Probably not many here will agree with this, but I’ll post it anyway. Just something I wrote over the week that condenses some of my recent thoughts about AI:
20 Likes
Thanks @JasonTheBirder for a thoughtful article. For me, a photograph has (or should have) a close connection with ‘authorship’; I think of photography as being essentially ‘a person making a picture with a camera’. (Yes, I know that even in analogue photography that was already a complicated matter: the film came with characteristics built in, as did the chemistry for development. But in a sense, those things were integral to ‘making a picture with a camera’.)
In that earlier time, the great Italian photographer, Gianni Berengo Gardin, wrote in the introduction to one of his books, ‘No Photoshop was used in making any of the photographs in this book’, or words to that effect. It was important to him to affirm his authorship of his pictures.
In our more complex digital world, I try to keep ‘processing’ to a minimum, and close to the simple, basic things I would have done in a dark room to make a clear, open print. Please appreciate that I’m not saying everyone should work like this, or that people shouldn’t make pictures with whatever tools they wish. I’m only saying what I myself do to maintain my sense of authorship, and hope that this ‘authorship’ is present in my pictures.
So I’m extremely grateful to the darktable developers for making the AI features in dt ‘off by default’, to be turned on only if people wish to use them.
2 Likes
That’s true, and I also agree with that positon as well. However, the main point of my article was not about personal artistic integrity, even though I think that’s an important topic in itself. My point was more that standing against AI could be a useful approach to take even if it does not compromise personal artistic integrity, simply because it sends a strong signal that AI as a whole is not a good thing. In other words, I think human beings should take a stand against AI even if they theoretically could use it responsibly on an individual level.
3 Likes
I welcome and use dt AI object masking. I welcome, but don’t have current use, for the AI de-noising tools. I do not object to AI agents having being used in the programming of these.
I don’t use, and possibly never will, any of the chat tools.
The general-public tools are poor, immature, and generally a very bad, but over-hyped product. Even an enshitified search engine is a better information access tool. Vomiting internet is not my idea of a tool that I want.
My basic objection: it’s a lousy product. Excel would not be where it is today if a double-digit proportion of its results were simply wrong.
Where would I draw the line? Hmmm… For now, I am ok with knowing that I don’t have to. I can take each tool and say yes or no to it.
The term is a marketing tool anyway. Machine learning, neural networks, etc, have been around for a very long time. Doing some very clever stuff. It really is impossible to write off everything that current salespeople want to call “AI.”
1 Like
Thanks Jason for sharing your thoughts on this matter!
I understand (I think) your position and the analogy with DDT.
I’m not fond of what smartphones brought to us.
I’m generally unhappy about big-tech’s role in life as I’m with the direction politics presently take.
But I’m not sure at all that AI is a bad thing generally. There are lot’s of good things to be accomplished with it.
But I do see a need for regulation and verification.
And I’m convinced politicians will be far to late doing so, even if - which presently is not the case - we the people demand them to do so.
Quite possible we are all just frogs in a pan with slowly warming water, not jumping out in time. Which quite probable is NOW.
2 Likes
I’m one of the people that does agree with it. The current crop of LLMs are offered by highly immoral parties, with no regard whatsoever for the environment and the communities that happen to find themselves close to the datacenters, not giving the slightest suggestion of a damn about copyright and attribution. That’s before we get to the whole cognitive decline that is, by now, a well-documented side effect of using them.
A lot of people will say “well yes, but if you set aside the ethical considerations for a second…”, and I do not want to. So no, I will not use these things willingly if at all possible.
This doesn’t include the models that are trained in an ethical and responsible manner, for a specific task. Those are OK with me; I work at a university, I see plenty of use that I can support — the tech itself is neat, and so on.
But apart from that, yes, I’m a staunch opponent.
3 Likes
I think that the companies, and their megalomaniac billionaire bosses, for whom it is a tool for world domination, might be!
I totally agree with @dnzm, in his post prior to this one.
1 Like
Please give some examples.
I think when we discuss AI here, most people think of LLM’s, and image generation.
The AI - or a much better terms machine learning is much more used than that.
https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(23)00298-X/abstract
Here is a study from the Lancet, showing that these algorithms did not let to less detection of breast cancer while improving ergonomics for the radiologist. I call that a good thing.
Here another article: https://pubs.rsna.org/doi/full/10.1148/ryai.2020190208 Showing machine learning is leading to better diagnostics. I call that a good thing.
Just two examples in medicine where machine learning is helpful.
Here is an article about how machine learning is used to ‘weed crops’ AI-Powered Weed Detection and Removal in Agriculture and is reducing chemical use. (Which is really an issue). I think that this is a nice development, reducing pesticides…
If we only see the development of machine learning in connection with big-tech, we are missing the bigger picture.
—-
Just for the record, I am not defending LLMs etc here, pure making a comment that machine learning can do good in my opinion
2 Likes
I second the critical voices.
AI / machine learning is a tool. A tool is just that.
The problem is, people mistake these LLMs as the Oracle of Delphi.
Outsourcing thinking is dangerous.
I remember, one day our daughter wanted to take us on a trip to some lake. It was an estimated 3h ride. So: ‘Google Maps’ → location → destination ->go!
After 1,5h the battery of the mobile drained, charging was impossible due a defective cable- she wanted to go all the way home, in order to get the right cable. She was completely lost. I asked her, if she had a map (that old school thing printed on paper), she had. I said, well then just go ahead.
She was more than skeptical, even though I explained her that there’s an index, we just need to know the name of the next village, look it up and are ready to go. It was like magic to her.
That made me so sad and angry at the same time.
What are they teaching our kids in school- she was in her mid 30s back then, mother of two, master degree in psychology…
Our son is just the same- no navigation, no drive.
Back in the days I knew all telephone numbers I would more or less regularly call, now I’m proud if I can recall my own number…
Laziness, disorientation and giving up thinking,
And we have grandchildren too, thinking about their future, makes me feel grateful that I’m old…
2 Likes
I do partly agree with @JasonTheBirder. But I see i a bit more sophisticated.
I don’t think that AI itself is the problem, but the big tech companies who are offering it and the users who are using it in a way they shouldn’t.
If I see some advertisement where there is propagated, that removing things or people on a photo is the most common thing in the world, I get really upset. Faking has become that normal that people stopped thinking about it.
But again, not the removing of the unwanted parts is the issue, but the intent behind. If you offer that as an arty picture where it is stated or obvious, that it is altered, I wouldn’t have a problem. But most people don’t do that, they cheat others and on the long run themselves.
AI itself is nothing I find ethical critical. The question is, how and with what data were they trained. And here I see a real concern, because NONE of the LLMs of the big tech companies are trained with ethical aspects in mind (it’s more the other way round). On top many or the users don’t use them in a responsible way. On top I as well have huge concerns regarding our environment. I think the LLMs are unbelievable resource hogs which will speed up climate change and water shortage enormously.
This said, I think specialized AI models used responsibly can be helpful. But without any regulation this will and already does lead to more problems than it is helpful.
I have nothing against some helpful AI powered local running tools like denoising or masking. But they should be used carefully and responsibly.
But there is one thing you should be always aware of:
Where is the fun. Where is the proud after you made a great edit, if you just tell the AI I want a great looking photo, and then you get it. Without fiddling around, without thinking how you can achieve a special look, without tackling some technical difficulties, editing pictures is getting boring. And at some point it starts to lose any meaning. Just like the quadrillion meaningless photos on all the smartphones on the world.
4 Likes
I am very skeptical of any argument that equates the lack of the ability to navigate using a paper map with “outsourcing thinking”.
I never learned to use a rotary phone. My parents never learned telegraphy. When I am in a new city that I’ve never been to before, I often struggle with the doors of the metro exactly once until figuring them out. Learning these things from scratch on the spot when you need them for the first time can be tricky. That does not make any of us less able to think, more disoriented or lazier than our ancestors.
1 Like
Why didn’t you teach your children to read a paper map?
It is def a meme that the older generation points in horror at the skills the younger generation, but do so without the knowledge that its their own failure 
1 Like
There’s medical applications, mostly around supporting diagnostics and research, things like pattern recognition. Pattern recognition also helps in “cyber security” (don’t get me started on my opinions regarding terms starting with “cyber”), where they’re pretty good at the random fuzzing and finding weird edge cases that humans are plain bad at spotting, or which are so dependent on several things falling exactly in place that testing them all would take too much time. I know of a project where someone fed the dead sea scrolls through a network and have it determine how many monks worked on it, and who did which part.
There’s plenty of useful (or at least educational) use cases and uses. It’s just that the industry at large is such a disgraceful money grab.
2 Likes
They are sold as such. But yes, people seem to be far too short of critical thinking and all too willing to lap it up.
I unashamedly love navigation software. I love arguing it during the first part of the journey, which I usually know better, or simply have preferred routes, and I love the confidence it gives me when I approach the unknown areas. I often use it for over-view decision making. Then it is just a different medium serving the same purpose.
Mostly technology has compensated for areas in which I was weak, like numbers and memory, or growing weaker, like keeping a mental image of road directions.
True, my handwriting has declined. But this dates way back to the days when I preferred to use a manual typewriter to pen and paper! 
But this would worry me deeply: if I found myself “asking ai” to write this post. That “need,” should be a recognised medical condition!
1 Like
Because we learned it in school, and I never made a test to find out if they can- I didn’t know they can’t.
Also I never tested if they can solve 1+1…
Somethings should be common sense, I think.
2 Likes
This, exactly this, this is the lazynes of the older generations. ‘It is just common sense’. Yes in YOUR world.
But in a world where google maps is (almost) always available, then that is common sense. Actually, in that world it is just plainly stupid to fiddle with maps, spend lots of money on paper maps. Because there is a more efficient way to do it. And you know what is the biggest flaw with maps? They get outdated fairly quickly…
I can read a map and a compass. Not because the generation above me thought me that. Because I started to travel to places where there is no cell phone reception. I thought myself.
And you know what… that map is also ‘outsourcing thinking’…
I you find a flaw in a generation… look at the parents who raised them…
1 Like
I find it harder and harder to separate the two. At least in the big agentic LLM world, the product can’t exist without the companies that already have the money and power to make them work at the scale they are at, and even then the cracks are showing that these companies maybe can’t either. I certainly can’t create a model the likes of OpenAI or Anthropic. I certainly can’t get away with stealing the sum total of human output. I certainly don’t have the political influence to drive policy. I certainly didn’t get invited to stand side by side with the president and other world tech leaders (unless my invite got lost in the mail, then I apologize).
“It’s just a tool” but it’s a tool that none of us can recreate for ourselves. It’s a tool that requires being yoked to big tech companies and/or governments that can and do change things on a whim. So for me, even if the tools were perfect and had no negative issues, just the fact that these tools are tied to the existence of large companies and governments who control access to those services makes me wary of embracing them at all. It makes me scratch my head how much the FOSS world has embraced these tools as well.
That smaller, open, more targeted tools like those that are in darktable now don’t particularly bother me in the same way, I just have very little interest in them.
5 Likes
- drug design
- study of protein folding
These are machine learning applications, little to do with chatbots.
1 Like
Normally i wouldn’t comment about any AI thing any more. But this is about about medicine. There are “thousands” of medical papers around about stuff like this and in many many cases the conclusion are a) it’s helping b) as usual more studies are needed and c) current status of AI assistance is not safe. Not mentioning young docs using AI tools with absolutely frightening suggestions. In last 1/2 year their suggestions would have killed at least 2 patients on the ITS under my supervision.
Or - in current german “Linux Magazin” - just read it. Or decisions in many open source projects …
9 Likes