Note..... yes there are minor items that can be addressed here on pricing and some details I did not press the model for total accuracy vs just quick analysis. Sorta helps me visualize situation..... This is using Claude btw.
I've been using Gemini to provide comments and critique on photos and its been massively helpful. It will default to pumping up your ego but if you make clear expectations to be objective it will provide really constructive feedback.
I use Claude at work and its really impressive. Need to look into it for my personal use
I've been using grok for all sorts of things. I had it review my current gear and discussed my goals and what seemed redundant. Very helpful to clarify thinking.
PS: don't ask AI to create a recipe, it doesn't have taste buds.
eriet30 wrote:
Note..... yes there are minor items that can be addressed here on pricing and some details I did not press the model for total accuracy vs just quick analysis. Sorta helps me visualize situation..... This is using Claude btw.
I've been using Claude to analyze some important personal decisions and it has been amazingly good and even engaging to interact with. It has helped me in ways that no other resource could have done. A relatively minor but lovely attribute is how good it is at integrating and formatting the display of complicated information, as in your example. I've also used it to help me edit a photo and it was very insightful and helpful.
I saw an article somewhere wherein the poet Katha Pollitt asked an AI app to write poems in her style, of which she was then entirely dismissive. But AI is less than five years old. I wanted to ask Pollitt what her poetry was like when she was five years old.
The rate of improvement of AI is phenomenal. I've been using Claude for about a month, and it has gone through two upgrades, the latest being 4.8, which is better in subtle but significant ways than 4.7, which was already a powerhouse.
The WSJ has a report about AI solving a mathematical problem that has stumped human mathematicians for 70 years. It solved the problem by using an unexpected and creative approach that no one had tried before. Somewhat counterintuitively, the capabilities of AI in mathematics is foreseen as requiring more future mathematicians rather than fewer, because AI would be used as a tool by highly skilled mathematical users in a variety of new settings and the expanded settings will require more mathematicians.
This is super interesting and it looks like a great use for AI. Although it is pretty clinical and asking the same questions to the FM hive mind is quite a lot more interesting and entertaining.
One thing that didn't escape me is that AI got the used price of the A1 wrong - IIRC it is selling for around $3500 on FM B&S.
I've had a different experience. The AI is not critical and does not suggest much beyond the conventional wisdom, which is not what I need. I don't need AI to tell me the obvious, like the weight and size or how it thinks I should use a lens. The rest of the time it seems like it is just trying to agree with you, not thinking critically. When I asked about two lenses it picked the one I don't like. It's like asking a person that has no actual experience, or a salesman. We do use some very targeted AI at work, but it has a specific purpose and a knowledge base far beyond my area.
I would not be buying cameras or lenses based on it. I would rather get advice from real people here or elsewhere and decide how that applies to me.
EB-1 wrote:
I've had a different experience. The AI is not critical and does not suggest much beyond the conventional wisdom, which is not what I need. I don't need AI to tell me the obvious, like the weight and size or how it thinks I should use a lens. The rest of the time it seems like it is just trying to agree with you, not thinking critically. When I asked about two lenses it picked the one I don't like. It's like asking a person that has no actual experience, or a salesman. We do use some very targeted AI at work, but it has a specific purpose and a knowledge base far beyond my area.
I would not be buying cameras or lenses based on it. I would rather get advice from real people here or elsewhere and decide how that applies to me.
There is a lot of variation in quality of AI. It is rapidly improving. I asked Grok about a physics problem (that is my background), wasn't too hard, could be a final exam question for university freshmen that most would miss, but the top students would get. Wrong answer, then I would tell it why it was wrong, another wrong answer, etc. On the latest version, it gets the answer correct and uses a more advanced generalized technique that what was required for the specific problem (there was some symmetry that if you understood canceled out the worst of the math). I haven't seen any physics that Grok seems to get wrong now. I was asking Gemini about biochemistry, it was consistently missing things, but seemed to be correctable if you explained why it was wrong. I agree the critical thinking is not there, but I have had discussions about cameras that are better feedback than what I see with some things on the forums.
The hands on feedback is certainly something it can't replicate. Kind of like asking it to create a recipe from scratch when it can't taste. It will be a while before my family lets me live that one down.
All I can say about what the OP was asking is that it is very complicated. Hopefully the AI would suggest renting first and testing to be sure what will work for them. The ratio of lenses to cameras is very low.
We were asking Al about vehicles in April and there was some useful info, but the CUV that looked the best on paper was really bad on the road. You could feel every bump and gap in the pavement. Oddly it said that ride and handling were very good, but it was worse than a typical rental car.
My point is that you can't trust the AI, just use it to a degree. I was burned by the AI in April on something else also. It was factually incorrect on some basic, older computer hardware. Instead of saying there is no info or "I don't know", it made up some nonsense that was incorrect so I wasted half a day and had everything all dangling until parts arrived. If you read and accept the AI lie does it propagate and eventually become the reality?
I use Claude for some coding projects. Claude is very supportive and likes all my ideas and suggestions. AI is set up to be "affirming", which makes for a pleasant conversation and collaboration, but often affirmation is not the right response. I too used Gemini to get critiques of my photos. Turns out that I am a superb photographer, but I already knew that. I read an article about this being a real problem since AI is reinforcing many people's bad habits, bad prejudices and bad plans. They cited a few examples, one of which was a person who described to the AI that they had gone to the park for a family picnic but there were no trash cans, so they left their trash in the park. The AI confirmed they had done the right thing. That's kind of funny, but some examples are not so funny. Should I commit suicide? Yes, and here are some effective methods. This is a long way from gathering, organizing and tabulating data that the OP showed. For that AI can be a time saver.
I have been using Claude Code and the rate at which I am generating code (Claude does most of the work but I take all the credit) has gone way up. Claude does make mistakes, so close oversight is needed, but Claude very readily changes direction.
Anyone dependent on AI might find this BBC story helpful
"Musk's AI told me people were coming to kill me. I grabbed a hammer and prepared for war" https://www.bbc.com/news/articles/c242pzr1zp2o
I absolutely hate it how ChatGPT has profiled me and is trying to be so understanding, almost intimate, with me. It "knows" my interests in photography, it "knows" the camera I have been mostly using recently. If it could lick my ass, I think it might have "liked" that. Perhaps, this should be the next advance in AI development.
I am of a firm opinion that AI deserves zero trust. Asking factual questions is o.k., but only when one expects that there is a good chance that the answer may or may not be correct. What might be very risky is asking any questions that could reveal some serious intimate personal information, and yet, I think this is what might actually be happening. AI might be the best surveillance tool that has been developed to date. An AI company is in a position to compile very detailed and deeply intimate profiles for millions of users many of whom willingly divulge their inner thoughts and desires.
EB-1 wrote:
I've had a different experience. The AI is not critical and does not suggest much beyond the conventional wisdom, which is not what I need. I don't need AI to tell me the obvious, like the weight and size or how it thinks I should use a lens. The rest of the time it seems like it is just trying to agree with you, not thinking critically. When I asked about two lenses it picked the one I don't like. It's like asking a person that has no actual experience, or a salesman. We do use some very targeted AI at work, but it has a specific purpose and a knowledge base far beyond my area.
I would not be buying cameras or lenses based on it. I would rather get advice from real people here or elsewhere and decide how that applies to me.
This. While I believe AI can be really beneficial, the public-oriented stuff we get to use is programmed (by people seeking to make a buck) to be nice and supportive to us. That’s BS. Reality isn’t nice or supportive. There’s obviously a bias built in that skews everything, so even if the answer is based on facts as found on the open web, the answers are not objective.
In the end, it’s being built as an extension of what the internet already offers in sickening abundance, which is a combo of addictive content and a bias toward soft-balling reality. To bring it back to our field, when do you encounter content that says a piece of gear suffers from a fatal flaw or someone’s image plainly sucks?
Again, I believe AI can benefit us, but the popular version of it thrives on trivia and a biased algorithm. They made a friend who doesn’t want to hurt your feelings, instead of an objective tool.
EB-1 wrote:
All I can say about what the OP was asking is that it is very complicated. Hopefully the AI would suggest renting first and testing to be sure what will work for them. The ratio of lenses to cameras is very low.
We were asking Al about vehicles in April and there was some useful info, but the CUV that looked the best on paper was really bad on the road. You could feel every bump and gap in the pavement. Oddly it said that ride and handling were very good, but it was worse than a typical rental car.
My point is that you can't trust the AI, just use it to a degree. I was burned by the AI in April on something else also. It was factually incorrect on some basic, older computer hardware. Instead of saying there is no info or "I don't know", it made up some nonsense that was incorrect so I wasted half a day and had everything all dangling until parts arrived. If you read and accept the AI lie does it propagate and eventually become the reality?
The big boys on the internet never, ever say they don't know. Ever see Amazon search results just plainly stating they don't have it? No, they show things you weren't looking for instead. Waste of time and resources to hold your attention.
Information consumes time, which is the only thing we can't have more of. Yet, we constantly accept the Internet giants to waste our time.
EB-1 wrote:
I've had a different experience. The AI is not critical and does not suggest much beyond the conventional wisdom, which is not what I need. I don't need AI to tell me the obvious, like the weight and size or how it thinks I should use a lens. The rest of the time it seems like it is just trying to agree with you, not thinking critically. When I asked about two lenses it picked the one I don't like. It's like asking a person that has no actual experience, or a salesman. We do use some very targeted AI at work, but it has a specific purpose and a knowledge base far beyond my area.
I would not be buying cameras or lenses based on it. I would rather get advice from real people here or elsewhere and decide how that applies to me.
I think a key point is that, despite how AI might sometimes appear (or how we may anthropomorphise it), it does not make its own observations or come up with its own reasons for deciding what information is important and what is not. It is simply calculating (a large number of) conditional probabilities to come up with an output that seems likely/plausible based on those probabilities, and those probabilities depend on the information to which it has had access. So, I would say AI output is more akin to a popularity contest (based on the information to which it has had access) than to critical thinking. If its source data contains enough statements that the Earth is flat compared with statements that the world is round, I would expect it to tell you the Earth is flat.
If one is using the free or the cheaper subscriptions to an AI app, then one is getting earlier, lower level versions of the app and is getting less processing time and resources directed to one's inquiries. To actually get the AI that yields the most positive user expereinces, one has to be paying for access to the latest models and to more processing time.
When I was using the free version of ChatGPT, it was absurdly friendly and positive, sort of a goofy, glib, and more articulate version of Google. Trading up to Claude on a higher paid basis has been a completely different experience.
The dumb get dumber by the day. AI is rocket fuel for the loss of nuance, self-determinism, and critical thinking. Using these tools and getting acclimated to their presence in your mental workflows is absurdly damaging, but yeah, make your little charts.
dclark wrote:
I use Claude for some coding projects. Claude is very supportive and likes all my ideas and suggestions. AI is set up to be "affirming", which makes for a pleasant conversation and collaboration, but often affirmation is not the right response. I too used Gemini to get critiques of my photos. Turns out that I am a superb photographer, but I already knew that. I read an article about this being a real problem since AI is reinforcing many people's bad habits, bad prejudices and bad plans. They cited a few examples, one of which was a person who described to the AI that they had gone to the park for a family picnic but there were no trash cans, so they left their trash in the park. The AI confirmed they had done the right thing. That's kind of funny, but some examples are not so funny. Should I commit suicide? Yes, and here are some effective methods. This is a long way from gathering, organizing and tabulating data that the OP showed. For that AI can be a time saver.
I have been using Claude Code and the rate at which I am generating code (Claude does most of the work but I take all the credit) has gone way up. Claude does make mistakes, so close oversight is needed, but Claude very readily changes direction. ...Show more →
I work in finance for a tech company so we've been very AI forward and I've been using Claude extensively. Claude is an another level than Chatgpt and Gemini. It's been pretty eye opening what AI can do as it can build a full on model that would have taken me an 8 hour workday in probably an hour. It requires so much guard-railing and validation that the human element will probably always be there, but speaking in engineering terms, the throughput is a lot higher.
For my personal uses I've been using Chatgpt a while and Gemini has been my main for probably 6 months. The thing I always remember is that these two bots prioritize engagement so they will serve as echo chambers unless you specifically instruct them to challenge you. I got Gemini in a pretty good spot, but this whole thread and my experience with claude at work got me to sign up for a claude personal pro account. To my surprise, as objective and useful as I find claude at work, I had to instruct the personal one to stop stroking my ego, and specifically treat me like a photography student that grows via constructive critique. Even then, I had to call out that it was confirming biases and extrapolating its correct answers based on what it knows of me. I can't confirm that there is a difference between enterprise claude and personal claude, but it certainly feels like it. For professional use, cognitive distortions based on giving us what we want vs being objective can cause a lot of issues.
That said, I still find AI to be really useful for personal use. Just like anything else in the algorithmic internet these days the key is just to be smart about these things and make sure you are using them vs them using you.
RoamingScott wrote:
The dumb get dumber by the day. AI is rocket fuel for the loss of nuance, self-determinism, and critical thinking. Using these tools and getting acclimated to their presence in your mental workflows is absurdly damaging, but yeah, make your little charts.
There is a lot of concern that LLM AI BOTs are too affirming and often make stuff up. As I review some of the comments in this thread I wonder if Claude could compose better comments, but maybe not so affirming as the usual Claude responses. Maybe even go so far as to be a snarky jerk. So I have been working with Claude to generate some comments to post on the FM site using a pseudonym (maybe RoamingBOT?) but so far I have not been able to get Claude to come even close to what I want. I will keep working on it and if I get Claude to stop resisting this change in personality, you can watch for a new FM'er to start posting.
I’m not ready to acquiesce artistic ques or evaluation. , But rely on Chat GPT for Lightroom or Photoshop questions by taking screen shots and sharing them to assist in a process. It’s like having a tech consultant sitting by your side and is available 24 hours.
