←back to thread

627 points cratermoon | 1 comments | | HN request time: 0.214s | source
1. rob_c ◴[] No.44462789[source]
Learning to use LLM effectively is like learning how to use scripting in office. Useless to a lot of people. Dangerous if you get it wrong. Full of many bad things from the internet. But still the most powerful thing to happen to computing in a generation. And probably soon about to be turned off by default for security reasons to follow the analogy.

Just because you failed to use an LLM effectively the first time, or it doesn't live up to your version of the hype doesn't mean you have a magic window into how sh1t they are. Many people who use them regularly know just how bad they are and what it's like to move beyond the original context window length. But having auto complete on steroids is still the best thing with regards to making computing "fun" again in a generation. No more boiler plate, no more beating your head against a desk looking for that minor coding bug that could have been fixed by a awesome regex if only you had the time to learn it. No more having to break your stride and go through about 20 websites with popups to find that someone else solved the problem got you, just getting on with stuff and having fun whilst doing it.

Edit: no retort, just flagging and down-voting... lovely