Live data from Hacker News

Kenyan moderators decry toll of training of AI

theguardian.com

61–70 of 91 posts

Re: Kenyan moderators decry toll of training of AI

#61
post #50

Earlier quoted context omitted.

If a worker effectively performs the required tasks for $1.46/hr, this demonstrates the work's intrinsic value quite explicitly. The onus to make a higher wage to improve their lifestyle is on the individual, given the freedom they have to allocate their time as they see fit. It's not incumbent upon others or businesses to pay more for tasks if a lower-cost, higher-value solution exists.

> It's not incumbent upon others or businesses to pay more for tasks if a lower-cost, higher-value solution exists. in the US, we've collectively decided (through elections) that there should be a minimum wage, from which it follows that we've also decided that the "wages have no bottom" argument, i.e. they should be able to go as low as possible without limit, is unethical and detrimental to society.

Yes that’s true. I’m not making the argument that “wages have no bottom”. As another poster has noted, the wage paid in this case is above the local minimum. But as the US participates in global trade, solutions that are extremely uncompetitive like those that rely on relatively higher costs of US unskilled labor, are bound to not be chosen. Forcing the decision makers to use sub-optimal higher cost solutions ultimately makes them uncompetitive and increases the risk of failure. I haven’t seen any logic to justify that.

Re: Kenyan moderators decry toll of training of AI

#62

Earlier quoted context omitted.

What's embarrassing about it? Don't you shop around when you need to hire a contractor or buy a car?

Buying/paying local helps my local economy.

I think there's times where this makes sense and times where it doesn't. Your local economy is probably better served by being paid to do something else. They probably have a comparative advantage at doing another activity. You'd be better off paying into your local economy where it makes strategic sense.

We're not talking about exporting critical manufacturing skills here. We're talking about doing a crappy job that's only worth doing and paying for if it's dirt cheap.

Re: Kenyan moderators decry toll of training of AI

#63
post #57

Earlier quoted context omitted.

It's almost exactly a good thing though. >"It's cheaper to live there" yeah, I wonder why that might be. It's incoherent reply - if a job does not require high qualifications, besides language knowledge, going for cheapest place is the obvious thing. >The numerous unbelievably wealthy global companies all doing this in the same very specific areas surely have the area's best interests at heart, and want to raise its…

Pretty much all this summarizes as "yes, it's valuable work to the individual". Broadly I agree - a lot of it is clearly a pretty big positive. But that doesn't make it not exploitative . For the same reasons that wealthy places don't generally allow child labor and have worker safety and environmental protections... but outsourced stuff frequently does not. The companies outsourcing are clearly aware of this, but mo…

> Pretty much all this summarizes as "yes, it's valuable work to the individual".

Individuals are making up the societies. Few dozen of people working relatively high income, are lifting whole countries up - that's how globalization have worked.

>But that doesn't make it not exploitative. For the same reasons that wealthy places don't generally allow child labor and have worker safety and environmental protections... but outsourced stuff frequently does not. The companies outsourcing are clearly aware of this, but most do or allow things that are blatantly illegal where they're headquartered.

Now you're not actually accusing them of anything, but _comparing_ to other potential bad things.

Those jobs aren't pleasant - as is working 12 hour shift in an ironworks - but they are damn better than subsistence farming and dying when there's draught one year.

And after those bad jobs, come progressively better ones.

Re: Kenyan moderators decry toll of training of AI

#64
post #39

Earlier quoted context omitted.

If you worked for Sama and was sent to work in Kenya, would you take a salary cut just because your cost of living has reduced? Pay people what they are worth, not based on where they choose to live.

You would not take this option because you prefer living in the U.S. (or wherever) and being employed there, but that's not the relevant counterfactual for the people doing this work, and it seems like it's preferable to other options they actually have, else it would hardly attract workers. Don't get me wrong, companies have an obligation to guarantee some level of workplace safety and healthy conditions etc.! But a…

That Kenyan workers don't have more favorable counterfactuals for employment doesn't place companies' behavior beyond critique, because the (collective) behavior of companies that employ workers in Kenya is partially determinative of what the counterfactuals themselves are.

If you fall off a bridge into a river and are drowning, and I'm the only person around, it's still exploitation if I try to make you promise to pay me $10k before I'll throw you the life buoy. It doesn't matter that, counterfactually, the alternative is death. It doesn't matter that you'd be far worse off if I wasn't around to offer you my help (at the maximum price I can extract).

Likewise, the fact that many Kenyans have no strong employment prospects doesn't mean that companies moving their labor to Kenya aren't wrongfully exploiting the situation. And that's leaving aside the fact that centuries of exploitation have created the economic conditions in Kenya. Benefiting from that is something akin to making money off the guy in the river, and it turns out your dad is the one who pushed him in to begin with.

Re: Kenyan moderators decry toll of training of AI

#65
post #46

Earlier quoted context omitted.

(I have edited to flesh these out a bit more, but have stopped since there is a reply) I've kinda waffled in my opinion of outsourcing through the years, but one issue I keep coming back to is that you can trivially view it as massive, intentional wage suppression. "It's cheaper to live there" yeah, I wonder why that might be. The numerous unbelievably wealthy global companies all doing this in the same very specific…

> The numerous unbelievably wealthy global companies all doing this in the same very specific areas surely have the area's best interests at heart, and want to raise its standard of living, so it must be something else. That’s the point of capitalism—-it works if you have only your own interests at heart. Wealthy Global Company A enters a market and makes huge profits. Wealth Global Company B (and maybe C, D, E, F, e…

> but prices also fall in Country A to make up for the fall in lost wages

I feel like the problem is that this part is usually skipped over.

Re: Kenyan moderators decry toll of training of AI

#66

Somebody once told me "yeah, without some slavery, THAT also wouldn't work", and I see it so often in products and concepts around us day by day. This isn't about the income gap. It's extracting work from a certain class for pennies, and create immense wealth from it on the other side of the globe without those people even knowing what it is about, let alone take part or benefit from it long term. After centuries wha…

Yeah outsourcing is terrible for the countries being outsourced to https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?location...

So China's GDP is a good measure for Kenyan well-being?

Re: Kenyan moderators decry toll of training of AI

#67
post #3

Why were human data labelers involved in the creation of ChatGPT? I thought the approach was to let the neural network predict the next word in a large corpus of text?

I don't know the scope of OpenAIs data cleaning or RLHF efforts. But I'd guess this is a big part of what's made them the leader. It's one thing to train an LLM on a big text corpus, as researchers do. It's another to actually wade into it and do the curation to turn it into a real product. I know there are public instruction datasets out there, but my guess is OpenAI has put a lot of money into getting cleaner data…

It's part of it but it's not necessary to get humans to do it. If you're aware of the Claude Models, they use reinforcement learning from AI feedback and they're great. The best Claude model is only behind GPT-4

Re: Kenyan moderators decry toll of training of AI

#68
post #3

Why were human data labelers involved in the creation of ChatGPT? I thought the approach was to let the neural network predict the next word in a large corpus of text?

You need people to evaluate the answers to improve the output and adjust the weights, there is a lot of manual labor involved at training every single one of these models, it’s not just magic.

There doesn't have to be manual labor. Anthropic (the Claude Models) use reinforcement learning from AI feedback (dubbed Constitutional AI) and it works great too.

Re: Kenyan moderators decry toll of training of AI

#69
post #64
post #39

Earlier quoted context omitted.

You would not take this option because you prefer living in the U.S. (or wherever) and being employed there, but that's not the relevant counterfactual for the people doing this work, and it seems like it's preferable to other options they actually have, else it would hardly attract workers. Don't get me wrong, companies have an obligation to guarantee some level of workplace safety and healthy conditions etc.! But a…

That Kenyan workers don't have more favorable counterfactuals for employment doesn't place companies' behavior beyond critique, because the (collective) behavior of companies that employ workers in Kenya is partially determinative of what the counterfactuals themselves are . If you fall off a bridge into a river and are drowning, and I'm the only person around, it's still exploitation if I try to make you promise to…

My perspective on this is more like "the locals will work x amount of time to get by either way, and you are offering them to work x amount of time and get a better living, or maintain their standard of living at lower x." (replace "time" with "time x enjoyment lost from missed leisure time", i.e. opportunity cost of time, if you will). You win because you get the work done cheaper or get to get more work done than you would "back home", the worker wins because they have a better alternative they didn't have before.

This hinges on the fact that I assume that if you take a job for OpenAI over, say, working at the local market, or your local family business, or doing subsistence farming, ... (I don't know the employment history of Kenyans doing this) you do so because you consider the job at OpenAI, in your own judgment, the better alternative. This assumes some amount of agency in the decision, yes — I would not argue the same if you e.g. forced prisoners to do the work for below minimum wage or whatever and then went and said "hey at least they get two dollars a day more than zero" (I exaggerate). Also, again, this assumes you don't mislead them about the mental toll and then hold them "captive" with sketchy contracts or whatever.

It also seems to me that you have an underlying assumption that companies in the global north could collectively decide to outsource as much labor as they are currently outsourcing, at wage levels matching their "home countries", which I don't think is economically true.

As for your analogy, I think I see the equivalence you draw, but I'm not sure I agree with it. For me the most natural read would be that the "payment" here is the workers' time, the ring you throw is the pay for the job. In that case, the payment (time spent doing some kind of work) happens either way (weird world where you always have to pay for getting saved by some strange law of the universe, I guess), and you're just offering a better ring, or the same ring for less time.

Re: Kenyan moderators decry toll of training of AI

#70

Earlier quoted context omitted.

Emergency responders have to deal with what I'd imagine the typical person would consider traumatic but we've agreed that their jobs are necessary in order for society to be efficacious. Not everyone has the same physiological response to disturbing content and it would appear people that take up these distressful emergency jobs are less likely to ever even have any traumatic stress responses in some studies IIRC. Ma…

You don't perceive any significant difference between emergency work and AI training moderation?

One of them seems to have actually traumatic situations.
Post reply on HN