When every tech company laid off a ton of workers and claimed AI, they were committing fraud to their investors, nothing more. They needed to lay these people off due to overhiring, these employees underperforming and needing to be fired (but they didn't want to risk a lawsuit), or economic headwinds and thought it made them look like better leaders. Do you really think they wanted to do last year's business with 15% less cost? If they had actual useful AI, they'd repurpose all of these engineers, if not hire more so they can 10x their output and crush their competitors.
We've seen job loss to automation before. It typically is pretty slow. It typically results in reduced hiring before mass firing. Most people get repurposed. In most cases, the majority of people are retained and the company focuses on increasing output or expanding offerings. For example, if you're DeWalt tools, you use automation to make power tools with less humans and start expanding to hand tools and ladders and tool storage, like they did over the last 15 years. Your net hiring usually expands. As the automation gets deployed, you typically retain a huge portion of the staff during the transition, just in case. It's a process that unfolds over many months or years, not weeks, like Salesforce.com, Google, Facebook, etc did. And to repeat...if these were useful employees, they'd repurpose them to expand offerings.
I don't know if these AI CEOs were ignorant and believed their bullshit or were lying. However, we've all used these AIs. They're not dogshit, but they're not magic either. They're definitely not ready to actually replace human beings. I use Claude daily and it produces huge errors every day. I use Gemini frequently...with very simple product questions "Which $x has features $y?" It fails the majority of the time. For example, I asked it "What's the cheapest 10' USB4 cable with USB-IF certification?" It returned results that were 6'6" and cables more expensive than the ones in my Amazon cart...yet it provided links to Amazon. Pretty much every product search I've ever made failed. My point being...I can't imagine and easier question for AI. I gave a length, I gave a specific token and asked for the cheapest...it failed on all 3 fronts.
This shit ain't ready to replace anyone. If you're paying a human being to do it today, you need more accuracy than today's models can provide. If you can tolerate their level of errors, you weren't paying anyone before. Sure...it can replace your old automated chat bots...maybe reduce your offshore outsourcing spend slightly...but no...it's not that they're humanitarians. It's that their technology doesn't work well enough to be evil.
AI is a lot more impressive and fearful in our minds than in reality.