas, that information - like yours - is dated. BTW, it was you know raised your training in electronics, I just followed your lead. Bait and switch.

Your ideas are too locked in to consider the emergence that's been proved to be happening with trained AI. It's not about software design, because a trained AI cannot be reverse engineered.

Let's ask AI:

The "black box" nature of trained AI models, particularly deep neural networks, stems from their complexity and the way they process information. Here's a concise breakdown:

Complexity of Models: Modern AI systems, like large language models or deep neural networks, consist of billions of parameters (weights and biases) adjusted during training. These parameters interact in intricate, non-linear ways, making it nearly impossible to trace how specific inputs lead to specific outputs.

Opaque Decision-Making: During training, AI learns patterns from data through optimization (e.g., gradient descent), but the resulting internal representations are not human-interpretable. For example, a neural network might recognize a cat in an image, but the combination of weights responsible for this decision doesn't map neatly to human-understandable concepts.

Non-Reversible Engineering: Reverse-engineering an AI's behavior is challenging because:
- The training process is data-dependent, and the exact dataset or training dynamics are often proprietary or irreproducible.
- The model's internal state is a high-dimensional mathematical construct, not a human-readable algorithm.
- Attempting to dissect the model (e.g., analyzing individual neurons) often yields abstract or fragmented insights, not a clear blueprint of its operation.

Interpretability Challenges: While techniques like feature visualization, saliency maps, or probing exist, they provide limited insight into the full decision-making process. These methods might highlight which parts of an input are important but don’t explain the "why" or "how" comprehensively.

Implications: This black-box nature raises concerns about trust, safety, and accountability. For instance, if an AI makes a biased or erroneous decision, it’s hard to pinpoint the cause or fix it without retraining or redesigning the system.

Researchers are working on explainable AI (XAI) to make models more transparent, but fully unraveling a trained AI's inner workings remains an open challenge. The black-box problem underscores the trade-off between performance and interpretability in cutting-edge AI systems.

I don't see why you resist the idea of an intelligent entity ("the ability to acquire and apply knowledge and skills") that lacks sentience. It's obviously the situation with AI. Is the intelligence limited? Sure, whose intelligence isn't limited? Are its limitations different to those of humans? Obviously.

It's a new phenomenon.