Hacker News Reader

Comment by ilaksh | Hacker News Reader

Robotics will rapidly pick up improved capabilities. With so many people out of work and so much automation being consolidated, what do you think all the entrepreneurs and researchers will look to next? It will be breakthroughs and efficiency improvements in robotics .

I think this "learn a trade" plan has maybe 5 years max. By then the trades will also be automated.

View on HN

I'm not a gambling man but if I was I would put a very large amount of my net worth on a bet that no trades will be automated in 5 years time. Not a single one.

I think trades will be around for a century. The problem is dealing with crazy existing stuff, like houses that are not to code and have been repaired by the last 3 owners, all of whom made it up as they went. Good trades people are able to figure out what the last idiot was thinking, know how it is supposed to work, and can make it work (without callbacks) for a long time with minimal intervention.

I remember a college professor who cashed out his retirement in the mid-1990's to start his own online college. He went broke in less than a decade, and we still have not removed professors from the colleges.

As a teacher, I do think that ChatGPT will replace most textbooks and teachers, but I don't know how long that will take.

Progress is slower in the world of atoms than in the world of information. I watched the progress from the first DARPA Grand Challenge for self-driving cars (2004, no car made it to the finish line) to the Urban Challenge in 2007 (many cars succeeded in a much more complicated environment) and thought that self driving vehicles would soon be commercialized. But there are 10,000 "last 1%" cases left to handle and we still don't have good enough systems to replace truck drivers or cab drivers today. Replacing jobs like plumber or electrician looks even harder than self driving vehicles because it requires far more intricate physical manipulation and perception than driving on public roads.

It's fine to automate 99% of it if the 1% can be somewhat efficiently taken over by humans. LLMs suffer the same issues.

Reply on news.ycombinator.com