A former work colleague, whose native language was not English, was telling me once about her frustration with a family member and how she was dealing with something, and Colleague said to her family member ‘are you stupid???’.

I do love non-native English.

Hi, PalFishers.

How you doing? Y’ok?

So teacher pay was cut on PalFish by let’s-call-it 40% (there’s a lot of variation in this because of tiers etc). Before we get to the other hoo-ha, let’s pause here for a moment.

Parent class fees went up and teacher pay rates went down because (something like) private investors bailed because they were told they couldn’t operate for profit. Or something. Whatever. The point is, money from outside the company that allowed it to operate as it was, with the staffing levels it had, charging parents what it did and paying teachers what it did — disappeared more or less overnight. And in order to attempt to keep operating, the company got rid of a lot of staff, at least doubled the workload of the (presumably cheaper) replacements for that staff, and committed apparent suicide by charging customers more for the same product while paying contractors less to deliver that product.

And your takeaway from this is … greed?

If they have to charge customers more for the same product *while cutting their costs by 50 percent* — baby, someone else was paying the shortfall before, and they stopped.

Why was someone else paying for millions (billions?) of hours of hundreds of thousands of adults of different ages and physiognomies and millions of children as they age over months and years on video talking in English in a vast array of accents and skill levels, expressing a vast array of emotions and varying levels of self control, in various states of exhaustion or energy, in situations ranging from calmly sitting at a desk to riding on a motorcycle, demonstrating different vocabulary using TPR and being super expressive to make meaning as clear as possible?

Think about how facebook and google etc. make money off your data—the whole ‘if you’re not paying for it, you’re not the customer, you’re the product’ thing. The ostensible customers of online English classes taught by native English speakers to Chinese children (the parents) were not actually covering the cost of the English teaching. The Chinese children’s online English teaching market paid double what online English teaching was paying in other places, even wealthy places. I’ve said it before and I’ll say it again: the English classes were a loss-leader for a different product. At least one of those products is a massive amount of data for machine learning—and not random, anonymous data, either — data linked to specific, known individuals with attached PII.

AI can detect feelings from a face, which is apparently being used to ‘rapidly identify criminal suspects by analyzing their mental state… to prevent illegal acts including terrorism and smuggling‘ in China. It can reconstruct faces using only audio. They’re already making virtual people who look and sound like the real thing: ‘Chinese government media outlet Xinhua has a virtual news reader who doesn’t have a mind of its own, and Fable Studio or Virtual Humans will create and operate a synthetic character or influencer for any brand that wants one‘. AI can imitate a human on video well enough to make it impossible to distinguish real from fake, and biometric security technology apparently struggles to stay one step ahead of the deepfake posse to keep all our digital everything secure. Evergreen (‘Evergreen’ what? Inc? Ltd? Technology Services?) has this to say about some of their products:

Emotion AI in face detection measures the facial expressions using any optical sensor like a standard web/ smartphone camera, detecting a human face in real-time, in a pre-recorded video or images. Computer vision algorithms identify the main points of an individual’s face: eyes, the tip of the nose, eyebrows, corners of the mouth, etc., and track their movement to decode emotions. By comparing this gathered data to a vast library of template images, facial expression detection software can determine the person’s feelings based on the combination of facial expressions. Advanced emotion AI solutions like those provided by Affectiva or Kairos can measure the following emotion metrics: joy, sadness, anger, contempt, disgust, fear, and surprise. Additional software features may include facial identification and verification, age and gender detection, ethnicity and multi-face detection, and much more.

https://evergreen.team/articles/emotion-ai.html

‘A vast library of template images’… This all comes from somewhere.

Remember when VIPKid teachers found out that AI was leaving feedback for students’ class performance? And the brainwave headband that detected students’ attention levels? That was ages ago. We’ve come a long way, baby.

Anyway, all that’s old news, and yeah, we’re now surplus to requirements. And the memory of all the ‘If the US doesn’t value its teachers enough to pay them fairly, China will’ news articles and facebook posts etc., all the Cindy Mi ‘invest in Chinese online education companies because it’s an infinite market with 16 million or whatever new customers/products born every year’ is blending in my (mental) ears with ‘this is below minimum wage!!!’ and similar outrage and the subsequent confused silence at everyone being summarily ‘fired’.

Funnily enough, I actually think it’s a good call, the pulling the plug on all this. Kids and families were being over-stretched in many ways. I do hope that overall this is better for the people of China seen in the aggregate.

I hope.

But what I *think* is that the CCP is teetering on the brink of losing the mandate of heaven.