The Ease of LLMs

My interest is always piqued when I view folks’ description of LLMs as making work “easier.” I don’t find that LLMs make work easier nor do they provide any meaningful supply of any human need. They make computerized work faster—they accelerate progress—and that is “better” for companies and give workers an approximation of ease because of the apparent lack of labour.

Unfortunately, you’re still accountable for the outputs and/or integration, which now means you’re a primarily a details reviewer, which I often find is a harder job to perform without having done the original labour (thinking, writing, doing) or at least possessing some artifice of experience. You’re reverse-engineering intent from output, which is closer to auditing than editing. And most people aren’t trained or very adept as auditors of work; let alone their own work. This type of work is actually quite...hard.

I also think LLMs make work harder because now you’re racing your ghost, sort of like a ghost car in a racing game. You’re now competing not just other labour, but your own labour. If you’re not self-replacing, we may have to replace you. Work is perpetually accelerated; you with it. You’re not getting left behind are you? Who is that lazy ghostly so-and-so who took 10 seconds when it could have been 9 seconds? Speed as proxy for competence.

Hyperfixating on optimizing your labour is also laborious, the sort of hypercritical and hypercyclical metawork that leads to burnout. It compounds because it feels productive, so we evaluate ourselves as getting more done, freeing us up to...? It also tends to defer the actual output indefinitely. There’s always one more tweak to the workflow before you’re “ready.”

LLMs might also make work harder because your labour is what grants you access to healthcare and other “social niceties” in certain jurisdictional hells. In much of the world personhood grants access. Work is more fraught under AI governance because now you’re always looking over your shoulder while also self-defeating/-competing.

I don’t hate technology—in fact, I love tech—it’s mostly that I have a hard time naming a single social/care need that AI helps humans with at the moment. It’s like “what if we inverted Maslow’s hierarchy” and stated everyone who didn’t agree that self-actualization was paramount was a boorish luddite.

As for Maslow’s, does AI reach the bottom layers (physiological and safety needs)? And where it does, is it in provisionment or displacement?

More importantly, what are you doing reading this? Get back to work!