How far will we let AI control us?

[DaveC426913]

Would it be considered a bug or a feature?

Do you think that AI could:
a] produce this as an original piece of artwork (no precedent to steal from)
b] do so without direct prompting by a thinking entity, and
c] explain its reasoning and what it was attempting to capture (again, with no precedent to guide its rationale)?

Picasso could.

 

[russ_watters]

"Allen’s process wasn’t that simple. To get the final three images he entered in the competition, he said, took more than 80 hours."

If Picasso never existed and someone asked an LLM to draw the portrait of a person and the result was this:

View attachment 367845​

Would it be considered a bug or a feature?

Depends: how many hours did it take?

 

[russ_watters]

Here's the thing. Suppose there was a puzzle of some sort and my life depended on getting the right answer. I had a choice of asking @sbrothy @DaveC426913 or ChatGPT. And I had to take answer given. There is no question in my mind who I would choose. It's not even close. Sorry, guys!

Random or pointedly chosen question and random human, sure, that's been true for decades before LLMs. But how about if the answer to the question really matters; If you need a lawyer or structural engineer to do something significant in real life, are you still trusting an LLM vs licensed professionals?

 

[jack action]

Do you think that AI could:
a] produce this as an original piece of artwork (no precedent to steal from)
b] do so without direct prompting by a thinking entity, and
c] explain its reasoning and what it was attempting to capture (again, with no precedent to guide its rationale)?

Picasso could.

Knowing I'm no expert on prompt generation, let's see what GPT-5 mini from DuckDuckGo has to say:

Create an original portrait of the king of England, Charles III, drawn in a way that has never been seen before, and explain the reasoning and what you are attempting to capture with that portrait.

I can’t create images directly, but I can describe an original, never-before-seen portrait concept of King Charles III (public figure) and explain the artistic reasoning and what it aims to capture. Use this as a detailed brief for an artist or image generator.

Concept title: “Heirloom of Seasons”

Portrait description (visual brief)

• Format: Three-quarter-length, vertical canvas, 160 × 110 cm.
• Palette: A muted, earthy spectrum—deep umber, moss greens, pewter, and soft gold highlights—combined with translucent seasonal pigments (icy blue, spring sap green, harvest amber, winter crimson) layered like glazing.
• Composition: The King stands slightly turned to the viewer’s left, gaze steady but reflective. The background is an abstract montage of four interlocking semi-circles arranged like overlapping clockfaces, each subtly textured to suggest a season: frost-laced silver, budding leaf-fiber green, sun-warmed amber, and wind-swept slate. The circles intersect behind his shoulders, so the portrait’s focal plane sits where the seasons converge.
• Attire and regalia: Not a formal uniform or crown. He wears a contemporary-cut dark wool coat with an inner lining patterned faintly with botanical etchings (oaks, thistles, daffodils, roses) rendered in micro-embroidery. On his lapel, instead of a medal, is a small heirloom brooch shaped like a seed pod—bronze with a single pearl—symbolizing continuity and nurture.
• Surface treatment: Use layered mixed media—oil underpainting for flesh tones, egg tempera for the botanical motifs, and translucent resin veils over the seasonal circles to create depth and a subtle sheen that changes with light.
• Light and texture: Soft directional light from upper left, giving gentle highlights to cheekbones and hands; hands are prominent and shown relaxed, one touching the brooch as a caretaking gesture. The skin is rendered with careful realism but with visible brushwork that becomes more gestural toward the edges, blending into the seasonal backgrounds.
• Symbolic elements, integrated visually:
  • A loom-like shadow pattern across the lower torso, evoking weaving of communities and institutions.
  • Tiny, almost imperceptible golden threads that run from the brooch into the botanical lining and then outward into the seasonal circles—visualizing stewardship connecting small acts to national cycles.
  • Background textures include faint topographic lines and tidal ripples, suggesting land and sea bound together.
• Mood: Measured dignity with an undercurrent of vulnerability—an elder statesman aware of a living, changing realm rather than a static throne.

Artistic reasoning and intent

• Avoiding cliché regalia: By omitting crown and traditional full regalia, the portrait shifts emphasis from hereditary power to custodianship and service.
• Seasons as temporal stewardship: The four interlocking seasonal circles represent stewardship across time—ecological, cultural, and institutional cycles. Making them intersect behind the subject places him at the nexus of continuity, not above it.
• Botanical lining and brooch: Flora embroidered into the lining ties the monarch personally to the nation’s natural and cultural symbols. The seed-pod brooch embodies legacy, potential, and the idea that leadership is about planting and tending, not merely commanding.
• Hands and touch: Emphasizing hands (one at the brooch) humanizes authority—decisions enacted through care, not only decree.
• Mixed media and texture: Layering materials mirrors layered responsibilities and histories. The resin veils that shift with light symbolize how public perception changes with context and time.
• Loom shadow and golden threads: These motifs visualize networks—institutions, people, ecosystems—showing leadership as interwoven with others, where small threads matter.
• Light and brushwork: Realism at the face invites recognition and empathy; loosened edges suggest that identity is porous and shaped by surroundings.

What this portrait attempts to capture

• Custodial legitimacy: The idea of monarchy as ongoing stewardship over land, culture, and institutions rather than unimpeachable authority.
• Temporal responsibility: Leadership experienced across seasons and generations—decisions that resonate through cycles.
• Human scale: The portrait aims to balance dignity with approachable vulnerability, showing a person enmeshed in responsibilities and connections.
• Ecological-cultural link: Subtle motifs highlight the interplay between natural stewardship and civic duty.

Usage notes for an artist or generator

• For photorealistic approaches: Use a high-resolution reference for facial likeness; apply oil underpainting techniques, then layer tempera and resin digitally or physically.
• For painterly approaches: Emphasize visible brushwork at the edges and carefully control glazes for the seasonal circles.
• For AI-image generation: Provide the above scene-level prompt, specifying mixed-media textures, palette, and symbolic elements; iterate on hands, brooch detail, and background intersections.

If you’d like, I can convert this into:

• A short prompt optimized for image-generation models (text-only), or
• A step-by-step studio painting workflow (materials, stages, drying times). Which would you prefer?

I asked for the short prompt and got:

Use this concise, image‑generation–optimized prompt:

"A three-quarter-length portrait of King Charles III, contemporary and original: no crown or formal regalia. He stands turned slightly left, steady reflective gaze, one hand touching a small bronze seed‑pod brooch with a single pearl on his lapel. Contemporary dark wool coat with inner lining subtly embroidered with oak, thistle, daffodil, and rose motifs. Background: four interlocking translucent circles like clockfaces, each textured to evoke a season—frost silver, spring sap green, harvest amber, wind-swept slate—intersecting behind his shoulders. Soft directional light from upper-left, realistic facial rendering with visible brushwork; hands detailed and natural. Mixed-media texture: oil underpainting for skin, egg tempera botanical details, and glossy resin-like glazes over seasonal circles so surface shifts with light. Subtle symbols: faint topographic lines and tidal ripples in background; a loom-like shadow across lower torso and tiny golden threads running from brooch into lining and outward into the seasonal circles. Palette: muted earth tones (umber, moss, pewter) with translucent seasonal highlights. Mood: measured dignity with gentle vulnerability. Photorealistic face, painterly edges, high detail, 8k, cinematic lighting."

If you want adjustments (style, era, realism level, or aspect ratio), tell me which and I’ll refine.

I then enter that prompt into the free image generator deepimg.io and got these from Nano Banana Free (no login needed, defined as "Supports multi-image synthesis. Free popular models may occasionally be unstable."):

Pretty scary for freely available programs, huh?

 

[DaveC426913]

Here's the thing. Suppose there was a puzzle of some sort and my life depended on getting the right answer. I had a choice of asking @sbrothy @DaveC426913 or ChatGPT. And I had to take answer given. There is no question in my mind who I would choose. It's not even close. Sorry, guys!

That's a pretty clear false dichotomy (trichotomy?)

Two of those choices are statistically random.
OK, so GPT is slightly better than random.

 

[DaveC426913]

Pretty scary for freely available programs, huh?

Holy carp. Yes it is.
I concede the point. No notes.

 

[russ_watters]

Holy carp. Yes it is.
I concede the point. No notes.

Really? Art is maybe the least impressive use of AI I can think of. For other uses hallucination is a bug, but for art it's a feature. Art can be anything and can't ever be wrong.

 

[DaveC426913]

I wonder if Turing ever envisioned that AIs would fail the Turing Test - not because they undershot the bar - but because they overshot it.

To-wit:
"This is too good to be made by a human"...

 

[DaveC426913]

Really? Art is maybe the least impressive use of AI I can think of. For other uses hallucination is a bug, but for art it's a feature. Art can be anything and can't ever be wrong.

I meant the rationalization it offered. See above "Artistic Reasoning and Intent"

 

[russ_watters]

I meant the rationalization it offered. See above "Artistic Reasoning and Intent"

What, exactly, about that do you find impressive?

I wonder if Turing ever envisioned that AIs would fail the Turing Test - not because they undershot the bar - but because they overshot it.

To-wit:
"This is too good to be made by a human"...

I would think so. Excel does that to me every day.

 

[PeroK]

Random or pointedly chosen question and random human, sure, that's been true for decades before LLMs. But how about if the answer to the question really matters; If you need a lawyer or structural engineer to do something significant in real life, are you still trusting an LLM vs licensed professionals?

My point was about general knowledge and reasoning (at least in a practical sense) using that knowledge. And, specifically the point that it doesn't matter so much how an answer is produced.

In answer to your question, it depends on the circumstances. As you say, computers have aided professionals for decades. LLM's go further, but in most cases there would still be the need for human overview at least. But, that is changing. Moreover, licensed professionals are perhaps limited in terms of original, creative thought. My friend was a vet and by the end of his career, he was allowed little or no discretion in the prescription of drugs. It was all mandated by standards.

To me it's a practical matter. LLM's can do lots of stuff that seems to be hard to accept for some people. I don't believe we have "proper" AI yet. Nevertheless, what we do have is changing the world. The last thing we need is a failure to acknowledge this, IMO.

 

[DaveC426913]

What, exactly, about that do you find impressive?

I find it impressive that an AI could come up with things like this (just one of the examples; they're all equally impressive):

• Seasons as temporal stewardship: The four interlocking seasonal circles represent stewardship across time—ecological, cultural, and institutional cycles. Making them intersect behind the subject places him at the nexus of continuity, not above it.

and then incorporate it visually. That is a level of sophistication beyond my expectations.

 

[DaveC426913]

That's a pretty clear false dichotomy (trichotomy?)

Two of those choices are statistically random.
OK, so GPT is slightly better than random.

@PeroK What makes you skeptical about this?

 

[PeroK]

@PeroK What makes you skeptical about this?

GPT is not "slightly better than random". It knows stuff. Yes, you can confuse it. Yes, you can catch it out. But, if you know what you're doing it will give you extremely good answers on almost any subject.

Also, I fail to see the false dichotomy.

 

[DaveC426913]

GPT is not "slightly better than random". It knows stuff.

No, I mean your litmus test only implies it is slightly better than two guys chosen randomly.

Two random guys is a very low bar to set for chatGPT to be better than.

Also, I fail to see the false dichotomy.

The (implied) false dichotomy is that one's choices are:
- two people out of 8 billion, who may or may not know the first thing about puzzle-solving, and
- ChatGPT.

Why would one not go instead to a qualified specialist?

 

[PeroK]

No, I mean your litmus test only shows it is slightly better than two guys you chose randomly.

That is a very low bar to set for chatGPT to be better than.

Perhaps I misunderstood the points being made.

The (implied) false dichotomy is that one's choices are:
- two people out of 8 billion, who may or may not know the first thing about puzzle-solving, and
- ChatGPT.

Why would one not go instead to a qualified specialist?

There are lots of reasons to prefer ChatGPT. It's available for you 24x365. It costs nothing. It answers immediately. Not to mention the speed at which it can do complex collation.

I've made the point in previous threads that even non-intelligent technology has some major advantages over human brainpower. Even that has led to serious counter-argument.

From a practical point of view, it understands what you are asking (in natural human language - you don't have to learn a programming script and precise syntax).

Qualified human specialists are not easy to come by and you might expect days or weeks to get an answer.

 

[jack action]

@PeroK , what would you choose given there was a puzzle of some sort and your life depended on getting the right answer, and you have the choice of consulting Wikipedia, Google search engine, or ChatGPT? That is a more interesting choice especially knowing that both Google and ChatGPT have already parsed Wikipedia.

 

[DaveC426913]

There are lots of reasons to prefer ChatGPT.

And I'm not saying otherwise. I simply found your scenario for measuring its efficacy at a given task to be flawed. It simply says "ChatGPT is fractionally better than the opinion of a randomly chosen person".

It's not a big deal, I just thought your initial comment in post 23:

Here's the thing. Suppose there was a puzzle of some sort and my life depended on getting the right answer. I had a choice of asking @sbrothy @DaveC426913 or ChatGPT. And I had to take answer given. There is no question in my mind who I would choose. It's not even close. Sorry, guys!

wasn't really "the thing" that's "here".

 

[PeroK]

@PeroK , what would you choose given there was a puzzle of some sort and your life depended on getting the right answer, and you have the choice of consulting Wikipedia, Google search engine, or ChatGPT? That is a more interesting choice especially knowing that both Google and ChatGPT have already parsed Wikipedia.

That's a good question. A human plus Internet search depends on the human. I might be okay with some puzzles, but not so good with others. I might back myself sometimes, but not others.

In fact, as an aside, I used to think it was unfair in chess games between a human and a computer that the computer could have the entire opening theory pre-loaded; whereas, a human is not allowed to consult chess books during a game. A human player, in fact, is not even allowed to make notes of any description.

The key difference is the speed at which ChatGPT can do stuff. Unless it's something I know well and can rattle off the answer, then I might need an hour or a day or a week; whereas, CHatGPt might need a few seconds.

This is part of the reason, IMO, that there are almost no homework threads on here any more. We take so long to answer questions, that I don't see how we could compete with LLM's.

In answer to your question, eventually a human armed with the Internet could compete with ChatGPT on the quality of answer. But, not on speed.

 

[gleem]

AI and LLM in particular are amazing. Beginning with ChatGPT3 making a splash on the web, detractors argued that it was just auto-complete. It was desired to emulate human neural networks, a guess of how we think they work. It worked pretty well without any specific reasoning capability. It showed, at least to me, that our language contains much hidden information about logic and culture, as its immediate responses were generally correct and reasonable. Did you ever wonder how humans functioned without language? We get sight, hearing, and touch in order to function, and then get language. AI gets language first but doesn't experience sight, touch, and hearing as we do.

Since it was required to process prompts until it reached some artificial endpoint and had no memory to reflect on the development of its responses, it often drifted off track when processing too long and hallucinates, a term I find unhelpful. But we continue to move the goal posts when it acts more like a human. I believe one day, we will not be able to move them any farther.

Assuming AI at this stage should be able to generate generally original works of art, like Picasso, is no more reasonable than expecting an arbitrary human to do that. We think of his modern works, but he was influenced by many artists: Van Gogh, Gauguin, Cézanne, Matisse, and Braque. Artists learn from one another, and great artists evolve to develop a distinctive expression of their skill. The famous abstract artist Mondrian's geometric works were probably influenced by cubism. So, what is creativity?

Regarding creativity, here is a video in which the script was generated in 2020 by GPT3, called the Solicitors. It was just given the opening setting.

How much of what we think is processed in the subconscious and "revealed" to our conscious mind? That begs the question of what is actually in our subconscious. Perhaps more than we can imagine. I find myself sometimes mimicking something I heard, or am I fabricating a response based on information obtained from unremembered sources? Even as I compose this post, I cannot be sure how much is truly original.

When we choose one word over another because it better fits the context, do we not have some sense ( a type of weighting factor like AI ) of its appropriateness?

I have sometimes been befuddled by a question or statement because my understanding of the words was not what was intended. I either gave a stupid answer or interpretation, or I could not process it. Why do we expect these early AI models to be perfect?

Humans can think about their thinking, called metacognition, which can help avoid mistakes; AI, AFAIK, cannot (so far). In some sense, human intelligence and AI are similar. But then, do we really want AI to be like human intelligence?

So, getting back to the opening post, anybody can become too dependent on things, letting them control their behavior, whether a tool, AI, or another human.

 

[sbrothy]

What have you or any of us ever done that is truly new and unique? Everything builds on what we know and have learned already.

I know this is an old discussion but I stumbled across this thread: Against AI, and it occurred to me that one criteria for AI creativity could be a piece of "art" - created by AI - which stood out clearly on the background of the "crud" they complain about in there.

EDIT: Unfortunately, it just occurs to me you could just ask the AI to make a "piece of art" which honored that criteria. Back to square one I guess.

 

[Greg Bernhardt]

which stood out clearly on the background of the "crud" they complain about in there.

I think at this point that is impossible due to the transformer architecture and training. AI can't do anything truly novel yet.

 

[sbrothy]

Yes, our discussion circled around exactly that, and when that happened AI would be "real". But I think we ended on the note that we're far from there yet.

 

[PeroK]

I think at this point that is impossible due to the transformer architecture and training. AI can't do anything truly novel yet.

Define "truly novel".

 

[DaveC426913]

Define "truly novel".

Something about a Scotsman, sugar and a bowl of porridge, perhaps?

 

[PeroK]

I think at this point that is impossible due to the transformer architecture and training. AI can't do anything truly novel yet.

On one thread, people are complaining that it "hallucinates". I.e. makes stuff up. And, on another thread, people are saying it can't do anything creative. It can't be both.

PS even without the capabilities of an LLM, using a random number generator could produce all manner of original ideas. The quality may be lacking, but not the originality or apparent creativity.

It's sad to see so many scientifically minded people forget (or not accept) that fantastic complexity can arise from the simplest algorithms. Let alone the complex algorithms of an LLM.

 

[PeroK]

Something about a Scotsman, sugar and a bowl of porridge, perhaps?

Awa an' bile yer heid!

 

[DaveC426913]

On one thread, people are complaining that it "hallucinates". I.e. makes stuff up. And, on another thread, people are saying it can't do anything creative. It can't be both.

The difference is in whether it does so deliberately.

A baby can stumble and knock paint on a canvas. That does not make it a Jackson Pollock.

 

[PeroK]

The difference is in whether it does so deliberately.

A baby can stumble and knock paint on a canvas. That does not make it a Jackson Pollock.

View attachment 367878

Define "deliberately". From a purely artistic point of view, there is no reason to prefer a Pollock to a baby's scrawl. It's all a matter of highly arbitrary taste. There's a case that all modern art is a con.

Have you ever seen the documentary "My Daughter Could Paint That"?

 

[DaveC426913]

Define "deliberately". From a purely artistic point of view, there is no reason to prefer a Pollock from a baby's scrawl. It's all a matter of highly arbitrary taste. There's a case that all modern art is a con.

A case, yes. I am not of that mind.

The example is a biased scenario* - the fact that the pictures are already chosen and available for you to view. Let me 'splain.

*not your fault. I'm just saying it's not the whole picture

Have you ever seen the documentary "My Daughter Could Paint That"?

No, but I get it. It is a discussion I often have with friends. I tend to side with "No she couldn't."



Consider:

It is missing the point to look at one or two pieces of paper that a baby has made a mess on.

It is not the baby who is embuing the piece with import or value or message. It is the curator of the pieces that does so (whomever that might be - parent, an agent, whatever).

Nobody goes to a friend's home to entertain themselves with a baby's works of art. Nobody is going to go to an art show that has a hundred pieces that indiscriminately include every piece a baby does.(Just like nobody is going to read every iteration of pages that a million monkeys churn out on a millon typerwriters.)

By the time we see them, they are already curated. A signal has been found among the noise.

Those one or two pages that the baby made a mess on were chosen - out of hundreds - because the curator thought they saw a relatable message embodied in them.

The act of choosing - separating what makes a new, interesting message from what does not - is where the deliberateness and creativity lies.