Hacker News Reader

Comment by falcor84 | Hacker News Reader

>One key finding, for example, is that only about 23% of the wages paid to humans right now for jobs that could potentially be done by AI tools would be cost-effective for employers to replace with machines right now.

>While this could change over time, the overall findings suggest that job disruption from AI will likely unfurl at a gradual pace.

I'm not quite following their logic here. Isn't 23% absolutely massive? And seeing that AI capabilities are maturing exponentially, this is a tectonic shift in the economy, even if it's technically "gradual" in the sense that not everyone will be immediately replaced.

View on HN

Yeah I agree, it's like saying "oh hey don't worry only a quarter of you guys will lose your jobs... but hey! Not gonna happen tomorrow!". Seems to completely ignore the impact that will have on people reliant on this income.

> are maturing exponentially

What does this even mean?

I was mainly referring to how compute costs keep decreasing exponentially year over year, even though the strict version of Moore's Law has slowed down. TFA's analysis is based on the cost of replacing an employee with a bot, so even if AI software stopped advancing, it'll continue getting exponentially cheaper to replace a typical employee. So at the most optimistic scenario, the current employment rate will persist only if everyone is really good at The Red Queen's Race.

This is simply wrong, because it is not a matter of cost. ChatGPT can be available for free and it would still not replace most jobs in its current state. Likewise the current state of vision-based self-driving is not good enough to replace even a single driver, even if the cars would be free. So certainly either software advances are needed OR increasing the training data or parameters will further improve the models (which is not necessarily the case).

I'm not following. If tech ChatGPT4 becomes available literally for free, or at least such that it can be re-run thousands of times for each task with some basic glue code and/or human intervention in between, then I see it replacing a huge chunk of knowledge work. There have already been quite a few impressive examples of running a whole team of LLM agents with different system prompts, communicating with each other and the external world, but price is still a limitation.

And as for self-driving, it is absolutely already replacing quite a few drivers, even if you just consider Waymo. I for one would ditch my car tomorrow if their service was available in my city for a reasonable price.

Reply on news.ycombinator.com