Much as I think GPT-5 is the bees' knees, this seems to both overestimate AI's performance (error rate is now really low, but not low enough to run without supervision) and underestimate CEO stupidity and greed (of course they'd fire workers en masse if they thought they could actually replace them).

It’s the ticking time bomb in the global economy, and every CEO knows it: AI is already powerful enough to replace millions of jobs. So why haven’t the mass layoffs begun? The answer has little to do with technology and everything to do with fear. Corporate leaders are quietly waiting to see who will be the first to pull the trigger.

Take Palantir’s CEO, Alex Karp. During an interview with CNBC in August, he said: “We’re planning to grow our revenue … while decreasing our number of people.” Karp then continued: “This is a crazy, efficient revolution. The goal is to get 10x revenue and have 3,600 people. We have now 4,100.”

The subtext is clear: Palantir already considers 500 of its employees to be a surplus that AI could replace. It could increase its revenue by 10x while reducing its workforce by almost 12.2%.

I think it's worth quoting one of the comments as well :

If by replace, you mean 'very poorly substituted' then sure, AI could 'replace' lots of people.

Exactly. Some of what LLMs can do is bloody impressive and anyone who says otherwise is lying or not paying attention. But while I think GPT-5 has crossed a threshold from being "sort of useful as a supplement" to "actually genuinely useful for all kinds of stuff on a regular basis", the proviso to that is that it's still useful only as an assistant. It's not ready to replace anything very much.

#AI

https://gizmodo.com/the-real-reason-you-havent-been-replaced-by-ai-yet-2000641235