Live data from Hacker News

AI Has a Discovery Problem

mhacevedo.com

31–39 of 39 posts

Re: AI Has a Discovery Problem

#33
post #12

If I know the code behind the button, I don't have to press it to know what it will do. If I know how LLM work, and what it is trained on, I don't have to "explore" it to understand what it can do for me. The only variable is how deeply one knows how "LLMs" work, and that makes up all the on going arguments by lay people...

That's like saying if you truly know how a hammer works, carpentry can hold no surprise for you.

Re: AI Has a Discovery Problem

#35
"The discovery problem" is the biggest bottleneck to all human behaviour- travel, business development, academic rediscovery, literature, etc etc. Saying "my natural language input device is too intimidating to use" would be considered borderline insane 3 years ago.

Re: AI Has a Discovery Problem

#38
We're all ants. You can't see the whole sky by design. That's the "guard rails" they talk about.

The non-technical marketing person can't use it because they don't understand programming tools. AI is by programmers for programmers. AI agents might make perfect sense to someone who understands a posix pipe. Marketers don't.

Re: AI Has a Discovery Problem

#39

I was trying to teach a co-worker some agentic stuff today, and this is totally true. Especially if they are using gimpy tools like Github Copilot. The harness matters. None of my super-powered CLI + skills stuff worked like how it does in a more powerful harness like Codex or Claude Code. Try telling someone to imagine it could do something amazing, if they switch to a completely different tool.

[flagged]
Post reply on HN