I will admit that I have not done a lot of research into the topic and might be slightly biased as a Kenyan.
Edit: The tasks that are currently outsourced are data labelling NOT training of AI models.
This is just to ask whether anybody else is concerned about a tendency by major tech companies to outsource labour needs to the global south? The services outsourced are either training of AI models or for a longer period, content moderation of social media sites.
There are instances where this work is outsourced to countries such as Kenya or India. For reasons that I can only assume which include cheap labour, lax labour right enforcement and fewer regulatory restrictions in these countries. Similar to the reasons that guide the outsourcing of manufacturing jobs to the global south.
My general worry is that in future, the global south shall become the training ground for more harmful AI projects that would be prohibited within the Global North. Is this something that I and other people should be concerned about?
It is hard to make any confident claims when so much depends on the details. As a very simplistic example, a US-based firm could outsource manual content moderations to a Kenyan for a wage that is low by Kenyan standards (as described in the links you shared), or it could pay a Kenyan an wage that is excellent by Kenyan standards to do a “good” job.
I care about labor rights, and it is no surprise that some US-based companies pay small amounts of money to people in underdeveloped countries for unpleasant tasks. I dislike it, but it doesn’t seem very tractable to me. Imagine how incredibly hard it would be for the US government to adopt a law that all outsourcing must comply with US labor law.
There is also the old argument that this job may be terrible, but people choose to do it because it is the best option available to them, it is better than the alternatives. Many young people from rural areas migrate to cities and work very unenjoyable jobs in factories because they find it preferable to staying in the countryside and being a farmer. I’m not fully convinced by this argument, but I do think that there is some aspect of it that makes sense: these people are choosing to do this job rather than other jobs.
What you have stated is entirely true. However, there are also contentions that at least in Kenya, Meta violated Kenyan labour laws in its use of a content moderation firm. Meta tried to claim rather unsuccessfully that they shouldn’t be bound by Kenyan laws since they aren’t based in Kenya. This can be found here and here, I think it is slightly indicative that the use of outsourcing companies is an aim to limit their liability in jurisdictions outside of the US. The case hasn’t been concluded since it has been sent to mediation. It is what raised this question in the first place.
It does seem bad if prohibited dangerous models could be trained in other jurisdictions, but the issues there—about regulating developers and datacenters—seem quite distinct from data labeling and the like, which seems to be the sorts of services that are currently outsourced in the way you described.
You’re 100% correct. It is data labelling (an oversight on my part)
Due to current outsourcing being of data labeling, I think one of the issues you express in the post is very unlikely:
Maybe there’s an argument about how:
current practices are evidence that AI companies are trying to avoid following the laws (note I mostly don’t believe this),
and this is why they’re outsourcing parts of development,
so then we should be worried they’ll do the same to get around other (safety-oriented) laws.
This is possible, but my best guess is that low wages are the primary reason for current outsourcing.
Additionally, as noted by Larks, outsourcing data-centers is going to be much more difficult, or at least take a long time, compared to outsourcing data-labeling, so we should be less worried that companies could effectively get around laws by doing this.