Machine Dreams

I wrote most of this series last year, and it pretty much all holds up. There really hasn’t been any progress in tech. This video was written about the constant threat of “robots taking our jobs”. This is the one area where there have been some major developments. In the face of more automation than ever before. There’s currently widespread worker shortages.

I won’t change anything, because it is still a prominent take, and I’m sure it’ll come back with a vengeance. But rather than being in opposition to mass opinion, this video’s now more like kicking them while they’re down.

MACHINE DREAMS

We’re not the first society to get technophobic, towards the end of Ancient Greece they were worried that the catapult would replace the job of the soldier. And there are cases of engineers or inventors being rejected or killed because they had discovered technologies that would threaten the very fabric of society. Technologies like: a glass bowl, or a pulley.

But maybe that’s not a fair comparison. That’s technophobia sure, but it’s not really automation. 

At the start of the industrial revolution, it was quite popular for rich dicks to demonstrate their technological power with automatons. Like a little hand-cranked robot. Like there was a duck that shat.

But probably the most influential one, was “the Mechanical Turk”. The Mechanical Turk was an automaton that could play chess. It was this robotic man, and a desk full of machinery, with a chessboard on top. And it toured around, they’d always make a big show of showing everyone the inner workings, and it would beat people at chess. It famously beat Napoleon at chess. I don’t know how good he was at chess, but that’s a guy that we’ve heard of.

Now there were two-parts to the Mechanical Turk. The first was the mechanics of moving the automaton, to have it pick up chess pieces and move its head; the second, was the illusions to hide the chess player sitting inside.

Now this sort of frivolous bullshit was happening because it was peacetime, there was a lot of economic prosperity, and there was the technological opportunity. And ultimately those forces would grow to form the industrial revolution: There’s the physical possibility (the resources were available and the steam engine had been floating around for a couple of centuries), the economic opportunity, and skilled workers were being paid a lot to produce very little. 

So the industrial revolution happened. I mean, obviously it’s a very hotly debated topic that doesn’t have a single tidy answer. Except for that, that is the answer, I solved it. That’s why the industrial revolution happened.

Uh, and the new technologies made production processes much more efficient. Which dramatically increased the amount of work, dramatically increased the amount of goods produced. Though it did kind of fuck everything up.

But it’s interesting to look at because the effect was the opposite of our current fears, things were automated and it actually increased the amount of work. It wasn’t great for quality of life, but the result was just sort of shuffling people around the economy. [Joke]

But maybe that’s not a fair example, that’s not actually automation. Of course “Automation” was invented by the Ford Motor Company in the post-war boom. They were thriving, their production workers were demanding more pay, there were a lot of new technological developments. So, the Vice-President for Manufacturing, introduced a new word, ‘automation.’ And at great cost they were able to replace much of their workforce with machines. Which increased their productivity. So it started a wave of automation across the globe. Which did two things, it displaced workers, and it made the goods cheaper because they cost less to produce.

There’s only so much car you can buy, so people started to spend their extra money on services. And again, instead of destroying jobs, there was a transfer of supply and demand, as those factory workers became service workers.

Our current wave of automation hype comes from the new, disruptive, tech companies. The new, disruptive, tech companies rely on a business model of snapping up all potential users, and essentially locking out all competition. They lose billions upfront and the idea is that eventually they’ll have underbid or absorbed all competition, so as the only option left, they’ll have total market dominance and make super-profits.

And then once they’re in that position, instead of using their record profits and record amounts of borrowed money to develop new tech, or raise wages, or genuinely invest. 

They do stock buybacks
buy up competition
protect their ip
influence laws
Which has conveniently been reclassified to count as “investment spending.”

[streamline. pointify. more about them ensuring stagnation with a little sprinkling of “they have nowhere left to put the money”]

Basically what all that means is that every app looks like Instagram which looks likeSnapchat which looks like TikTok which looks like Snapchat which looks like Instagram. 

So really the new, disruptive, tech companies are in the business of ensuring stagnation. 

And if we subtract those top few tech companies from the s&p, you’re left with 494 large-cap companies with no growth.

Or at least that had been the case for a decade when I wrote this, and it’s a pain to check, so it could have changed, but overall, that’s the general trend. And it ultimately forms a larger percentage than the one preceding the dot com crash

While the markets are looking pretty good, those tech companies are masking a larger stagnation. The entire market is being kept afloat by corporate borrowing, most of it dogshit. […is “Covenant Lite” the corporate equivalent of the subprime loans behind the Global Financial Crisis.]

[Joke]

One of the things that freaks the economists out is that this computer age hasn’t affected the productivity statistics. You’d think adding computers would make us more efficient, but actually, this has been the worst decade for productivity since the late 18th century. One of the few times it did spike was immediately following the Global Financial Crisis. Output fell a lot, but hours worked fell even more. And people who kept their jobs were scared of losing them so worked harder. Meaning there was a net increase in productivity. 

So. Faced with their employees demanding a livable wage in 2013, McDonald’s came up with a new approach. They started to publicise that they didn’t need employees at all. Their CEO was like “McDonalds is going to be all robots, so if you’re working, you should take what you can get. And in fact we should be allowed to pay you even less.”

And to go with this, they introduced the self-serve kiosk. Now the self-serve kiosk does look like automation. But if we think about it for a second, it hasn’t actually automated anything at all. You push the buttons on the screen instead of an employee, so they’ve automated the task of communication, but that’s about it. It hasn’t actually reduced McDonalds staffing requirements at all.

Of course the jobs that can be replaced by this task automation, are all the made up ones. Accountant, CFO, real estate agent. What you’re not going to replace is service work, teacher, fast-food employee, y’know, just all the essential workers. Automation is happening but not where we think, and not as much as we think.

In a lot of the cases where we hear about machine learning being used, the story’s being distorted. Half the tech startups claiming to use AI, don’t use it at all. Boston Dynamics robots have more in common with a remote control toy than machine learning. You hear about AI detecting cancer, that’s not a robot that can divine some sort of hidden truth that we humans can’t see. That’s after this algorithm receives 1000000000 human labelled examples of what is/isn’t cancer, it can somewhat accurately classify examples itself.

Which isn’t particularly useful. In fact we’ve seen that ML is medically useless, so many people were trying to solve Coronavirus with AI. Not one did anything of value.

Now in AI research there was a massive push to solve the coronavirus. To create an algorithm that could predict the case numbers, diagnose patients, or help with a vaccine. Hundreds of predictive tools were made. You might not have heard about this though, because none of them were successful. Not one. Every traditional statistical model easily beat out pulling patterns randomly from the past to try and predict the future.

[Recap More Data]

But right now, that’s all we’ve got, just add more data, and hope improvement happens. 

And this is a problem that Amazon identified early on. If you want to train a voice recognition model, you need an insane amount of examples of transcribed audio. If you want to detect penises you need an insane amount of examples of what is/isn’t a penis.

So they set up a marketplace where people can log on, do a bit of this data entry, and get paid jack shit. They called this MTurk, after the mechanical turk, as a sort of cute wink at their hidden labour exploitation. 

But while that’s a good service for researchers and small scale projects, it’s not quite exploitative enough for a big company developing a smart device. So when Amazon entered the voice recognition game with Alexa, they started hiring thousands of contractors in India, each with the job of transcribing ~700 failed Alexa calls every day (because someone had an Australian accent), they transcribe those, reinput that back into the training, and Alexa improve by a millionth of a percent. And if those contractors ever drop below 97% accuracy in their transcription, that’s it, they’re gone.

So whether it really is machine learning, or if it’s just tech startups lying to attract investment, it requires a massive amount of hidden work. And sometimes it is easier and cheaper to just outsource your “AI” to underpaid humans.

Now, Amazon has set up a few overpriced convenience stores. Their selling point is that you don’t go through a checkout. You walk in, grab what you want, walk out, and you get charged. And it is quite a feat of combining current computer-vision technologies into a nice bit of entertainment and brand building, while getting a good consumer dataset to sell.

Now the thing is, even though there’s no cashiers, they have a full staff of people walking around the store, making sure you’re interacting with it correctly, there’s security, and on-site tech. They accept any losses and anyone can remove anything from their bill, no questions asked. And then behind the scenes, you have a massive workforce of internationally outsourced labor: there’s like a thousand people whose job it is to review the surveillance footage, and the algorithms, and manually verify and refine them. Double-checking everyone’s purchases in the hopes of narrowing down its failure rate. 

They’ve taken the convenience store, which is usually staffed by what, 2 people? And they’ve exponentially increased the required staffing. It’s the threat of the McDonalds kiosk and the illusion of the mechanical turk.

And it doesn’t really have anything to do with actual automation.

Automation is being very specifically deployed to juice maximum productivity out of as few people as possible. The warehouse worker or office worker who works alongside automation, works harder. Rather than robots taking jobs, which they can’t do, they don’t have agency. That’s an executive decision, to re-allocate work. Your boss is taking your jobs.

But it’s not that jobs are being replaced, it’s that tasks are being automated, and then basically the same amount of work is consolidated to fewer people. This tech which could ease workloads, is being used very specifically to increase them. Isolating people and turning employees into contractors. 

And that has nothing to do with the technology, it has nothing to do with upcoming advancements in technology (the tech being used to do the automation that is happening has been around for decades). It’s about who is doing that automation, what are their incentives, and what work has value.

Because out of context, automation is an incredible tool that we love. It’s a staple of our fictional utopias. But our current market incentives override that, and demand that it be used in as exploitative a way as possible.

And even this looming threat of mass automation would just lead to some more shuffling around the economy. Which is scary for a lot of people. Because of course the work that will never be replaced, the work that matters to us the most, the essential work, has no financial value.

[TECH STAGNATION]

But realistically, the tech companies aren’t on the brink of causing mass automation. It’s not their focus. They’re not on the brink of anything. Their goal is just to kamikaze their way to the top of the market and then sit there, beating competitors away with a stick.

So automation will continue, but the immediate concern isn’t that a dancing robot will take your job, it’s that these companies are actively exploiting you and the world now, in this age of stagnation.