Hacker News Reader

1. These jobs are situated in Africa for the precise reason that other usual outsourcing countries have become increasingly expensive and prohibitive for this type of work. These workers are exposed to the most vile things the human mind can produce and they do it for starvation wages day in and day out.

2. Their leverage is their sanity and willingness to be able to be exposed to unconscionable content. Many places in the world would categorically block this kind of gig simply for being damaging to the mind, and rightly so. And even in countries where this work is permissible, it would otherwise expose the company to future liability in terms of emotional trauma.

3. This 80% figure is a number fabricated from thin air. These workers don't work in call centers. They are content moderators in some shape or form -- a job that requires not just reading comprehension but being able to weigh certain current events, cultural queues, unspoken and spoken language, and other data that you can't magically feed into a machine.

S

These workers are not slaves, they are free to work anywhere else. If moderating posts on a computer really is worse than say working at a farm, or a a clothing factory, then the workers would leave.

They don't work in call centers of course. But the workers who currently work in call centers, can swoop in and compete for their jobs. I can tell you that plenty of people on HN, would rather deal with moderating awful content, than to talk to endless waves of people who can't read 3 step instructions. And I would prefer either compared to say diving in sewage.

If you don't think GPT-4 can replace offshore call centers very rapidly, then you need to talk to those companies that do employ them. Onshore call centers handle more complex stuff that is harder to outsource, offshore ones not so much.

The workers could leave. Or they could unionize and attempt to improve their employment conditions. Maybe it will work. Maybe the company will fire them and replace them with other workers.

But I have to imagine that the workers’ own assessment of their leverage is more accurate than someone on the other side of the world who read an article headline. And it would appear that at least some of them, based on that assessment have decided to give unionizing a shot.

I'm not sure if saying "quite a sad understanding" is really an appropriate or deserved response in this scenario.

Like anything, there's a lot more beneath the surface. Many white collar folx tend to entirely write off Africa and other third world countries and misunderstand some of the dynamics there.

Yes, of course there will be a decrease in demand, that does not make unionization bad.

You also seem to underestimate the ability for African people to band together into collective movements. Yes there will be scabs, of course, but this is a step forward in the fight.

I do find myself feeling frustrated when someone approaches the comment section on a journey that has taken technical workers _several years_ through the courts system and then offers a dismissal that seems to suggest that the problem could be solved if they thought of it differently or didn't realize how shortsighted they were.

I do hope Africa is able to follow through on its ambitions to have serious punching weight in the global sphere, all problems aside. It is a beautiful continent with many beautiful cultures in it. It certainly feels a bit like a kind of home for me.

It is true that Africans can band together in movements, that's how they forced Europeans out.

It also true, those 'collective movements' don't necessarily result in good outcomes. Look at Kenya on the map. To the east is Somalia, to the north is Ethiopia, South Sudan, Sudan, all countries locked in civil war between different 'collective movements'.

The big opportunity for Africa in the AI era, is to use AIs to serve local healthcare/education/software development needs. There weren't many teachers/doctors/SE engineers etc to begin with, so they have little to lose from mass deploying AI.

I'm not sure unionising to get higher wages from moderation work will make a big impact. I am really curious to see the counterargument here, why will unionisation work in this case?

I can't claim expertise on Africa, something I would add though is that many Africans from Kenya, for example, that I've talked to, generally blame British imperialism for many of the problems today. Because they drew country borders willy-nilly around tribes, sometimes even splitting tribes between countries, and culturally the tribes only vote for their own, the countries are ruled by majority tribes which become the rich and leading class. Ethiopia has this problem too for example. Then tribalism + continued tensions just keep going from there.

Of course there is the rampant corruption. There are also collective community policing strategies too in some communities, for example.

In the African Union I have high hopes for the continent, but still I think there is a a very large, very large, very large barrier to overcome because of the infighting.

As far as the unionization, I think if handled right it can have lower downsides for the amount of bargaining power that it provides. It seems that a lot of African culture thrives on ad-hoc person-to-person movements, so this feels very much within the lines of what could work in Africa. But tribalism inside of each country and between countries could still keep any local movements from having a larger impact within the continent even if, say, a local union is successful.

In any case, it basically moves the Nash Equilibrium but doesn't necessarily abolish it. If handled poorly, or if the company retaliates, then sure, if the company is willing to run the press gauntlet and such. But at the very least, I have hopes.

Also I know of the conflict in Somalia, Ethiopia, South & North Sudan, etc. I haven't checked in on Kenya in a while. I hope all is well over there, it is generally a very beautiful and safe country with a few notable exceptions. Has something happened, or is it in the list mainly due to proximity to the conflict (or something like that?) :'( </3

Kenya is a true bright spot (and really the only one) in the horn of africa region. They have typical internal tensions, but they are a real democracy, I don't think they are interested in civil war given their neighbour's poor examples.

That being said, "Africans from Kenya, for example, that I've talked to, generally blame British imperialism for many of the problems today" is a typical and unimpressive mentality. Tribal conflicts existed way before the British did.

For example, Sudan existed before the British came, and the Arab vs black fights in Sudan is not a British creation. Somalia is mono-ethnic, doesn't stop it from endless wars. Ethiopia was not colonised beyond a short Italian occupation, their current civil war is still the result of ethnic conflicts.

African unity is also a questionable concept. West, Central, East and Southern Africa are four different continents in terms of culture, ethnicities, climate, and size. The only consensus is 'no-changing-borders-by-force', yes, they know that those colonial imposed borders is the only thing keeping africa from falling into a bloodbath free-for-all.

That's why AI is really africa's greatest hope. They have not invested much into white collar knowledge and industries, and AI allows them to leapfrog much of it. It'll be very interesting to see how the african public reacts to AI.

I really and truly hope so. I have a lot to learn about the continent, and I love it much. I've done some work on significantly reducing the experiment cost for small experiments on GPT and convnet models (WikiText and CIFAR10, respectively), and one of the big things that I thought of was how it could help Africans in AI specifically because it allows for good research test signals but isn't prohibitively expensive/slow.

On A100s this particular CIFAR10 network trains in under 7 seconds for example and the network (fp16) is <10 MB IIRC, the relative gains over other previous training speeds is even better on older cards IIRC due to it being pretty memory limited at this point. It could be used as a backbone in other networks and is nice because it is good to run experiments on and also can directly deploy at scale pretty easily. In my experience, cycle time is actually one of the biggest resources for most smaller research initiatives.

I just haven't had the opportunity to talk to anyone yet to see if there's a way that I could interface with the African AI community over it. I don't know how Colab works but they work pretty darn well in the free version of Colab, and both are annotated to help a moderately well-motivated learner pick up some of the ropes themselves.

It seems like you know a little about the space. If you do know anyone or have any pointers as to where I could/should look to talk to people about this, please do let me know! I think OSS can do a lot and would love to give a bit to the community, for sure.

I'm really not sure there's much of an investor interest in helping Africans train their own AIs, your data center costs will be way higher in Africa. And the talent pool there to help with AI is still very small.

There will be a large interest in selling AI to Africans, GPT-4 today, lightly modified, can already radically improve African education. The vastly underpaid and overworked teachers no longer need to waste time in preparing lessons, they just need to control the classroom and perform assessments. Also GPT-4 probably dominates your average English tutor in Africa...

The problem is the cost, a GPT-4 subscription is far beyond the means of an average African. The bottleneck is with server-side inference costs, once that is solved, I expect a AI gold rush in Africa, where western companies compete to sell AI subscription services. Your average African already owns a cheap andrioid phone with 3g connection, which is good enough for an AI chatbot.

> To the east is Somalia, to the north is Ethiopia, South Sudan, Sudan, all countries locked in civil war between different 'collective movements'

Even my limited knowledge of the situation is enough to tell that this is neither historically accurate nor a fair portrayal.

This does not sound like a good faith argument.

No collective movements means no American revolution, no Civil war, no French Revolution, no Glorious revolution in Britain.

We'd still bow down to Kings and have slavery

> 3. AI will annihilate probably 80% of the offshore call center business in a few years, the workers laid off from call centers, can 1-1 slot into these AI-RL training jobs. Heck AI will also annihilate 80% of those offshore moderation jobs for say facebook. So more competition.

I am dreading that day. Most often when calling a business I need someone to fix something and often it can only be fixed by someone that can make a decision that is kind of "outside of the rules". AI will be trained to always stick to the rules, so I foresee an even more frustrating call center future.

The present flock of LLMs is not exactly good at that whole "stick to the rules" thing beyond surface compliance. GPT-4 was touted as a significant improvement in that regard, but 1) it was jailbroken within days, and 2) most of the methods used to do so still work today.

My experience is that in big companies people stick just as much to the rules (or "processes") as any LLM would. You wouldn't want to know how often sensible and straightforward requests get obstructed because "there's no process for that".