How the AI industry profits from catastrophe

In the olden days, working conditions in places like coal mines, textile factories and garment factories, to name but just a few, were pretty dreadful. The capitalists got rich while bearing down hard on the workers.

Much scaled down now, although stories are still told about life in garment factories and down on the farms – where I read just today that conditions are so grim that even Ukrainian refugees can’t hack it.

Rather different were the home workers, exploited in a different way for the privilege of balancing the duties of home keeper with those of bread winner. So when we were little, the lady in the flat under us, supplemented her husband’s income by assembling slippers for some derisory sum per pair.

Rather more recently, we had the arrival of Mechanical Turk on the scene, a member of the (trade union hating) Amazon family. The business model here was that people with clerical work to be done could package it up and recruit workers through Mechanical Turk to do it online. And there were lots of places where there were lots of people prepared to do this sort of home work.

And at around the same time, a whole industry grew up around the huge demand from various AI flavoured projects – self driving cars being one of the biggest – for labelled data on which to train the AI. In the case of self-driving cars I think the idea is that you look at an image of a street scene and tell the computer what you can see: cars, lamp posts, pedestrians, whatever. And for this particular application, accuracy is very important.

Accuracy and speed being delivered by robust, computerised management of the workforce.

On this story though, Amazon did not deliver the goods, while companies like Appen (reference 2) and Scale AI (reference 3) did. They are now doing very well, with lots of employees in troubled parts of the world, for example Venezuela. 

In what is now a very competitive industry in which, if you want to get rich, you have to force speed up, accuracy up and labour costs down. Odd that Amazon didn’t get the idea, as this is very much their mantra.

On this story also, there is an upside. The pay and conditions might not be great, but it is a way for a lot of people to get paid employment of some sort. Paid employment of any sort being in short supply in places like Venezuela.

But in the round, depressing how the rest of us, comparatively well-off, are only too happy to dump part of our cost of living on unseen others.

PS: and while we are on robust management of the workforce, I also read this afternoon at reference 6 of a 'very tragic case involving RaDonda Vaught, who was an ICU nurse [and] who was recently convicted in Tennessee of criminally negligent homicide and gross neglect of an impaired adult. She accidentally administered a paralytic medication, vecuronium, instead of a sedative ...'. I know little more than this about the case, but if we are going to start sending medical people under pressure who make mistakes to jail, we might soon be rather short of medical people.

References

Reference 1: How the AI industry profits from catastrophe: As the demand for data labeling exploded, an economic catastrophe turned Venezuela into ground zero for a new model of labor exploitation – Karen Hao,  Andrea Paola Hernández, MIT Technology Review – 2022.

Reference 2: https://appen.com/. One of the older big players.

Reference 3: https://scale.com/. One of the younger big players.

Reference 4: https://psmv3.blogspot.com/search?q=mechanical+turk. Previous notice of Mechanical Turk.

Reference 5: https://psmv4.blogspot.com/search?q=mechanical+turk. More previous notice of Mechanical Turk.

Reference 6: Are All Medical Errors Now Crimes? The Nurse Vaught Verdict - Robert D. Glatter, Megan L. Ranney, Jane Barnsteiner, Medscape - 2022.