To get those labels, OpenAI sent tens of thousands of snippets of text to an outsourcing firm in Kenya, beginning in November 2021. Much of that text appeared to have been pulled from the darkest recesses of the internet. Some of it described situations in graphic detail like child sexual abuse, bestiality, murder, suicide, torture, self harm, and incest.
OpenAI’s outsourcing partner in Kenya was Sama, a San Francisco-based firm that employs workers in Kenya, Uganda and India to label data for Silicon Valley clients like Google, Meta and Microsoft. Sama markets itself as an “ethical AI” company and claims to have helped lift more than 50,000 people out of poverty.
The data labelers employed by Sama on behalf of OpenAI were paid a take-home wage of between around $1.32 and $2 per hour depending on seniority and performance. For this story, TIME reviewed hundreds of pages of internal Sama and OpenAI documents, including workers’ payslips, and interviewed four Sama employees who worked on the project. All the employees spoke on condition of anonymity out of concern for their livelihoods.
[…]
Documents reviewed by TIME show that OpenAI signed three contracts worth about $200,000 in total with Sama in late 2021 to label textual descriptions of sexual abuse, hate speech, and violence. Around three dozen workers were split into three teams, one focusing on each subject. Three employees told TIME they were expected to read and label between 150 and 250 passages of text per nine-hour shift. Those snippets could range from around 100 words to well over 1,000. All of the four employees interviewed by TIME described being mentally scarred by the work. Although they were entitled to attend sessions with “wellness” counselors, all four said these sessions were unhelpful and rare due to high demands to be more productive at work. Two said they were only given the option to attend group sessions, and one said their requests to see counselors on a one-to-one basis instead were repeatedly denied by Sama management.
[…]
One Sama worker tasked with reading and labeling text for OpenAI told TIME he suffered from recurring visions after reading a graphic description of a man having sex with a dog in the presence of a young child. “That was torture,” he said. “You will read a number of statements like that all through the week. By the time it gets to Friday, you are disturbed from thinking through that picture.” The work’s traumatic nature eventually led Sama to cancel all its work for OpenAI in February 2022, eight months earlier than planned.
[…]
That month, Sama began pilot work for a separate project for OpenAI: collecting sexual and violent images—some of them illegal under U.S. law—to deliver to OpenAI. The work of labeling images appears to be unrelated to ChatGPT.
What about spending all day being abused by people in a call center?
I mean sure we’d all like to make enough money to live a full life with any job but that’s sadly not a reality and the point you’re missing is that economies don’t work the same as the US in every country.
I live in Argentina, I make 25k a year as a software developer and I’m on the top 1% of highest earners on the country
I strongly disagree. I have read and seen a lot of messed up things on the internet, I much, much, prefer it to the couple weeks I spent helping out a friend at a part-time service job. (And I was doing it with good friends in a casual environment.)
One Sama worker tasked with reading and labeling text for OpenAI told TIME he suffered from recurring visions after reading a graphic description of a man having sex with a dog in the presence of a young child. “That was torture,” he said. “You will read a number of statements like that all through the week. By the time it gets to Friday, you are disturbed from thinking through that picture.” The work’s traumatic nature eventually led Sama to cancel all its work for OpenAI in February 2022, eight months earlier than planned.
Is not worth high pay, but I would say psychologically damaging your employees and then not even giving them the counseling tools to help them is absolutely worth high pay. You should not have to endure things like that for an ‘above the median’ wage in a country where ‘the median’ is still being very poor. I see this as not much better than defending other corporations making poor people in Africa work in mines for a decent wage relative to others in their country but not giving them safety equipment. And they still die poor.
I obviously prefer people aren’t in poverty at all. But I have far more sympathy for the miner risking their lives than someone reading something disgusting/disturbing on the internet, it is not anywhere near close.
You don’t understand how massive psychological damage can be as bad as seriously endangering someone’s physical health?
Just because a graphic description of a dog being raped while a child watches doesn’t bother you doesn’t mean it won’t bother anyone else. In fact, I would wager that it would be pretty disturbing for most people to read that, let alone read that sort of thing for hours every day.
And then there are the ones who are just as low-paid but have to look at images instead. Again, you may not be bothered by CSAM, but I would wager that most people would find looking at that all the time very hard to deal with and it could easily result in PTSD.
Getting crushed in a mine collapse harms everyone. As unfashionable as it is, the vast majority of people, that I know at least, have experienced far more traumatic things than you could ever get from third person observation.
I hate gore, I hate seeing people dying, I hate hearing about those sorts of things. They seriously upset me, but to compare that discomfort to anything like someone working (maybe enslaved) in a mine in essentially anywhere in Africa is ridiculous. Risking on a daily basis, painful death, painful suffering than death, likely slow death from dust inhalation, severe maming, etc.
If you really believed reading it were that dangerous, it is evil of you to even summarize it as you did and risk serious harm to others.
No, you’re right, you should be. We don’t want to normalize this shit, it should continue to shock and offend.
These are the dark sides of modern technology. The kids working cobalt mines. The workers being paid pennies to categorize data so bad that it is traumatic to even read it. I can’t imagine how the people who have to look at pictures can do it.
I feel like I could handle some dark text here or there, but if I had to do it for 40-50 hours a week? Hundreds of passages every day. That would warp me pretty quickly.
The last quote danced around it but if the implication is that they were seeking out and collecting CSAM which is a sex crime to access, possess and distribute, why the fuck are the boards of both companies not in prison and on the sex offender list?!
Isn’t CSAM classed as images and videos which depict child sexual abuse? Last time I checked written descriptions alone did not count, unless they were being forced to look at AI generated image prompts of such acts?
That month, Sama began pilot work for a separate project for OpenAI: collecting sexual and violent images—some of them illegal under U.S. law—to deliver to OpenAI. The work of labeling images appears to be unrelated to ChatGPT.
This is the quote in question. They’re talking about images
I really find this a bit alarmist and exaggerated. Consider the motive and the alternative. You really think companies like that have any other options than to deal with those things?
If absolutely nothing else and even assuming for the sake of the argument that work of this nature is completely justified, they still have to answer for the fact that they severely underpaid foreign workers in clickfarms to do this and traumatize themselves on their behalf presumably so no one in the West had to.
Personally, my opinion is very strongly that if you can’t develop a technology without committing such serious ethical breaches, for example seeking out and accumulating CSAM, then it’s either too early to develop that technology or it’s not worth developing at all. One may counter this with something like “well it’s basically inevitable that unscrupulous people will harm others to develop technology” but I would also argue that while that is true, the inevitability of something doesn’t make the act itself any less unethical.
As a bit of context: The reason why even accessing and possessing CSAM is illegal almost everywhere in the world is because the generally accepted philosophy around this kind of material is that every time someone views it for any reason, it victimizes that child all over again, which is also very consistent with the opinions of actual CSAM survivors so I don’t feel it’s something that the rest of us can really question. I obviously cannot speak on their behalf in any way, but my guess would be the vast majority of CSAM victims do not want photos and videos the most terrifying and traumatic moments of their lives being used in this way, especially not by a for-profit company so they can develop a product with the goal of making themselves richer.
Consider the impact on human psychology. Not everyone has the guts to read and even look through these. And even though they appear to have, it still scars them inside.
Maybe There is no alternative for now, but don’t do that to people with such low paycheck. Consider even the background of these people who may work on these tasks to not even live, but to survive. I would have preffered to wait 10 years than to indulge these horrifying tasks to those persons.
I’m sure there are lots of people who are in jail for creating/sharing or even making a profit off of these content. They could do that work ? But then again, even though it bothers me less than people who has no choice to live their lives, that is still an Idea I find ethically very questionable.
Very much yes police authorities have CSAM databases. If what you want to do with it really is above board and sensible they’ll let you access that stuff.
I don’t doubt anything that OpenAI could do with that stuff can be above board, but sensible is another question: Any model that can detect something can be used to train a model which can generate it. As such those models are under lock and key just like their training sets, (social) media platforms which have a use for these things and the resources run them, under the watchful eye of the authorities. Think faceboogle. OpenAI could, in principle, try to get into the business of selling companies at that scale models they can, and have, trained themselves, I don’t really see that making sense from the business POV, either.
This reminds me of an NPR podcast from 5 or 6 years ago about the people who get paid by Facebook to moderate the worst of the worst. They had a former employee giving an interview about the manual review of images that were CP andrape related shit iirc. Terrible stuff
This is actually extremely critical work, if results are going to be used by ai’s that are going to be used widely. This essentially determines the “moral compass” of the ai.
Imagine if some big corporation did the labeling and such, trained some huge ai with that data and it became widely used. Then years pass and eventually ai develops to such extent it can be reliably be used to replace entire upper management. Suddenly becoming slave for “evil” ai overlord is starting to move from being beyond crazy idea to plausible(years and years in future, not now obviously).
Extremely critical but mostly done by underpaid workers in poor countries who have to look at the most horrific stuff imaginable and develop lifelong trauma because it’s the only job available and otherwise they and their family might starve.
Source
This is one of the main reasons I have little hope that if OpenAI actually manages to create an AGI that it will operate in an ethical way. How could it if the people trying to instill morality into it are so lacking in it themselves.
True. Though while its horrible for those people, they might be doing more important work than they or us even realize. I also kind of trust moral judgement of oppressed more than oppressor(since they are the ones who do the work). Though i’m definitely not condoning the exploitation of those people.
Its quite awful that this seems to be the best we can hope for regarding this. I doubt google or microsoft are going to give very positive guidance whether its ok for people to suffer if it leads to more money for investors when they do their own labeling.
https://time.com/6247678/openai-chatgpt-kenya-workers/
[…]
[…]
[…]
Gonna leave this here.
So they paid Kenyan workers $2 an hour to sift through some of the darkest shit on the internet.
Ugh.
What? And here I am doing it for free…
They could have just given 4chan a $1 bounty per piece and they would have gleefully delivered until Lambo.
They are problaby the ones writing those pieces literature
All while getting all high and mighty about how AI is poised to rid humanity of the need to make humans do degrading jobs, mind you.
In some countries 2 bucks an hour puts you above the median
“Above the median” should not be the standard for having to spend all day reading about racism and rape.
What about spending all day being abused by people in a call center?
I mean sure we’d all like to make enough money to live a full life with any job but that’s sadly not a reality and the point you’re missing is that economies don’t work the same as the US in every country.
I live in Argentina, I make 25k a year as a software developer and I’m on the top 1% of highest earners on the country
I strongly disagree. I have read and seen a lot of messed up things on the internet, I much, much, prefer it to the couple weeks I spent helping out a friend at a part-time service job. (And I was doing it with good friends in a casual environment.)
You’re welcome to strongly disagree that this:
Is not worth high pay, but I would say psychologically damaging your employees and then not even giving them the counseling tools to help them is absolutely worth high pay. You should not have to endure things like that for an ‘above the median’ wage in a country where ‘the median’ is still being very poor. I see this as not much better than defending other corporations making poor people in Africa work in mines for a decent wage relative to others in their country but not giving them safety equipment. And they still die poor.
I obviously prefer people aren’t in poverty at all. But I have far more sympathy for the miner risking their lives than someone reading something disgusting/disturbing on the internet, it is not anywhere near close.
You don’t understand how massive psychological damage can be as bad as seriously endangering someone’s physical health?
Just because a graphic description of a dog being raped while a child watches doesn’t bother you doesn’t mean it won’t bother anyone else. In fact, I would wager that it would be pretty disturbing for most people to read that, let alone read that sort of thing for hours every day.
And then there are the ones who are just as low-paid but have to look at images instead. Again, you may not be bothered by CSAM, but I would wager that most people would find looking at that all the time very hard to deal with and it could easily result in PTSD.
Getting crushed in a mine collapse harms everyone. As unfashionable as it is, the vast majority of people, that I know at least, have experienced far more traumatic things than you could ever get from third person observation.
I hate gore, I hate seeing people dying, I hate hearing about those sorts of things. They seriously upset me, but to compare that discomfort to anything like someone working (maybe enslaved) in a mine in essentially anywhere in Africa is ridiculous. Risking on a daily basis, painful death, painful suffering than death, likely slow death from dust inhalation, severe maming, etc.
If you really believed reading it were that dangerous, it is evil of you to even summarize it as you did and risk serious harm to others.
That’s actually about 3x what the average Kenyan makes, sadly.
I’m shocked and I shouldn’t be… Poor people
No, you’re right, you should be. We don’t want to normalize this shit, it should continue to shock and offend.
These are the dark sides of modern technology. The kids working cobalt mines. The workers being paid pennies to categorize data so bad that it is traumatic to even read it. I can’t imagine how the people who have to look at pictures can do it.
I feel like I could handle some dark text here or there, but if I had to do it for 40-50 hours a week? Hundreds of passages every day. That would warp me pretty quickly.
The last quote danced around it but if the implication is that they were seeking out and collecting CSAM which is a sex crime to access, possess and distribute, why the fuck are the boards of both companies not in prison and on the sex offender list?!
I mean, I know why, but
I’m sure there’s some loophole there, maybe between countries’ laws. And if there isn’t, Hey! We’ll make one!
Isn’t CSAM classed as images and videos which depict child sexual abuse? Last time I checked written descriptions alone did not count, unless they were being forced to look at AI generated image prompts of such acts?
This is the quote in question. They’re talking about images
They could be working with the governments of relevant countries to develop filters and detection systems.
IIRC there are a few legitimate and legal reasons to seek CSAM, such as journalism, and definitely developing methods to prevent it’s spread.
I really find this a bit alarmist and exaggerated. Consider the motive and the alternative. You really think companies like that have any other options than to deal with those things?
If absolutely nothing else and even assuming for the sake of the argument that work of this nature is completely justified, they still have to answer for the fact that they severely underpaid foreign workers in clickfarms to do this and traumatize themselves on their behalf presumably so no one in the West had to.
Personally, my opinion is very strongly that if you can’t develop a technology without committing such serious ethical breaches, for example seeking out and accumulating CSAM, then it’s either too early to develop that technology or it’s not worth developing at all. One may counter this with something like “well it’s basically inevitable that unscrupulous people will harm others to develop technology” but I would also argue that while that is true, the inevitability of something doesn’t make the act itself any less unethical.
As a bit of context: The reason why even accessing and possessing CSAM is illegal almost everywhere in the world is because the generally accepted philosophy around this kind of material is that every time someone views it for any reason, it victimizes that child all over again, which is also very consistent with the opinions of actual CSAM survivors so I don’t feel it’s something that the rest of us can really question. I obviously cannot speak on their behalf in any way, but my guess would be the vast majority of CSAM victims do not want photos and videos the most terrifying and traumatic moments of their lives being used in this way, especially not by a for-profit company so they can develop a product with the goal of making themselves richer.
Consider the impact on human psychology. Not everyone has the guts to read and even look through these. And even though they appear to have, it still scars them inside.
Maybe There is no alternative for now, but don’t do that to people with such low paycheck. Consider even the background of these people who may work on these tasks to not even live, but to survive. I would have preffered to wait 10 years than to indulge these horrifying tasks to those persons.
I’m sure there are lots of people who are in jail for creating/sharing or even making a profit off of these content. They could do that work ? But then again, even though it bothers me less than people who has no choice to live their lives, that is still an Idea I find ethically very questionable.
Very much yes police authorities have CSAM databases. If what you want to do with it really is above board and sensible they’ll let you access that stuff.
I don’t doubt anything that OpenAI could do with that stuff can be above board, but sensible is another question: Any model that can detect something can be used to train a model which can generate it. As such those models are under lock and key just like their training sets, (social) media platforms which have a use for these things and the resources run them, under the watchful eye of the authorities. Think faceboogle. OpenAI could, in principle, try to get into the business of selling companies at that scale models they can, and have, trained themselves, I don’t really see that making sense from the business POV, either.
This reminds me of an NPR podcast from 5 or 6 years ago about the people who get paid by Facebook to moderate the worst of the worst. They had a former employee giving an interview about the manual review of images that were CP andrape related shit iirc. Terrible stuff
Hold on, why exactly do they need people to label this shit?
How else will the AI be able to recognize that such text is “bad”?
This is actually extremely critical work, if results are going to be used by ai’s that are going to be used widely. This essentially determines the “moral compass” of the ai.
Imagine if some big corporation did the labeling and such, trained some huge ai with that data and it became widely used. Then years pass and eventually ai develops to such extent it can be reliably be used to replace entire upper management. Suddenly becoming slave for “evil” ai overlord is starting to move from being beyond crazy idea to plausible(years and years in future, not now obviously).
Extremely critical but mostly done by underpaid workers in poor countries who have to look at the most horrific stuff imaginable and develop lifelong trauma because it’s the only job available and otherwise they and their family might starve. Source This is one of the main reasons I have little hope that if OpenAI actually manages to create an AGI that it will operate in an ethical way. How could it if the people trying to instill morality into it are so lacking in it themselves.
True. Though while its horrible for those people, they might be doing more important work than they or us even realize. I also kind of trust moral judgement of oppressed more than oppressor(since they are the ones who do the work). Though i’m definitely not condoning the exploitation of those people.
Its quite awful that this seems to be the best we can hope for regarding this. I doubt google or microsoft are going to give very positive guidance whether its ok for people to suffer if it leads to more money for investors when they do their own labeling.