How do humans with very little exposure to grotesque violence or extreme content universally label such content so well? This is not graduate level data that needs labeling.

What is missing in an AI model for it to intuitively understand what content is extreme from very few labeled sample in training?

> Murmu, 26, is a content moderator for a global technology company, logging on from her village in India’s Jharkhand state

> With just four months left on her contract, which pays about £260 a month

Earning US$350/mo working remotely in a village in one of the poorest states in India is an extremely competitive given that the alternative would be spending 12 hours sewing fast fashion for Zara earning US$130-150/mo [0], doing bit piece ag labor for around US$100/mo and participating in MGNREGA for US$50/mo, become a housewife, or become a Naxalite/Maoist insurgent to earn a couple thousand dollars when surrendering [1].

Content moderation means interacting with extremely depressing and horrid content, but someone needs to do it, and once models get good enough we would start seeing articles about how "all the good 100% remote first jobs with no barrier to entry" are being automated to oblivion.

Yes it sucks, but the alternative is becoming a migrant worker or working in light manufacturing where QoL is worse. Heck, we used to see similar articles about Chinese workers for Apple barely 14 years ago in then equally poor Sichuan [2], but you don't see those kinds of articles anymore.

Development takes time and the fact that US$350/mo remote data annotation and content moderation jobs are now penetrating into villages in what used to be the Naxalite/Maoist/Red Corridor where bombings and gun battles were a part of normal life just 10 years ago [3] is a massive step up developmentally - it means that there is robust enough internet, literacy, banking, and public services penetration for the seeds for a services economy to form.

Edit: Thanks for the downvotes westerners - my family is from these kinds of villages in India and Vietnam. The alternatives are extremely bleak - especially for a tribal woman like Ms Murmu at the bottom of the social and patriarchal hierarchy.

[0] - https://theprint.in/ground-reports/industries-finally-return...

[1] - https://www.thehansindia.com/news/national/18-yr-old-maoist-...

[2] - https://www.nytimes.com/2012/01/26/business/ieconomy-apples-...

[3] - https://www.newindianexpress.com/nation/2016/Nov/23/six-maoi...

Eww. Like 19th c. children in dangerous factories, abusing poor people's mental health sifting through the Global North's cavalcade of depravity. There must exist more productive and honest uses of people's time, and some jobs shouldn't be done for any amount of money. Some jobs done risky ways shouldn't be done by human beings at all in dangerous manners (coal mining without safety equipment, loom maintenance while running, carrying sulfur chunks out of active volcanoes) because they lower us all. "But they're making money" is not a good enough excuse because that's a false choice as there infinitely other activities, and any number of safer activities or similar tasks done with meaningful precautions are needed, desirable, and could be done instead.

[dead]

[flagged]

When VCs and investors keep saying 'There will be new jobs', they never tell you exactly what they are - on purpose.

Now we know that it is actually being a data labeller, AI tutor and content moderator, but in very low wage countries such as in India.

This is the post-AGI reality. 'Abundance', but not for you.