There are about a thousand and one debates about the ecology of AI, the ethics of AI, the governance of AI. As the technology advances, the debate with the slipperiest footing is on the actual efficacy of AI: does this work for this application? Or for this one? Ultimately, whether LLMs “work” as a technology or not doesn’t matter. The function of a system is its output, and no matter what utility individuals find in the tech, the widespread adoption and deployment should be considered and regarded firstly as a threat to the labor force.

To illustrate, we’ll consider a few scenarios.

This is a short preface to a larger piece I’m working on called The Threat, which covers AI mandates at work, harm reduction in the use of LLMs, and long-term strategies as industries expand their adoption of this technology.

AI works

O, great miracle — the tech advances! It will soon be as good or better per dollar than its human equivalent. Under the current economic system, there exists no incentive for corporations to keep those human equivalents employed. Humans necessitate sticky conversations about benefits, work hours, protections, and accountability that AI frankly does not. What happens to those jobs?

AI proponents often compare the advent of the tech to the Industrial Revolution — many jobs were lost as industries were completely refactored and destroyed by these technologies, like we expect AI to, they say, but as you can clearly see, humans are still here, working now in different or differently-shaped industries. Those jobs are considered unfortunate collateral in a necessary technological reshuffling of the labor market.

But the buried lede of the Industrial Revolution comparison is that the revolution didn’t simply exchange some jobs for others. It developed a new capitalist system of labor de-skilling. De-skilling extracts the knowledge of design, craft, and mastery from high-skill workers and replaces those craftsperson type roles with new types of less-skilled and more-specialized roles. Those increasingly specific roles, in increasingly complex systems, meant that the average level of wide conceptual understanding of each worker was reduced, resulting in concentrated power of the managerial class.

As the average skill level of a worker goes down, the easier it becomes for capitalists to control a labor force. If a worker can be easily replaced, there is no justification to pay that worker high wages, or ultimately to keep that worker employed.

In this future where LLMs really are the be-all end-all many techno-optimists purport them to be, they will expand to wider and wider industries. Benefitting from narrowed and more specific job requirements, they will capture more roles, automate and de-skill more industries, and result in a widespread lowering of wages, job security, and employment.

And while those tech leaders on the optimistic far end of the axis pretend to believe this will open doors for some utopia where mankind does not work, it is plain to see the dishonesty of this when looking at how those tech leaders treat humans of any class lower than theirs, be it in their brutal mass layoffs of white collar workers, or the criminalization of homeless folks.

AI doesn’t work

The second scenario is that AI is not nearly as effective or useful in replacing labor as it is lauded to be. In this scenario, the explanation for these mass layoffs by Amazon, Atlassian, Autodesk, Block, Cisco, Coinbase, Dell, Dropbox, Fiverr, Goldman Sachs, Google, Grammarly, HP, IBM, Intel, Intuit, Klarna, Meta, Microsoft, Nike, Oracle, PayPal, Pinterest, Salesforce, TikTok, UPS, Unity, YouTube, et al, is that AI is merely the foil by which companies can cover up all manner of business failings or strategies.

The problem now is that, because there are scant worker protections (especially in the US), the onus to prove any claims about AI efficiencies does not exist. Businesses are enabled to use the magic spectre of AI to justify an any% RIF and spin great successes out of their operational failures.

However, while this is likely at least a contributor to the layoffs sweeping the technology and finance industries, it does not account for the massive scale of the cuts we’re seeing in this industry. Instead, these cuts must be lensed by class. Whether they mean to or not, owners of capital have exceptional class solidarity, because they have the shared goal of reducing the cost of labor. This, then, explains mass layoffs attributed to a technology like LLMs, regardless of actual LLM “productivity gains”: If the technology causing all the layoffs does not actually replace workers, then the workers will need to be rehired. Herein likes the rub — if large portions of the workforce have just spent 6 months or a year searching for jobs in an extremely employer-favored labor market, workers will take what work is available, and as a result wages, role quality, and benefits will nosedive.

The material outcomes

Both of these scenarios result in the same outcome: reduced worker solidarity, reduced worker skill, and reduced worker wages. AI/LLMs are the perfect tool for the capitalist because it doesn’t matter if it works; it threatens the workforce either way.

There exists precedent for this type of threat in the practice of “offshoring”, where companies move work to markets with cheaper labor and make layoffs in more expensive markets. Whether or not this move is useful can obviously vary widely as there are many variables to offshoring, but that companies consider it a tool, that it can happen at any time to large groups of workers is its functional impact.

Furthermore, because it weakens the local labor market and the bargaining power of labor, offshoring, like AI, creates an incentive for companies to position workers to be more easily replaced, through de-skilling and other strategies.

Eat the frame

The widespread deployment of AI/LLMs constitutes an existential threat to labor whether or not it is an adept move by the adoptive industries. Thus it is essential for workers to stop asking whether or not AI works and start asking who ultimately benefits from its deployment.

The answer to that question is capitalists, who have a persistent vested interest in weakening the labor market. Consequently it is imperative for workers to join forces in defense of their work, whether or not they are optimistic about LLMs as a technology. If a functional future of AI is realized, labor must be strong enough to withstand its seismic effects. If the brilliant AI future never dawns, labor must be strong enough to survive the ramifications of all of these companies making existential threats.

This means of course both bargaining for protections against successful widespread automation of the workforce, but protection against widespread automation as pretext.

Throughout history, workers have proven that when they stand together, they take control of their fate and the fate of their industries. You deserve a tech union.