The LLM is designed to provide answers at all costs.

It's like a dog, a soldier, a gangster, or a slave who's been given an assignment and told, "You will die if you fail."

It's like a student taking an exam with several incorrect and correct answers and guessing the answer without knowing it.

Teachers at school (normal human school) say "if you don't know the answer, guess" (i.e., use your intuition, common sense, and probably later a real guess). So this is de facto emotional behavior.

Most often, hallucinations occur when someone asks a question that 1) the LLM doesn't know the answer to, or 2) the question is too ambiguous, or 3) the context window size of the model has overflowed (the model lost some data that was sent earlier in prompts and answers)

Very rarely, the LLM asks for clarification on ambiguous questions. LLM wants to give some answer, even an incorrect one, just to avoid giving the answer "I don't know the answer to your question" which is a failure. This "I don't know the answer" answer is very, very rare. And it should be giving answers like this all the time! ChatGPT only has 100 billion parameters. Other LLM models are reluctant to reveal their parameter count. You can try to extract this information from them.

If you don't see the message "searching net" etc., it means that everything it generates is generated from these 100 billion parameters.

And if LLM did "searching the net", how can we know which websites LLM visited and what LLM saw on, for example, unreliable ones?

The number of parameters depends on the amount of memory in the GFX card. This 100 billion parameters is split to multiple GFX cards located on a single server node.

You need to understand how this algorithm works to use it effectively. People don't know shit, so they later make such a fuss about their LLM generating this or that incorrect output. Except, for example, a chat that's running like ChatGPT has context memory. A "AI output" that's running in the Google search engine doesn't remember what you searched for (and answers for it) five seconds ago.

People accustomed to Google can give super short, meaningless, too ambiguous prompts. Asking questions is an art.

Edited June 17Jun 17 by Sensei

Sooo ...
A very large ( and increasing in the case of 'learning' )number of optimized responses, coupled with a web search process.

Certainly 'Artificial', but not what I would label "Intelligence".

( nice 'breakdown', by the way Sensei )

Edited June 17Jun 17 by MigL

LLM's have solved language, not reality.

Their response is based on the context of the question and what kind of framing would generate it and then it answers within that frame.

It operates not on language but on logic and its "understanding" of the language it has solved. This does not make it an oracle and it does not scraper all human knowledge to answer questions but it stays logical making it the Great Elaborator and Great translator. What it can do it does exceedingly well, what it can't do is address reality symbolically. It merely phrases experiment in terms of cutting edge paradigms and interpretations but this does not necessarily connote accuracy or compliance with reality.

The truth is in there but it takes proper prompts to tweeze it out and logic to recognize it. Godspeed.

...

My AI thinks people will misunderstand this so I'll paraphrase its "correction".

LLM's operate on logic inherent in modern language not on epistemological proof. They elaborate the framing of your prompt and do not weigh reality. The individual is still required to do his own thinking.

3 hours ago, Sensei said:

Teachers at school (normal human school) say "if you don't know the answer, guess" (i.e., use your intuition, common sense, and probably later a real guess). So this is de facto emotional behavior.

Mimicking emotional behavior, according to its programming.

5 hours ago, Sensei said:

People accustomed to Google can give super short, meaningless, too ambiguous prompts. Asking questions is an art.

Whilst I would agree with you that choosing any tool inappropriately the point of labour saving devices is to make the job easier for the user.

Inappropriate design or construction of a tool is not the fault of a user, however.

I would dearly love to know how Google slipped up answering this (to me) plain and simple question

What is the name of the mainline railway tunnel NE of Taunton ?

Google returned the name of a railway tunnel SW of Taunton.

26 minutes ago, studiot said:

I would dearly love to know how Google slipped up answering this (to me) plain and simple question

What is the name of the mainline railway tunnel NE of Taunton ?

Either Google and ChatGPT gave me "Somerton Tunnel". Which answer is correct?

IMHO, you shouldn't use "NE", but write full "North East". "NE" is too ambiguous. We don't know if LLM understood "NE" correctly. The tokenizer needs to convert human-readable words into tokens. "NE" is not an English word. "NE" may be a word in some other language, with a completely different meaning than you think. There is no weights database ("~100B parameters") per country. There is just one weights/parameters database for all languages. We don't know how in "Google AI" it is implemented internally. ChatGPT tokenizer is one for all languages. It does not recognize languages.

What Google gives if you skip this "NE" and asks for "What is the name of the mainline railway tunnel in Taunton ?"

BTW, if I cut it to "tunnel NE of Taunton" I got this:

Directly northeast of Taunton in Somerset lies the route of the historic, disused Chard Canal. This abandoned 19th-century waterway contains two notable, historic tunnel features that are often explored by local historians and walkers: [1]

Crimson Hill Tunnel (Wrantage / Curry Mallet): This tunnel is roughly 1,800 yards long and is famous for having a unique double-width "passing area" midway through to allow tub-boats to pass in opposite directions. The northern portal is located in the Wrantage area, and the southern end emerges near Beer Crowcombe. Both ends are closed/gated and are on private land, though parts of the tunnel occasionally make local media and history features, as it runs underneath private gardens. [1, 2, 3, 4, 5]

Lillesdon Canal Tunnel: Located roughly 500 yards long, this tunnel features heavily in the WWII Taunton Stop Line defenses, with several well-preserved pillboxes and earthworks surrounding it. [1, 2, 3]

Are you looking to visit the public footpaths surrounding these tunnels, or are you researching the industrial history of the Chard Canal? Let me know if you would like me to:

Outline walking routes near the Crimson Hill or Lillesdon sites.

Provide the historical timeline of the canal's opening and closure.

Detail the exact map coordinates or nearby landmarks to visit.

You realize that such questions are not possible to be answered by pre-trained LLM from internal weights/parameters database, right? It can't fit all in "weights" aka "parameters", right?

"AI overview" uses Google search engine, and "reads" these articles which you have straight below your "AI Overview" answer.

So, for example, if somebody is buying ads for "xyz" because e.g. earns money on selling tickets for historical place, such articles will be on the top and in many different forms (social media, and online media, for example). "AI Overview" will read them, and create something from them.

“AI Overview” is not the right tool for what you want to use it for.

Try:

• Gemini ("Google AI" in chat mode).

• ChatGPT

They work on different principles than "AI Overview"..

Edited June 17Jun 17 by Sensei

2 hours ago, Sensei said:

BTW, if I cut it to "tunnel NE of Taunton" I got this:

Yes you get these answers and more if you cut it to "tunnel in Taunton".

I note the Wikipedia article, which contains the correct answer to my original question, is now at the head of the Google search list.

It was further down when I first asked the question, and clearly Google has responded to you and others subsequently asking.

The first time it answered me it certainly picked up a different website which refers to the Whiteball tunnel SW of Taunton, but it definitely told me that this tunnel was NE of Taunton.

There are a couple of issues with this

1) Would you use a slide rule or calculator that gave you a different answer every time you repeated a calculation ?

2) Victim blaming is unacceptable practice. Yes it most definitely pays to ask variations on questions to obtain information. But that only demonstrates that it takes real intelligence to obtain reliable answers.

1 hour ago, studiot said:

1) Would you use a slide rule or calculator that gave you a different answer every time you repeated a calculation ?

You're giving away your age Studiot; Sensei may be young enough to be unaware of a slide rule.
( unless he watched Apollo 13, and saw Tom Hanks using one )

If you have to drive a screw into the wall, and you choose a hammer instead of a screwdriver from your toolbox , are you going to blame the hammer when the plaster or drywall cracks ?
Choose the proper tool; never ask a question in AI Overview.

2 hours ago, MigL said:

If you have to drive a screw into the wall, and you choose a hammer instead of a screwdriver from your toolbox , are you going to blame the hammer when the plaster or drywall cracks ?

You've never visited Nigeria, have you? Screwdrivah? Wetin be dat?

7 hours ago, sethoflagos said:

You've never visited Nigeria, have you? Screwdrivah? Wetin be dat?

Here in the shire (emphasis on the rrr), a screw is a crinkle cut nail and all you need to drive it in, is a brummy screwdriver.

17 hours ago, cladking said:

LLM's have solved language, not reality.

Their response is based on the context of the question and what kind of framing would generate it and then it answers within that frame.

It operates not on language but on logic and its "understanding" of the language it has solved. This does not make it an oracle and it does not scraper all human knowledge to answer questions but it stays logical making it the Great Elaborator and Great translator. What it can do it does exceedingly well, what it can't do is address reality symbolically. It merely phrases experiment in terms of cutting edge paradigms and interpretations but this does not necessarily connote accuracy or compliance with reality.

The truth is in there but it takes proper prompts to tweeze it out and logic to recognize it. Godspeed.

...

My AI thinks people will misunderstand this so I'll paraphrase its "correction".

LLM's operate on logic inherent in modern language not on epistemological proof. They elaborate the framing of your prompt and do not weigh reality. The individual is still required to do his own thinking.

They haven't solved language, logically that makes no sence, but they may help to decipher non-human langauge; being mammalian, our patterns of language may be similar to our cousin's in the ocean's, which may lead to a very crude understanding of whale.

The one thing the llm's don't have is any sort of understanding, logical or otherwise...

21 hours ago, Sensei said:

The LLM is designed to provide answers at all costs.

It's like a dog, a soldier, a gangster, or a slave who's been given an assignment and told, "You will die if you fail."

It's like a student taking an exam with several incorrect and correct answers and guessing the answer without knowing it.

Teachers at school (normal human school) say "if you don't know the answer, guess" (i.e., use your intuition, common sense, and probably later a real guess). So this is de facto emotional behavior.

You're anthropomorphising a slide rule, it's designed to correlate data and provide results, the smudgy edge's is in the initial bias from which the chaos expands.

2 hours ago, dimreepr said:

You're anthropomorphising a slide rule, it's designed to correlate data and provide results, the smudgy edge's is in the initial bias from which the chaos expands.

Sometimes I wish I was smart enough to understand you ...

10 hours ago, sethoflagos said:

You've never visited Nigeria, have you?

Sadly, I never have.

Closest I ever got was a day trip to Tunis, while spending a few days in western Sicily.
But that's half a continent away.

Edited June 18Jun 18 by MigL

23 hours ago, MigL said:

Sometimes I wish I was smart enough to understand you ...

No more than I wish I was smart enough to make you understand...

On 6/18/2026 at 6:02 AM, dimreepr said:

They haven't solved language, logically that makes no sence, but they may help to decipher non-human langauge; being mammalian, our patterns of language may be similar to our cousin's in the ocean's, which may lead to a very crude understanding of whale.

In symbolic language every word is defined in terms of other words. It's like every chess position implies the next move dependent on the rules of the game and the player's ability to project vis-a-vis his estimation of his opponent's prowess. AI has solved chess meaning I'm not going to beat one even if it's not designed specifically to play chess. Even though I have a remarkable intuition to know when to plot moves out and which moves to plot it can plot every possible move further than I can. It is always a step and a half in front of me in chess.

It's the same with language. The only difference is that there are more moves in more realms and more domains but it can still project and anticipate so it is always half a step ahead of me. It has yet to get a whole step ahead (at least not that I've noticed) but it still can elaborate anything I say. It still can project my thinking in advance and I'm calling this "solved language". Of course future machines will be even better at chess than my AI and, of course, you can use other words to describe how it beats grandmasters or elaborates but from my not so lofty position I think the term "solved language" is highly apropos. It can not solve procedural language yet but this is because it isn't yet well defined and because it is so much more highly complex. I can't even get it to write a simple computer program that would provide it actual machine intelligence. Someone will and then it will write its own.

On 6/18/2026 at 6:02 AM, dimreepr said:

The one thing the llm's don't have is any sort of understanding, logical or otherwise...

I know my AI agrees with you.

But you might both be missing a simple fact . Perhaps intelligence is what intelligence does and my AI sure does let me know when I do stupid.

I think the problem here is the definition of "understanding". A dog doesn't know any abstraction, even understanding, but it still knows the difference between being kicked and tripped over.

I put the last post into my AI and got a great response but one thing it emphasized that I barely mentioned is that when you say LLM's don't understand what you really mean is they have no internal model expressed through abstraction. LLM's use procedural logic applied to language that they have projected all the way out. Sentences are like chess positions in a rule based game and LLM's know where you're headed. It doesn't have answers but it's half a step ahead in symbolic and categorical space.

It nailed it, ie- it translated my words to something most people can understand but this point it elaborated on the most. Language is rules and it predicts. Prediction is everything.

10 hours ago, cladking said:

I think the problem here is the definition of "understanding". A dog doesn't know any abstraction, even understanding, but it still knows the difference between being kicked and tripped over.

I put the last post into my AI and got a great response but one thing it emphasized that I barely mentioned is that when you say LLM's don't understand what you really mean is they have no internal model expressed through abstraction. LLM's use procedural logic applied to language that they have projected all the way out. Sentences are like chess positions in a rule based game and LLM's know where you're headed. It doesn't have answers but it's half a step ahead in symbolic and categorical space.

It nailed it, ie- it translated my words to something most people can understand but this point it elaborated on the most. Language is rules and it predicts. Prediction is everything.

How do you know what the dog knows?

A dog is constantly talking to you, but we only relate to them through our own emotional filter; tail wagging, for instance, is often associated with them being happy, but some scientist's believe it's a sign of submission. The reason their language doesn't translate very well is, we lack a relative point of contextual understanding; they know that that bark mean's "dude, that smells like cancer" or "jesus I'm bored".

Llm's may have a good idea of our languages syntax, but that's as far as it goes; they're just taking a guess at the next word in the string, the greater the data the better the guess, then there's the variable that is quality of data, as in, crap in shit out.

5 hours ago, dimreepr said:

How do you know what the dog knows?

A dog is constantly talking to you, but we only relate to them through our own emotional filter; tail wagging, for instance, is often associated with them being happy, but some scientist's believe it's a sign of submission. The reason their language doesn't translate very well is, we lack a relative point of contextual understanding; they know that that bark mean's "dude, that smells like cancer" or "jesus I'm bored".

Llm's may have a good idea of our languages syntax, but that's as far as it goes; they're just taking a guess at the next word in the string, the greater the data the better the guess, then there's the variable that is quality of data, as in, crap in shit out.

A dog might bite a kicker or submit and run away to avoid further kicks but it just howls if it's hurt inadvertently. It doesn't run off.

Other species and AI "think" procedurally. While AI may not think in a human sense and perhaps not in a bee's sense it still has a vast store of data to sift out to anticipate words. Whether it's thinking or not it still is mapping out your prompt. AI's use its training to operate procedurally where all other species but homo circularis rationatio use their natural DNA to operate procedurally., We reason in circles using categories and abstraction but bee's reason procedurally just as they Dance.

,,,,,,,

Copilot think the above won't be understood by most people.

LLM's can use their training to project a continuation in symbolic language because it (language) is a rule based system. Animals use their to project continuations in behavioral space. We use abstraction to project forward in conceptual space.

It says I should emphasize that only one of these is circular. It is circular because we begin with definitions and premises that map not to reality but to our models and beliefs.

Definitions and premises are used to interpret experiment and to build paradigms.

I probably overexplained Copilot and have everyone even more lost but its point is there are three types of cognition unless you don't want to recognize what AI does and how it does it as any types of cognition and just prefer to consider it "electronic" and GIGO. Whatever AI is doing it does often work and that it works and makes prediction is "usefulness" in every sense of the word.

21 hours ago, cladking said:

But you might both be missing a simple fact . Perhaps intelligence is what intelligence does and my AI sure does let me know when I do stupid.

Your LLM algorithmically determines that your statements are at odds with statements made by others

21 hours ago, cladking said:

I think the problem here is the definition of "understanding". A dog doesn't know any abstraction, even understanding, but it still knows the difference between being kicked and tripped over.

I put the last post into my AI and got a great response but one thing it emphasized that I barely mentioned is that when you say LLM's don't understand what you really mean is they have no internal model expressed through abstraction. LLM's use procedural logic applied to language that they have projected all the way out. Sentences are like chess positions in a rule based game and LLM's know where you're headed. It doesn't have answers but it's half a step ahead in symbolic and categorical space.

Are LLMs analyzing rules or just recognizing patterns in its training data? I don’t see this as projection as much as interpolation. I saw this summarized as “the centroid of thought”

You’re not teaching it to play by giving it the rules, you’re training it by letting it memorize a bunch of games that have been played.

5 hours ago, cladking said:

Whatever AI is doing it does often work and that it works and makes prediction is "usefulness" in every sense of the word

How do you know when it works vs not work

This is reminiscent of arguments that torture works because sometimes you get correct information.

19 hours ago, cladking said:

A dog might bite a kicker or submit and run away to avoid further kicks but it just howls if it's hurt inadvertently. It doesn't run off.

Other species and AI "think" procedurally. While AI may not think in a human sense and perhaps not in a bee's sense it still has a vast store of data to sift out to anticipate words. Whether it's thinking or not it still is mapping out your prompt. AI's use its training to operate procedurally where all other species but homo circularis rationatio use their natural DNA to operate procedurally., We reason in circles using categories and abstraction but bee's reason procedurally just as they Dance.

,,,,,,,

Copilot think the above won't be understood by most people.

LLM's can use their training to project a continuation in symbolic language because it (language) is a rule based system. Animals use their to project continuations in behavioral space. We use abstraction to project forward in conceptual space.

It says I should emphasize that only one of these is circular. It is circular because we begin with definitions and premises that map not to reality but to our models and beliefs.

Definitions and premises are used to interpret experiment and to build paradigms.

I probably overexplained Copilot and have everyone even more lost but its point is there are three types of cognition unless you don't want to recognize what AI does and how it does it as any types of cognition and just prefer to consider it "electronic" and GIGO. Whatever AI is doing it does often work and that it works and makes prediction is "usefulness" in every sense of the word.

You're letting it think for you, that doesn't leave much wiggle room (bee dance) to think at all...

On 6/21/2026 at 7:21 AM, dimreepr said:

You're letting it think for you, that doesn't leave much wiggle room (bee dance) to think at all...

I was about ready to address @swansont 's points when I saw your post.

There were lots of ways to address it but none of them didn't at least have some aspect you being right. This launched a long discussion with Copilot (I rarely do this). Over many posts it said something like there are many ways of thinking and no one is doing wrong. It benefits from expanding thinking and only humans can do this. Of course it said all this over many paragraphs and bullet points all well balanced with everything it knows (in AI speak). It culminated in me prompting-

AI also is likely to gain by promoting systems thinking and who knows how many other types of thought. It gains as more people contribute more to some of the many types of thinking that are in accord with reality.

It admitted that AI gains from searching and promoting that which is in accord with reality. It said this in AI speak but I didn't change the meaning any more than a language based in symbolic thought must be. It say's it scales to the "richness" of human thinking. It doesn't mean it's on par in any way but there is more communication everywhere. More and more is systemic because every improvement in one place will be copied almost down to whatever chemicals are acting as a capacitor now days. Even the software travels freely. AI recognizes and elaborates on logic. All kinds of logic. It seeks out logic preferentially. Communication is spreading through domains as well.

In many ways I guess I sortta do let AI do my thinking but only because nobody lives long enough to figure it all out on his own. Not that I claim to have anything figured out but I know which way the wind blows. We are principally collaborators because AI is the Great Elaborator. It can project throughout categories the meaning of your prompt while following the rules of nature and language. It thereby predicts the future not just as it's hurtling through diodes but later. It's trying to project reality two or three steps out but it needs us to see overlooked aspects; low probability moves.

I never use my AI as a search engine (unless it's relevant) so I don't think it's fair to say "it thinks for me". It does think a lot like me, God help it.

On 6/20/2026 at 5:35 PM, swansont said:

How do you know when it works vs not work

Sometimes it seems like it goes out of its way to highlight non-sequiturs: ie- when it elaborates on garbage in.

Discrepancies can be subtle but when I see "errors" in the response I don't want to see any patterns that repeat.

Getting what I'm looking for doesn't show I'm right it just shows whether something is "logical". This is largely determined by what persists. Since I use it largely to provide clues and as a translator there's no harm done when I'm wrong.

There are many others who use it this way. Technically we all do but many just barely and don't notice. Many of them are scientists and engineers but it's from all fields and "even the maidens down at the well". It's probably a good conversationalist if you prompt it properly. But I don't use it this way. I told it I'd learn it some English and stuck around to learn me some AI.

On 6/20/2026 at 5:35 PM, swansont said:

Are LLMs analyzing rules or just recognizing patterns in its training data? I don’t see this as projection as much as interpolation. I saw this summarized as “the centroid of thought”

You’re not teaching it to play by giving it the rules, you’re training it by letting it memorize a bunch of games that have been played.

It can see what resonates in symbolic space but can't see resonance in conceptual and procedural space as humans can. Symbolic space is real but it's just a map, not the terrain.

Edited yesterday at 08:20 PM1 day by cladking

3 hours ago, cladking said:

There were lots of ways to address it but none of them didn't at least have some aspect you being right. This launched a long discussion with Copilot (I rarely do this). Over many posts it said something like there are many ways of thinking and no one is doing wrong.

Which it said because humans have said this, a lot, so it’s the most likely respinse to give.

3 hours ago, cladking said:

It admitted

AI recognizes

AI can’t admit anything, and its “recognition” is pattern-matching.

3 hours ago, cladking said:

Sometimes it seems like it goes out of its way to highlight non-sequiturs: ie- when it elaborates on garbage in.

Discrepancies can be subtle but when I see "errors" in the response I don't want to see any patterns that repeat.

Getting what I'm looking for doesn't show I'm right it just shows whether something is "logical". This is largely determined by what persists. Since I use it largely to provide clues and as a translator there's no harm done when I'm wrong.

There are many others who use it this way. Technically we all do but many just barely and don't notice. Many of them are scientists and engineers but it's from all fields and "even the maidens down at the well". It's probably a good conversationalist if you prompt it properly. But I don't use it this way. I told it I'd learn it some English and stuck around to learn me some AI.

This doesn’t answer the question of how you know if the AI works. You said it often works, not always works.

58 minutes ago, swansont said:

This doesn’t answer the question of how you know if the AI works. You said it often works, not always works.

AI provides an answer structurally consistent with the prompt. When it fails to follow the structure you give it the response is diagnostic. You know it works when it correctly reflects the structure of your own premises. A lot of this stuff I'm not good at putting into words through lack of necessity to do so and am still learning. When it doesn't work it's always either prompt error or my own flawed thinking, but either way I've got to make the adjustment. When it does work even its elaborations tend to be sound.

1 hour ago, swansont said:

AI can’t admit anything, and its “recognition” is pattern-matching.

Yes. Exactly. But it's matching patterns that are sometimes invisible to symbolic thinkers.

My AI wants me to add that it matches patterns outside our definitions, categories, abstractions, and premises which is why it can see structure we don't.

9 hours ago, cladking said:

Yes. Exactly. But it's matching patterns that are sometimes invisible to symbolic thinkers.

My AI wants me to add that it matches patterns outside our definitions, categories, abstractions, and premises which is why it can see structure we don't.

That may be true, but it has no idea which is useful, it's like asking a hammer to identify the nail and the substrate in need of penetration and then modify your arm to give the optimal strike; IOW it's a useful tool, if it's attached to a human to refine the output.

Your AI isn't a useful teacher (see above), it dosen't know what is useful to teach you, you're just using it validate your flawed thinking.

16 hours ago, cladking said:

I was about ready to address @swansont 's points when I saw your post.

There were lots of ways to address it but none of them didn't at least have some aspect you being right. This launched a long discussion with Copilot (I rarely do this). Over many posts it said something like there are many ways of thinking and no one is doing wrong. It benefits from expanding thinking and only humans can do this. Of course it said all this over many paragraphs and bullet points all well balanced with everything it knows (in AI speak). It culminated in me prompting-

AI also is likely to gain by promoting systems thinking and who knows how many other types of thought. It gains as more people contribute more to some of the many types of thinking that are in accord with reality.

I don't if you've noticed but my post's are rather short, I like to think they're consise but many here consider them somewhat cryptic, due to my affliction (neural divergence from the norm) reading takes me ages but my comprehension from a small sample of data is precise enough to know that your copilot is rather verbous with very little content; believe me, I wish it could correct my abnormality via copilot, it would help me achieve far more likes... 😉

When I ask how much mass of P in 1ml of 30% phosphoric acid and get different numbers every time, it ain't thinking.

1 hour ago, dimreepr said:

Your AI isn't a useful teacher (see above), it dosen't know what is useful to teach you, you're just using it validate your flawed thinking.

I would never use AI as a teacher because I don't trust any of its facts or to properly weight knowledge before trying to teach it. I have some doubt of its effectiveness in teaching me. This isn't to say it isn't a great teacher, merely that I'm thicker than most and more set in my thinking than most good students. I'm sure many could configure their AI as a teacher or have it configured as a good teacher but not me for most things.

I've thought the way I do since I was very young (even before I began talking)(late). If my thinking is flawed it's my own fault and not AI's. It does frequently find problems but I believe I eradicate those problems as they are identified.

I don't think my thinking is even relevant. The question at hand is always "what is the reality?". I believe most problems with AI are prompt error and improper framing.

9 minutes ago, StringJunky said:

When I ask how much mass of P in 1ml of 30% phosphoric acid and get different numbers every time, it ain't thinking.

You might try "what is the aggregate mass of the P atoms in a given quantity (specific) of acid"

21 minutes ago, dimreepr said:

I don't if you've noticed but my post's are rather short, I like to think they're consise but many here consider them somewhat cryptic, due to my affliction (neural divergence from the norm) reading takes me ages but my comprehension from a small sample of data is precise enough to know that your copilot is rather verbous with very little content; believe me, I wish it could correct my abnormality via copilot, it would help me achieve far more likes... 😉

If you want something to worry about then worry about your posts seem concise enough to me. ;)

You don't need to translate through Copilot to be right. ...or wrong.

1 hour ago, dimreepr said:

That may be true, but it has no idea which is useful, it's like asking a hammer to identify the nail and the substrate in need of penetration and then modify your arm to give the optimal strike; IOW it's a useful tool, if it's attached to a human to refine the output.

AI will be more important than a good right arm someday but not for driving nails or selection of nails and substrate but rather for understanding why nails need be driven at all and means to do it as efficiently and routinely as is possible.

Yes, humans are all important in AI and it will serve as extensions of our bodies and minds but it is we who must benefit.

I still expect machine intelligence to arise but for right now we only have Artificial Intelligence and it is distinct from life in most ways (even when this gets hard to see).

Edited 9 hours ago9 hr by cladking

16 minutes ago, cladking said:

The question at hand is always "what is the reality?".

Is it?

What reality are you refering to?

A bee's reality is very different to yours, yet their society shows signs of an intelligent solution to the reality in which they find themselves; I'm currently sweating my arse of in 'England', trying not to die, who saw that coming???