Has AI Taken My Job?

Write me an opening paragraph introducing ChatGPT.

"ChatGPT is a natural language generation model developed by OpenAI. Using artificial intelligence technology, it can automatically produce natural, fluent text. ChatGPT is built on the GPT-3 (Generative Pre-trained Transformer 3) architecture, and can automatically generate text responses based on user input."

Now introduce OpenAI for me.

"OpenAI is an artificial intelligence research company founded in 2015 by a number of prominent tech figures, including Elon Musk and Sam Altman. In just a few short years, OpenAI has achieved numerous breakthroughs, including the GPT series of natural language generation models and the DALL-E image generation model."

A Chatbot That Writes — And Might Wipe Out Most Jobs

Okay — everything in this piece so far, aside from the prompts themselves, was written by AI. Total time: about a minute? Given that Wealth Magazine might just cancel my fee for this issue, I'd better stop right there. But this time, everyone really should be panicking — myself and every other columnist included.

GPT models have actually been around for years — the current GPT-3.5 is already the fourth generation. But it wasn't until OpenAI released the ChatGPT chatbot that people really grasped how close AI had gotten, and how broadly it could be applied. For the past several weeks, every major social platform in Taiwan has been flooded with users signing up for GPT, asking it questions, and posting screenshots of their astonished reactions.

OpenAI's goal is general-purpose AI, and its approach has been almost brute-force: pour a staggering amount of compute and data straight into the system and train it. OpenAI is closely allied with Microsoft, and much of its resources come courtesy of that relationship. Microsoft has already built OpenAI's models into its Designer software, so before long every corner of the Microsoft ecosystem should have AI you can genuinely feel — writing slides and articles with little more than a mouse, or seemingly by sheer force of will, is only a matter of time.

Okay, so what's the actual relationship between these AIs and our future?

In the immediate term, the biggest impact is that certain jobs simply lose their meaning. Spend enough time with AI tools and you'll notice that work we used to consider meaningful starts feeling a lot less meaningful. Take illustration and simple copywriting — two needs that come up constantly in advertising work. Once images and press-release-grade articles can be generated on demand, the entry-level, heavily-trained roles built around producing them simply evaporate.

So does that mean everyone loses their job immediately? Not necessarily. People once assumed computerization and ERP systems would shrink office headcount and paper usage. In reality, in the early stages of digitization, both office headcount and paper consumption actually went up. Everything was digitized, sure, but for a long time messages — official documents, for instance — still needed paper as the medium, even though the underlying draft now lived on a computer. Headcount grew too, because more people were needed to route messages up and down the chain of approval. It wasn't until electronic signatures took hold and people got used to receiving information purely online that the numbers finally started shrinking. So in the early going, AI will indeed reduce headcount across various industries, and it will eliminate certain entry-level jobs that require heavy training.

Jobs Where AI Becomes Your Coworker Might Do Better, Short Term

But overall, one thing gets added while another gets subtracted. New roles emerge that require industry expertise specifically to communicate with AI — for a while, we'll still need creative directors and other people with genuine industry taste (or senior lawyers, senior accountants, and so on). But at some later stage, both sides of that equation start shrinking: AI will only need supervisors, and those supervisors won't even need deep domain expertise anymore — or they'll need a different kind of expertise, one specifically for supervising AI, not the kind traditional industry training produces. It's a bit like a CNC lathe engineer, who doesn't need to know how to hand-carve metal at all.

For now, people who already have deep experience and don't mind working alongside machines won't be badly affected — in the short run, they might actually do better than before. But the training pipeline for new hires is going to take a serious hit. If a workplace only needs veterans and has no use for rookies, where exactly does the next generation of veterans come from?

On the flip side, what machines genuinely can't replace should become clear within a few years — and there's at least a fairly long buffer period for it. Blue-collar work is going to be hard for AI to replace in the near term: delivery robots are still at a very early stage, and getting an electrician or plumber to your house for a renovation is still an absolute necessity. But conventionally defined white-collar work is about to get squeezed hard.

Until now, the strategy of the global giants was to move slowly, periodically releasing a demo just to remind the world how formidable — or how flush with cash — they are, like Meta's earlier text-free translation system. But spurred on by OpenAI, those giants may now start emptying out their warehouses of unreleased technology one item at a time. So over the next three years, plenty of industries are going to face disruptive shifts. The ones who'll lose their relevance immediately are the mid-sized, already-funded AI startups — particularly those working on language semantics or robotics — which now face the very real risk of having their technology leapfrogged entirely.

In the AI era, meaningful work and meaningless work will start to separate far more clearly than before. For those of us who are a bit old-fashioned, the painful part is watching things that used to be meaningful and "priceless" slowly become genuinely worthless. But technology, once it's happened, has happened — and we're in for a long stretch of adjustment ahead.