Using AI to critique your photos

[nikos79]

But it can't drive a Tesla without getting into an accident? Maybe, it's only good at certain things like processing trillions of math calculations quickly so it can solve unsolvable equations. Or learn to play great chess because it can go through all the permutations of possible moves quickly and select the best one. Computers have done that in the past. But that's not creativity. That's not innovation. AI seems to be the next computer chip. Faster, more circuits, just a higher speed idiot moving ones and zeros around. It can't think.

I agree with Tesla but the real world is not easily simulated. That is why I mentioned "immersion" for AGI. By the way autonomous driving in any environment is largely linked to AGI due to how challenging of a problem it is.
But for the math problems you are wrong, it didn't solve them with calculations it solved them with reasoning that's why it was so impressive.
As for chess you are right, although you are missing a crazy point. The computers who learnt to defeat humans and redefined whole chess openings from scratch were trained by playing only with themselves only by knowing the set of rules and nothing more. They also achieved it in half an hour. From zero to mastery. That goes beyond computations. This is mind-blowing

 

[MurrayMinchin]

AI side note:

My wife searched "reindeer" in Mac Photos and this came up...our Yorkie wearing a Christmas sweater sporting antlers.

Close, but...

 

[koraks]

it solved them with reasoning

We may have to discuss what 'reasoning' is - and what isn't. Not all association is reasoning in my book. If you said "association", then I would have agreed.

 

[warden]

I didn't say AI wasn't good at being a tool to help me to be more creative and knowledgeable. However, it isn;'t creative, but more something that mimics the creativity of humans.

You're right that it's still very new, and I haven't used it much. I have asked it questions and it does seem easier to use than Google search. It's very helpful in giving me answers to question,s although YouTube videos work better when I have to repair something. It helps give me ideas to questions I hadn't thought of sort of like a reference source. So for right now, I see it as a tool to aid me in being creative and to answer questions more easily. But it's not the thing being creative. I am.

Maybe I'm missing something. What do you think it is?

It sounds like you're using AI as a search engine for the most part, and as such it's certainly an outstanding one. I partly use it that way as well. And of course it's not human, which is as it should be, and why it's such an effective co-conspirator; It doesn't get tired, frustrated, distracted or drunk like humans do. (And that's also why autonomous vehicles are already safer than human drivers even though the systems are in their infancy.) Use AI with the prompt "Are autonomous vehicles safer than vehicles driven by humans, and if so, why. Please cite your sources."

It's not innovative or creative because it doesn't really have a mind.

I suppose someone is about to break out their dictionary and look up these words but my experience has been that AI is both wildly innovative and creative. It creates. It innovates. It doesn't need a mind for that, even though it can't do its tasks without you and leveraging outside resources. You, as the human interpreter and managing partner, can use its output to steer it and yourself to accomplish goals. You may not experience this if your use of AI is basically a super-search engine, but if you branch out from there I imagine you will.

 

[nikos79]

We may have to discuss what 'reasoning' is - and what isn't. Not all association is reasoning in my book. If you said "association", then I would have agreed.

Very good point as usual.
It is not associations but I see what you mean it is not reasoning in the sense that they know what they are doing or applying logic as humans do.
It is more a “closed system” you give it the formal rules and then by applying them it can lead to conjectures or theorems sort of as called “symbolic reasoning”

 

[MattKing]

But it can't drive a Tesla without getting into an accident?

The same might be said by the population of human drivers that have already tried, and definitely failed.

 

[Pieter12]

But it can't drive a Tesla without getting into an accident? Maybe, it's only good at certain things like processing trillions of math calculations quickly so it can solve unsolvable equations. Or learn to play great chess because it can go through all the permutations of possible moves quickly and select the best one. Computers have done that in the past. But that's not creativity. That's not innovation. AI seems to be the next computer chip. Faster, more circuits, just a higher speed idiot moving ones and zeros around. It can't think.

The tesla lacks the appropriate sensors for it to really drive autonomously. AI is the least of its problems.

 

[warden]

The tesla lacks the appropriate sensors for it to really drive autonomously. AI is the least of its problems.

We have autonomous Waymo Jaguars running around the city now. It’ll be interesting to see how well they do. Here It’s chaotic driving for sure. Anarchy, really. It will be a good test for a Waymo.

 

[Pieter12]

We have autonomous Waymo Jaguars running around the city now. It’ll be interesting to see how well they do. Here It’s chaotic driving for sure. Anarchy, really. It will be a good test for a Waymo.

We have had Waymos for quite a while here. The biggest problem (besides all the noise at the charging depot) is they pretty much strictly obey traffic rules unlike human drivers, often holding up traffic to make a left turn--they won't go into an intersection if the car hasn't determined it can safely clear it before the light turns red.

 

[Alan Edward Klein]

Everyone well-versed in autonomous systems knows that the failure rate of such systems is generally far below that of human-controlled systems. That's why human intervention in those systems is generally minimized. Look at plane crashes, nuclear power plant meltdowns etc. etc. - what goes wrong is the human factor, virtually without fail. It's not the autopilot that crashes the plane. It's the pilot deciding to disengage the autopilot because he thinks it's wrong.
Now, those Teslas may not be ripe for the roads - I really wouldn't know. All I do know is that they either potentially or already in actuality will drive safer than you, me or any other human being. Thinking otherwise is utterly delusional - but that's mankind for you: hubris.

But they don't drive safer, and they've been at it for years. And driving seems to be less than trying to create something or be innovative, where AI has to "think" outside given parameters, as there are with driving.

 

[Alan Edward Klein]

We have autonomous Waymo Jaguars running around the city now. It’ll be interesting to see how well they do. Here It’s chaotic driving for sure. Anarchy, really. It will be a good test for a Waymo.

Do they run on rainy or snowy days when the street lines are covered? Night? Are there other limitations of where, when etc they can;t run?

 

[Pieter12]

Do they run on rainy or snowy days when the street lines are covered? Night? Are there other limitations of where, when etc they can;t run?

Well, with the recent major blackout in San Francisco, they all came to a halt. Seems they couldn't deal with traffic lights being out among other things. And as of now, they aren't on the freeways except with a driver behind the wheel.

 

[Alan Edward Klein]

Well, with the recent major blackout in San Francisco, they all came to a halt. Seems they couldn't deal with traffic lights being out among other things. And as of now, they aren't on the freeways except with a driver behind the wheel.

What about when it rains or snows?

 

[Pieter12]

Rain is no problem. Can’t say about snow, it doesn’t snow here.

 

[Alan Edward Klein]

Rain is no problem. Can’t say about snow, it doesn’t snow here.

Well it snows in a big part of the country. What happens there? Can these programs navigate with obscured lane lines, crosswalks, low visibility, etc? Or do you need a dog sled?

 

[Pieter12]

Well it snows in a big part of the country. What happens there? Can these programs navigate with obscured lane lines, crosswalks, low visibility, etc? Or do you need a dog sled?

It could well maybe able to manage snow OK. And if not, it certainly isn’t supposed to be for everyone everywhere.

 

[Sirius Glass]

With all due respect Sirius, but your mention of 'heuristic' software in the other AI thread as well as what you've told about your work before makes it clear that your experience is in something entirely different from neural-network based models. This is a different animal from the ground up.

Regardless of the learning principles based on which examples and their order remain the same. AI can only be as good as its learning material and order of its presentation.

 

[koraks]

Regardless

No. You'll have to do some reading. A lot, in fact.

 

[nikos79]

Regardless of the learning principles based on which examples and their order remain the same. AI can only be as good as its learning material and order of its presentation.

There is a small truth to that in the old saying in the AI world “garbage in-garbage out” but the misconception that it is only as good as its data seems to linger even in many of my programmer colleagues.

Let’s do a small test. I asked the free non latest version of ChatGPT the following:

“Can you imagine a photo camera that works with a new technology neither analog nor digital?”

I assume that this was not present or formulated in the data as it is and it had to be some new original synthesis among the patterns. Here was the output (sorry for its length):



The “Cognitive Camera”

• Capture principle: Instead of using film or pixels, the camera records light and spatial information as patterns of neural-encoded thought — essentially storing how the scene feels and interacts with attention.
• Image formation: The photographer “frames” the scene with intention and gaze, and the camera interprets the visual field along with context, emotion, and conceptual focus.
• Output: The resulting image is not a static photo but a dynamic projection: the captured scene can shift depending on who is viewing it, highlighting details differently for each observer, almost like a living photograph.
• Controls: No shutter speed or ISO; instead, you “tune” your mental focus, adjust conceptual parameters (contrast of ideas, emotional depth), and the camera interprets these inputs into the light-matter interaction.
• Special feature: The camera can merge multiple photographers’ perspectives, combining their intentions into a composite “multi-conscious” photograph — almost like a collaborative dream image

Why it’s plausible (science-fiction-meets-future-tech)

1. Quantum / light field sensing: Could theoretically measure phase, polarization, and coherence patterns of light beyond classical imaging.
2. AI / brain-computer interface: Could integrate intentional patterns from a human mind to guide image formation.
3. Dynamic display: Using augmented reality, holography, or light-field projection, the “photo” could respond to the observer, unlike any analog or digital still image

 

[Alan Edward Klein]

There is a small truth to that in the old saying in the AI world “garbage in-garbage out” but the misconception that it is only as good as its data seems to linger even in many of my programmer colleagues.

Let’s do a small test. I asked the free non latest version of ChatGPT the following:

“Can you imagine a photo camera that works with a new technology neither analog nor digital?”

I assume that this was not present or formulated in the data as it is and it had to be some new original synthesis among the patterns. Here was the output (sorry for its length):



The “Cognitive Camera”

• Capture principle: Instead of using film or pixels, the camera records light and spatial information as patterns of neural-encoded thought — essentially storing how the scene feels and interacts with attention.
• Image formation: The photographer “frames” the scene with intention and gaze, and the camera interprets the visual field along with context, emotion, and conceptual focus.
• Output: The resulting image is not a static photo but a dynamic projection: the captured scene can shift depending on who is viewing it, highlighting details differently for each observer, almost like a living photograph.
• Controls: No shutter speed or ISO; instead, you “tune” your mental focus, adjust conceptual parameters (contrast of ideas, emotional depth), and the camera interprets these inputs into the light-matter interaction.
• Special feature: The camera can merge multiple photographers’ perspectives, combining their intentions into a composite “multi-conscious” photograph — almost like a collaborative dream image

Why it’s plausible (science-fiction-meets-future-tech)

1. Quantum / light field sensing: Could theoretically measure phase, polarization, and coherence patterns of light beyond classical imaging.
2. AI / brain-computer interface: Could integrate intentional patterns from a human mind to guide image formation.
3. Dynamic display: Using augmented reality, holography, or light-field projection, the “photo” could respond to the observer, unlike any analog or digital still image

Ask it what sources it used to make these conclusions. Maybe these have already been suggested in the literature it was trained on.

 

[Alan Edward Klein]

It could well maybe able to manage snow OK. And if not, it certainly isn’t supposed to be for everyone everywhere.

But the idea is that AI will be more advanced than special humans, certainly at least as good as regular humans. If it can't drive in the snow, or in other street conditions we run into ordinarily, what do you get with AI?

 

[warden]

If it can't drive in the snow, or in other street conditions we run into ordinarily, what do you get with AI?

Safety. Sobriety. Efficiency. Lack of road rage. Presence of 360 degree vision. Scanning for hazards better and faster than two elderly human eyes looking ahead. No macular degeneration or cataracts. Faster reflexes. Lack of distracted driving while checking your phone. More skill than the average 16 year old with a fresh license. Lack of drag racing. Lack of tailgating. An inability to run from the cops. Inability to drive while fatigued or fall asleep behind the wheel.

should I go on? Is your question even serious, Alan?

 

[Alan Edward Klein]

Safety. Sobriety. Efficiency. Lack of road rage. Presence of 360 degree vision. Scanning for hazards better and faster than two elderly human eyes looking ahead. No macular degeneration or cataracts. Faster reflexes. Lack of distracted driving while checking your phone. More skill than the average 16 year old with a fresh license. Lack of drag racing. Lack of tailgating. An inability to run from the cops. Inability to drive while fatigued or fall asleep behind the wheel.

should I go on? Is your question even serious, Alan?

Many of those things are lacking or not up to par. Here's AI's opinion.https://www.google.com/travel/flights?client=ms-android-verizon-us-rvc3&hs=BRTU&sca_esv=31db8cbb946c4eae&tbm=flm&sxsrf=ANbL-n7P4-snY40mit6so4CyCeYAHCZoqw:1768071951940&source=lnms&ved=1t:200715&ictx=111&q=What issues do automatic driving cars have



Automatic driving cars currently face a variety of significant issues spanning technology, safety, regulation, and social impact
. These challenges are hindering their widespread, fully autonomous deployment.

Technological and Safety Limitations

• Sensor and Software Errors: Cars rely on complex sensor and perception systems that can experience software glitches or errors, leading them to do the wrong thing at the wrong time.
• Difficulty with Complex Environments: Autonomous vehicles struggle with real-world complexities and "edge cases" such as construction zones, unclear lane markings, and unexpected obstacles or detours that human drivers might navigate easily.
• Adverse Weather Conditions: Snow, ice, and heavy rain can interfere with sensors, reducing the car's ability to detect its surroundings accurately.
• "Phantom Braking" and Detection Issues: Vehicles may perform abrupt, unnecessary stops (known as "phantom braking") in response to non-existent objects, which can cause rear-end collisions with human drivers. They also have issues correctly detecting pedestrians and cyclists, especially at dawn or dusk.
• Interaction with Unpredictable Human Behavior: Autonomous systems typically predict average behavior and struggle with the "not-average" and sometimes aggressive maneuvers of human drivers, which can lead to misjudgments in traffic.
• Need for Human Intervention: Many current systems are semi-autonomous and require the human driver to remain attentive and ready to take control. A false sense of security can lead to driver distraction and fatal accidents when a rapid transition of control is needed.

Cybersecurity, Legal, and Regulatory Issues

• Vulnerability to Hacking: As connected computers on wheels, self-driving cars are vulnerable to cybersecurity threats. A breach could allow remote control of the vehicle or access to personal data, with catastrophic consequences.
• Lack of Clear Regulation: There is currently no consistent and comprehensive federal, state, or local regulatory requirement for companies to prove their cars are reasonably safe before deployment. This uncertainty slows development and makes accountability difficult.
• Liability and Insurance Questions: In the event of an accident, it is often unclear who is at fault—the manufacturer, software developer, vehicle owner, or a passenger. This complicates legal processes and insurance frameworks.

Societal and Economic Concerns

• Job Displacement: The widespread adoption of autonomous vehicles could disrupt entire industries, leading to significant job losses for professional drivers, such as truckers, taxi, and rideshare operators.
• Equity and Access: The initial cost of the technology may mean benefits are exclusively available to the wealthy, potentially worsening existing transportation inequities.
• Increased Travel: By making commuting time more productive (e.g., allowing work or leisure during travel), AVs might actually encourage longer commutes and increase total vehicle miles traveled, potentially exacerbating congestion and environmental pollution unless paired with clean energy mandates.

Additional inquiries:

What progress is being made on the legal and regulatory front for autonomous vehicles?
What technologies are being developed to improve self-driving car performance in bad weather?
Explain more about 'phantom braking'
Potential Problems with Self-Driving Cars

 

[warden]

Many of those things are lacking or not up to par. Here's AI's opinion.https://www.google.com/travel/flights?client=ms-android-verizon-us-rvc3&hs=BRTU&sca_esv=31db8cbb946c4eae&tbm=flm&sxsrf=ANbL-n7P4-snY40mit6so4CyCeYAHCZoqw:1768071951940&source=lnms&ved=1t:200715&ictx=111&q=What issues do automatic driving cars have



Automatic driving cars currently face a variety of significant issues spanning technology, safety, regulation, and social impact
. These challenges are hindering their widespread, fully autonomous deployment.

Technological and Safety Limitations

• Sensor and Software Errors: Cars rely on complex sensor and perception systems that can experience software glitches or errors, leading them to do the wrong thing at the wrong time.
• Difficulty with Complex Environments: Autonomous vehicles struggle with real-world complexities and "edge cases" such as construction zones, unclear lane markings, and unexpected obstacles or detours that human drivers might navigate easily.
• Adverse Weather Conditions: Snow, ice, and heavy rain can interfere with sensors, reducing the car's ability to detect its surroundings accurately.
• "Phantom Braking" and Detection Issues: Vehicles may perform abrupt, unnecessary stops (known as "phantom braking") in response to non-existent objects, which can cause rear-end collisions with human drivers. They also have issues correctly detecting pedestrians and cyclists, especially at dawn or dusk.
• Interaction with Unpredictable Human Behavior: Autonomous systems typically predict average behavior and struggle with the "not-average" and sometimes aggressive maneuvers of human drivers, which can lead to misjudgments in traffic.
• Need for Human Intervention: Many current systems are semi-autonomous and require the human driver to remain attentive and ready to take control. A false sense of security can lead to driver distraction and fatal accidents when a rapid transition of control is needed.

Cybersecurity, Legal, and Regulatory Issues

• Vulnerability to Hacking: As connected computers on wheels, self-driving cars are vulnerable to cybersecurity threats. A breach could allow remote control of the vehicle or access to personal data, with catastrophic consequences.
• Lack of Clear Regulation: There is currently no consistent and comprehensive federal, state, or local regulatory requirement for companies to prove their cars are reasonably safe before deployment. This uncertainty slows development and makes accountability difficult.
• Liability and Insurance Questions: In the event of an accident, it is often unclear who is at fault—the manufacturer, software developer, vehicle owner, or a passenger. This complicates legal processes and insurance frameworks.

Societal and Economic Concerns

• Job Displacement: The widespread adoption of autonomous vehicles could disrupt entire industries, leading to significant job losses for professional drivers, such as truckers, taxi, and rideshare operators.
• Equity and Access: The initial cost of the technology may mean benefits are exclusively available to the wealthy, potentially worsening existing transportation inequities.
• Increased Travel: By making commuting time more productive (e.g., allowing work or leisure during travel), AVs might actually encourage longer commutes and increase total vehicle miles traveled, potentially exacerbating congestion and environmental pollution unless paired with clean energy mandates.

Additional inquiries:

What progress is being made on the legal and regulatory front for autonomous vehicles?
What technologies are being developed to improve self-driving car performance in bad weather?
Explain more about 'phantom braking'
Potential Problems with Self-Driving Cars

Your link opens a page for Google travel services. I’m going to take a break from this thread.

 

[cowanw]

Back on topic, it seems the one thing that AI has not scraped is any actual information on proper critiquing of photograph. Rather than using a model such as "Criticizing Photographs by Terry Barrett they seem to have taken scraped information like might be found on Flickr or here as examples of critique.
What is worse, it really does seem that AI writing seems to be becoming normalized and is having an impact on human authored work.
Up here in Canada, a lawyer who was arguing to get his privileges re-instated after revocation, used AI legal references which turned out to be entirely fictitious. He apologized stating he had no way of noting that AI output was other than reliable.
I have just finished reading Yousuf Karsh: Life & Work and, in my uninformed opinion, it reads like it might have been written to mimic AI assisted work. I am sure that is not true.
It does seem however that opinion is being increasingly presented as fact, in the absence of proof.