202309221117-AI_and_employment: Automation and the ethics of work

The promise of automation

Reads:

Bertrand Russell on the promise of automation and technological innovation:

If at the end of the War the scientific organization which had been created in order to liberate men for fighting and munition work had been preserved, and the hours of work had been cut down to four, all would have been well. Instead of that, the old chaos was restored, those whose work was demanded were made to work long hours, and the rest were left to starve as unemployed. Why? Because work is a duty, and a man should not receive wages in proportion to what he has produced, but in proportion to his virtue as exemplified by his industry. This is the morality of the Slave State, applied in circumstances totally unlike those in which it arose.

In old times times leisure time was a luxury only afforded by small privileged classes of society. The promise of technological innovation is increased productivity. In theory this would mean we could start working less and have more leisure time. In other words, the promise of automation is a more evenly distribution of leisure time for workers throughout our communities.

What is a pattern we see in practice?

If at the end of the War the scientific organization which had been created in order to liberate men for fighting and munition work had been preserved, and the hours of work had been cut down to four, all would have been well. Instead of that, the old chaos was restored, those whose work was demanded were made to work long hours, and the rest were left to starve as unemployed.

Work is not equally distributed amongst the population.

Towards AGI: automation squared

AI as an existential risk seems quite overblown but the following Statement on AI Risk (Center for AI Safety) is signed by many prominent figures in AI & Tech. The statement is:

Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.

Automation is reaching an unprecedented level with current developments in artificial intelligence.

OpenAI is a believer in AGI.

An AGI is an AI that is not only better in some specific task, but across a wide range of tasks. As such, AGI is the promise of automation squared!

Oude notitie omschrijven:

Er worden bijvoorbeeld aardig wat boeken geschreven over de existentiële dreigingen van Artificial General Intelligence (AGI) of zelfs van "super intelligence". Tot dusver was AI beter in specifieke taken, maar een AGI zou mensen evenaren of verbeteren over de hele linie. Dat klinkt als een ver-van-je-bed-show, maar OpenAI, het instituut achter de ontwikkeling van de meest bekende GenAI ChatGPT en Dall-E, ziet de ontwikkeling van AI op basis van large language models als een serieuze stap in die richting. Sterker nog, het ontwikkelen van veilige AGI is hun missie statement. Als je bedenkt dat tools als ChatGPT getrained kunnen worden op de "output van de hele mensheid" en dat zoveel informatie op één punt gebruikt kan worden, dan klinkt het steeds minder als science fiction.

Mocht super intelligence realiteit worden, dan is de vrees dat het al dan niet voortbestaan van de mensheid een kwestie van coulance is, net zoals mensen hebben besloten het uitsterven van de panda te voorkomen. De panda blijft bestaan zolang wij dat de moeite waard vinden, de panda zelf heeft bijzonder weinig te zeggen in deze zaak. Oftewel, de vraag is: past de mens nog bij de doelen en het grotere plan van de cognitief superieure AI, of zijn ze als een irritante vlieg die achteloos wordt dood gemept? Dat heet in AI safety onderzoek het AI alignment probleem: bijvoorbeeld, als machine learning modellen louter gericht zijn op het optimaliseren van een loss function, hoe zorgen we er dan voor dat dat doel in overeenstemming is met menselijke waarden (zoals: het niet doden van anderen)?

AGI and universal income

If AGI is achieved, the idea is that AI will be vastly more productive than human labor, to the degree that human labor becomes relatively useless. Our economic system and distribution of resources is anchored in the idea that humans work though. What happens to this system when AGI arrives?

OpenAI's Sam Altman is registering people for his crypto WorldCoin (through Iris scans) with the promise of paying a universal basic income to people once AI is the main source of production and wealth. source

AI: a democratic tool or a new social divide?

We are now in a phase where AI is becoming increasingly available to the general public.

Thesis 1: On paper then, this development could be democratic in the sense that it allows the broader community to reap the benefits of Gen AI in their daily work! (Underlying question to be answered below: who reaps the benefits of automatization, the workers or the employers?

Thesis 2: Increased productivity due to AI tools will cause a new social divide, between the automaters and the automated. People whose jobs will benefit from automatization will either 1) get more work because they can perform their tasks more efficiently (potentially taking away work from others) or 2) risk losing their job because it can be automated to a large degree; a single person can now do the job of five.

In other words, will there be a repetition of the pattern that was shown historically so far? This new social divide will, in our times, largely be drawn based on who is digitally literate or even more, AI literate. Technological innovation will create new more specialized jobs for a small group of digitally literate experts and take away work from others.

From my own article:

But instead of everyone then working half working days, a part of the population (those on the “right side” of automation) only seem to make longer days, whereas others become unemployed and see their life quality plunging (those made “redundant” by automation). If automation only contributes to the good of employers, then technology will only increase social divides along new lines, between “normal” workers and the those that are tech-savvy.

Even without automation we see this process within teams: Those who work efficiently will get more work to be done. Also in organisations you'll see that people who deliver good work will be increasingly sought after.

Who reaps the benefits of automation, workers or employers?

Question to the audience: assume you would be able to automate your current job to such a degree that you could do your 8h work in 2h? You would come in the office every day, work a few hours, and then slack off or go home while your co-workers plod along. Would this feel like cheating to you?

In our type of profession we are typically measured by results (we think). But is that really true? I would feel bad about the situation above, so apparently the results the deliver are not the whole story!

The situation where we can do the same work in a shorter time, and then have more leisure time, would be in the benefit of us, the workers.

Consider this insight from Bertrand Russell (in 1932!):

Because work is a duty, and a man should not receive wages in proportion to what he has produced, but in proportion to his virtue as exemplified by his industry.

We are also valued by showing up in the office early, staying late, and looking busy! So apparently "business" is a value that is deeply engrained in our culture, to the point where automating parts of our job without telling our employer feels unethical to many!

See the article on automating our own jobs: “Is it unethical for me to not tell my employer I’ve automated my job?”

Even if a program impeccably performs their job, many feel that automation for one’s own benefit is wrong. That human labor is inherently virtuous — and that employees should always maximize productivity for their employers — is more deeply coded into American work culture than any automation script could be.

And my own article:

We could say that these clever programmers participate in a form of grassroots automation that emerges bottom-up rather than being issued top-down by some executive in a reorganization. This creates a slightly different set of issues than the more straightforward story of “my job got replaced by a machine”. The difference is that in the latter case the process of automation is publicly evangelized, whereas what makes the case of self-automation poignant is the ethical question of disclosure: Should I tell others, or my employer? And should I share my scripts? And why, or why not? These considerations concerning work automation only become more relevant as AI technologies become more widely available.

Or... big tech!

If AI-based automation and tooling becomes part of our daily work (and it will be! it will integrated in our browsing, coding, even making powerpoint slides) we also become increasingly dependent on a few very large (and often non-European) countries.

So who reaps the most benefits if we become dependent on AI-based automation?! Big tech (and perhaps to some extent the niche experts with top-jobs there)! They are the only ones with the resources to train these big AI models.

Case study: coders

TODO work out questions asked with respect to this group.

How many of you already use ChatGPT in your daily work?

Will AI complement or replace their work?

Which parts of their work can an AI not do?

On which side of the societal divide will coders be? On the one hand they are tech savvy, on the other hand much code is boilerplate and follows common pattern so they are a candidate for automation (whereas garbage collectors are probably not automated any time soon!)

TODO

Techneuten machtiger worden in de samenleving.

Het technische riedeltje van AI hebben ze juist wel vaker gehoord.

The value of human labour in times of gen AI

Humans created the training data

In addition to not really helping those in need, deployment of this technology actually hurts workers: the systems rely on enormous amounts of training data that are stolen without compensation from the artists and authors who created it in the first place. (Bender & Hanna, 2023 source).

De-valuation: Humans as the checkers and correctors of the output of automated systems

Human labour is replaced with GenAI systems. In some businesses, AI output increasingly becomes primary and human labor secondary, demoted to checking and touching up the outputs of Gen AI.

In other words, industries which previously supported stable jobs turn into gig economies of low-paying freelance work.

Cf. what British economist Guy Standing calls the "precariat" (precarious + proletariat) the Standing-Precariat not bogus.

  • AI is already taking video game illustrators’ jobs in China
    • "But since February, these job opportunities have vanished, Yu told Rest of World. Gaming companies, equipped with AI image generators, can create a similar illustration in seconds. Yu said they now simply offer to commission her for small fixes, like tweaking the lighting and skewed body parts, for a tenth of her original rate."
    • AI-generated art was so skilled that some illustrators talked about giving up drawing altogether. “Our way of making a living is suddenly destroyed,” said a game artist in Guangdong, speaking on condition of anonymity for fear of being identified by her employer, to Rest of World. Yu, the freelance illustrator, said it was “despicable” that algorithms — trained on vast datasets that took humans decades to produce — were on the verge of replacing the artists themselves. Still, Yu plans to train AI programs with her own drawings to improve her productivity. “If I’m a top-notch artist, I might be able to boycott them. But I have to eat.”

Digital carcasses.

Re-evaluation

The quality and value of AI-generated

Cf. section in AI information pollution for links on the deterioration of AI model quality once synthetic outputs are reused as training inputs for models.

Will this lead to a re-evaluation of human labor?