the great theoretical chatbot job apocalypse

According to Anthropic, LLMs can obliterate most white collar jobs. Well, theoretically...

If the new Anthropic jobs report showing that the vast majority of white collar jobs are done for thanks to AI was a vodka, it would be called Absolut Bullshit. This is not cope or some wishful thinking. It’s that the infographic they shared to go with it makes this obvious because the vast majority of the obnoxious radar graph represents not what AI does today, right now, but its theoretical potential according to the company.

It’s like visiting a chiropractor for a sore back, mentioning that you have some trouble sleeping and have poor circulation in your legs, and what do you know, if you just give them your credit card and let them yank your head and limbs around while looking a tad to severely constipated, they’ll fix that too with 19th century spine magic.

For example, Anthropic claims that AI can do 94% of all tasks currently carried out by flesh and blood humans in business and finance. Sounds dire, right? Well, in the very same breath, they say that 66% of task coverage is purely theoretical. In other words, LLMs automate less than a third of all tasks today — ignoring questions of errors and overall quality, just a binary of whether it can or can’t do the task at all — and maybe, they will be able cover another two thirds because trust me bro, they could.

I mean, theoretically, one day I could be a disembodied intelligence flying through the stars using a warp drive, Bobiverse style. That is theoretically possible in a sense that you could argue it can mathematically work if everything goes exactly right and make some very bold assumptions about consciousness, energy, and computing.

What are the tasks that can be automated exactly? Whatever a couple of guys who worked for OpenAI in 2023 said they were based on their best guess as to what the people they were trying to replace actually do. In other words, LLMs could basically kick you out of a job because some tech bros said it’s theoretically possible because other tech bros said so three years ago, which is an eternity in AI land.

There’s also the bizarre assumption that given the relentless pace and pressures of job automation, the tasks they’re currently claiming can be done by an LLM did not rapidly change in response over that three year time frame, despite everyone being forced to use these tools, or else.

So, the bottom line here is that if it just took some grit and training to eliminate a few million office jobs a quarter thanks to chatbots, as per the report’s claim, companies would’ve 100% done it by now instead of 95% of pilot projects ending in failure as no productivity gains or additional revenues or profits materialized.

Unless virtually every company is being ran by absolute goobers who can barely use a computer — which I most certainly do not think is the case, but our AI bros apparently seem to — they would have absolutely figured all this out and we’d be in the middle of Great Depression 2.0 with a 30% unemployment rate and no jobs to be had as we all train to become plumbers and electricians, fix our own toilets and lights, then sit in a state of perma-panic, trying to figure out what’s next.

But there is an important conversation to be had here if we moved past the toxic PR of LLMs and started asking some profound questions. Because it is true that jobs as we know them today will be on their way faster than we can replace them at some point, and others exist only as a means to prop up the consumer economy, and could vanish overnight without any real impact to everyday life. Well, other than to the people who were laid off and now find themselves in job limbo.

At the same time, while acknowledging this fact, and that there are limits to how much you can grow the consumer economy, especially in the predatory manner in which it’s being handled now, we have no Plan B other than fantastic promises of free money by AI systems from people who are sealing themselves in doomsday bunkers just so they don’t have to give up a tiny percentage of their wealth for a fairly short amount of time to avoid a societal collapse.

Our culture still believes that our existence in general is a waste unless we either toil in a factory style job — that is for an average of eight hours a day, five days a week over forty or fifty years, or until we drop dead — or just have enough money to buy our way out of this arrangement, or never really enter it in the first place.

AI, for all its flaws and limitations, however, is really good at pointing out inefficiencies and major misallocations of human capital. Sure, the Anthropic report may be yet one more fear-based marketing effort, and yes, it may be years before an entire job sector is fully automated. Still, the fact that it’s even theoretically possible and we will run out of actual work to do, what’s next?

Flooding workers with the equivalents of corporate busy boxes for adults so we could prop up a limping economic model that’s venturing into fiscal cannibalism? We could argue that we already more or less do that more often than we’d like to admit. Oh, and it should be noted that people absolutely hate that because it makes them angry, and bored, and depressed, and fills them with existential dread and frustration.

What these job-pocalypse-is-coming reports are doing are tapping into the anxieties of workers on the verge of a mental breakdown, stirring up all of these unhealthy and anxious feelings, and using them to effectively threaten people into using LLMs, ready to jump on the latest bandwagon at the drop of a hat because their existence is being threatened.

This may be why today, AI is about as popular with the general public as new wars in the Middle East, kicking puppies, and Nazi salutes. After all, just because one is being forced to stay in abusive relationship doesn’t mean they may grow to like it. Just keep in mind that Stockholm Syndrome was never a real diagnosis, but one psychologist’s attempt to excuse an overzealous, reckless police force. The overwhelming majority of prisoners always resent their captors, be they robbers, pirates, or corporations.