I agree. Freedom and information go hand in hand, the reasoning behind the inclusion and the available information about each model should be made more prominent.
No worries, no offense taken. On the contrary, I am grateful for the time that you are spending articulating your thoughts and having a polite conversation
I agree. Even though I do not share (most of) the ethical concerns about the SAM model, the fact that different models sit at different points of the ethical spectrum should be emphasized. Probably I did not feel the need for this because I have been following the discussion about their inclusion in the PR, so I was under the impression that there was already a lot of transparency on this point. But it’s true that normal users don’t have this information, whereas it would be better if they did.
I disagree on the wording of the closing statement, not with the sentiment. We should not tell them that it is wrong. We should tell them the facts, highlight that some of the facts have ethical implications and let the users decide whether using them is right or wrong.
What is unethical about ML? The discussion is focusing on the models because it’s pretty much agreed upon that there is nothing unethical with ML, it’s just algorithms.
The first issue that was raised by Mica and others on this forum is that Meta is an “evil” organization whose products harm people.
The second issue is that SAM is a research project from META, whose researchers have endeavored to be transparent and ethical.
The third issue pertains to the sourcing of training data, which is today a universal problem in AI, and its ethical (and sometimes legal) collection and classification process.
Quickly, it seems to me that these three issues define the ethical categories we might include in an ethical use policy for darktable code.
Where does the code come from?
Has it been ethically developed?
Has the training data been ethically sourced?
A community can decide if an ethical use is 2 out of 3 of the above or all or nothing.
Actually I would want a definitive answer on “who the core devs are”. Without knowing that we cannot even define the structure of this “democracy”.
I don’t want to dig into it but mica’s post had this one glaring issue amongst many other concerns. This is more about “who evaluates these categories/models/tools and solved disputes” so that we don’t see another falling out because of this. Sincerely.
@Steven_Adler1
An additional point could be - if an existing tool already serves the same need then how do we judge the inclusion or exclusion of the tool? - this might be the core of this dispute as well and answering this will help us resolve the conflicting arguments and come to a common conclusion.
Preference of more ethically documented models over the model that already serves the same purpose should be noted.
Ofcourse the team already chose how they wanna offer the access but transparency side of things need a lot of work.
I am not an expert in the area, but perhaps it depends on the training data.
If text, images, and source code with various licences are processed by algorithms and what comes out is text, images, source code, or models for such things without any licence, I think is reasonable to ask “what happened to the licences?”.
To me it sometimes feels like the software equivalent of money laundering.
I’m speaking generally - I know nothing about the specific code or models being used in darktable.
There is no formal definition of “core devs”, this phrase was citing Pascal when he pinged some devs that constantly work on dt code over some years. (Unfortunately that ping list did not include some devs doing equally important things).
It’s about who is sharing the burden of keeping dt code in shape, doing code reviews and testing, reading issues systematically … We are all in the dt project for fun spending more or less time on it.
So it’s not about “democracy” - it’s about “how can we make dt better for all users” and/or for each individual dev, for what project do we want to work?
Sure, but these are issues with how data is sourced to train models, as we discussed above. Not with the algorithms, i.e., the ML. The infra that serves (or trains) the model is agnostic wrt the data that flows through it. That’s why the discussion has been focusing on specific models. My statement was in reply to @01McAc who mentioned “unethical ML”, which is misleading.
Thank you for giving a broader perspective . I wanted a dev to respond to that and I am satisfied with your answer. I appreciate you guys put the time and effort needed to not only keep this project maintained but also to keep improving it every day.
I only picked the words and phrases I read in this thread so it broadly answers a few concerns.
Again it only helps drive the conversation to a better place.
That was a lot to read. I have been reading a bit of the articles being shared earlier about the if the models are unethical or not. These articles are not necessarily only about image tagging but also about the moderation of post on social media.
But main issues that are raised in those articles:
Low wages: from a western perspective the workers are paid a very low income. From what I have read between $1.50 and $2.50 hour.
Job security the data-workers don’t have much of a job security. They can be fired the moment the work is done, or when contracts are cancelled.
Work pressure the workers report that they are pressured to process an unreasonable amount of images / moderations per hour. This seems to have to be improved over the years but the pressure is always: for you 10 others.
inappropiate content Data-workers especially when they are moderating are often presented with violent, child pornography and other disturbing content without any form of support how to deal with that mentality.
Bad working conditions, basically: long days, few breaks and not much daylight.
————-
This is a summary of what I have found so far.
Here are my thoughts:
The low wages are relative as I earlier mentioned. In those economies $1.50 / hour is not a minimum wage. For context: In some of the places I have lived in, the boys that were collecting the thrash on the street for recycling were making a $1 a day, which buys you 3 breads (I am not saying that is a good thing - this pure for context)
Job security. This is an interesting one. It depends much on your view on companies etc. Being from west Europe, I find this problematic, as we have strong laws for job security. But if you are from a country where your boss can fire you per email (as recently with Oracle) you may have a different view on this.
For me the combination of 3,4,5 is a real issues. Especially when 4 is the mix. But for the SAM models discussed here, 4, is likely less applicable as that is about tagging general images.
Nevertheless, I find this concerning. A lot on analogies have been made. What comes to mind here is the textile industry.
I hope the Meta researcher will respond with answers.
If these jobs are so bad, why do people take them? Are they being forced into labor? Or maybe it’s because having those jobs is better than not having them, or better of other jobs available on the market for the people who accept those jobs? Would they be better off WITHOUT those jobs? Is having an insecure job WORSE than having no job? Is having a low wage (for our standards) worse than having no wage? If you want to feed yourself and your family and the only job that you can get is this one, would you be happy to have this opportunity or would rather starve you and the rest of the family?
I am sorry, I really find this conversation surreal. I mean no disrespect, but come on, seriously.
EDIT: @martinus this is not against you personally, you are not the first one to raise these points, I am anchoring to your post just because you are the last one to list all these points very clearly, which makes the surreality of this argument even more glaring.
EDIT(2): Life is full of surprises. I would have never imagined that one day I would end up defending Meta in a conversation…
Is there necessarily a difference? If you at the outset have no possibility for income, anything is better than zero - but you are in reality forced and “anything” doesn’t always equal decency,
Seriously, if this the moral reasoning your applying? I am baffled.
As I spelled out, the in our eyes the low wage is not an actual low wage.
The problem is this: the companies providing these jobs don’t care about the workers (issue 3,4,5) because they know that without them they starve. That is the underlying corporate greed I see. If those issues were addressed it would be much less of an issue to me.
For me it almost sounds you are saying: they should be happy we bring them those jobs.
I am saying that no one is forced to accept those jobs. Every single person that accepts to do that job prefers that alternative to other alternatives, or lack thereof. So, yes, those jobs are adding value to the lives of the people who accept them, as confirmed by the fact that those jobs are being accepted.
You speak as if they were being forced to take those jobs, which they are not. It’s not slavery that we are talking about, even though you seem to be trying to give this impression.
I would take a job instead of starving to death, yes. But i would also know what that job is - slave labour. People work under all circumstances. Nothing to be accepted and supported.
@hannoschwalm@martinus and others who share this position. Please, answer the following question.
If I am poor and conditions force me to accept job X, how is removing job X from the equation going to make my life better?
If I chose job X over job Y it means that job X was better than job Y in the first place. So, if job X did not exist, would my life be any better?
Your reasoning is based on the assumption that, if job X was not there, I would get a fantastic job Z with much better conditions. But wait a second, if job Z is there waiting for me, why don’t I already pick that instead?
So, to reiterate. How is removing a job opportunity going to improve the living conditions of those for which this job is the best available option (as confirmed by the fact that they chose this job over other jobs)?
You talk about the freedom to choose. But in many of those countries and situations there is not much of a real choice. When struggling to survive you take anything to survive. It is not that most of those people make a very rational decision. It is a matter often of living or dying.
If job A pays or is better then job B, that doesn’t mean that job A is a good job or ethically right. Both can be very bad. I also would pick the lesser evil of the two.
And to answer your question,
—————————-
I have to say, I had the same way of thinking like you. Until I started traveling and saw things in the ‘3rd world’ with my own eyes. Till I saw the poverty and talked with people. Till I heard how people were used to test medicine without knowing what was given to them. Till my wife told me the stories of girls that she met that were sold as sex-slave so parents could buy food for their sons.
Yes those ‘data-worker’ jobs are not worst. And compared with jobs in mining or manufacturing they are not that bad. But that does not make them by default good either.
And to answer your question, no their living conditions will not improve if that job opportunity is removed for the job market.
But are the conditions really improving because of those jobs or are the worker simply trading one form of suffering for another? And who is getting the most benefits out of this? The workers, or in the end the big companies?
Business want to maximise profit by reducing their costs. Many of us would like to put what little pressure we can on businesses to share profits more fairly. If I stop engaging with a company, that is often about as much pressure as I can put on. It is next to nothing in terms of pressure, but what’s the alternative? To continue to reward companies for keeping poor people poor?
So, it’s not about wanting to stop offering people jobs. It is about want to offer jobs which pay a reasonable share of the profit and are ethical. Scam calling is a job, probably pays better than begging in the street. Doesn’t mean I have to like it.
For me, this conversation has run its course, and I doubt anyone is about to change their mind (or that would have happened before post 657). Best of luck to you all.
This is only anecdotal, but the majority of vehicle-pedestrian accidents I’ve seen reported on our local news (Colorado Springs, CO) involve pedestrians who choose to try to cross roads at places other than the marked and signalled crosswalks at intersections. Not that crosswalks are completely safe; I spent the better part of two decades walking to/from work, and the drivers I observed who ignore crosswalk boundaries were numerous.
I’d assert there’s choices to be made all around. And, physics is a harsh mistress…
