Hacker News Reader

Comment by ResearchCode | Hacker News Reader

I thought it would be the opposite. People heavy on the soft skills and light on tech are more easily replaceable by a large language model. They seem the most promising for slide decks or fill in Excel sheets, or generate a meeting protocol for a pointless meeting.

View on HN

Business domain knowledge is not used for creating slide decks and having pointless meetings. It is used to decide what to build, and to some degree how to build it.

LLMs are also good at reasoning. It would be good exercise to see if they can decide what to build.

They are good at generating plausible sounding tokens one after another. Often those will be good ideas. But this isn't the same as reasoning.

Depends on the "soft skills". ChatGPT doesn't look good at prioritization. It can't tell you which of two numbers is bigger, so that's not a great sign. Of course you can offload that to a calculator, but then you lose the actual LLM. So prioritization seems good for a while.

If your soft skill is turning requests into tickets and looking at PRs to figure out which tickets are already closed, yeah things ain't looking good.

I asked GPT-4 to choose the greater number from a pair of random numbers 0 to 1,000,000. It got 94/100.

I think the market will get squeezed from both ends: Both the leetcoders and the those only producing superficial stuff can be replaced. There is a lot in between that will remain useful for some time.

Reply on news.ycombinator.com