4 ms·
LLMs still completely won't admit that they're wrong, don't have enough information or that the information could have changed - Asking anything about Svelte 5
by flessner 2y ago
LLMs still completely won't admit that they're wrong, don't have enough information or that the information could have changed - Asking anything about Svelte 5 is an incredible experience currently.
At the end of the day it's a tool currently, with surface-level information it's incredibly helpful in my opinion - Getting an overview of a subject or even coding smaller functions.
What's interesting in my opinion is "agents" though... not in the current "let's slap an LLM into some workflow", but as a concept that is at least an order of magnitude away from what is possible today.
- gom_jabbar 2y agoWorking with Svelte 5 and LLMs is a real nightmare. AI agents are really interesting. Fundamentally they may represent a step toward the autonomization of capital, potentially disrupting "traditional legal definitions of personhood, agency, and property" [0] and leading to the need to recognize "capital self-ownership" [1]. [0] https://retrochronic.com/#teleoplexy-17 https://retrochronic.com/#teleoplexy-17 [1] https://retrochronic.com/#piketty https://retrochronic.com/#piketty
- baobabKoodaa 2y agoIt's fairly easy to prompt an LLM in a way where they're encouraged to say they don't know. Doesn't work 100% but cuts down the hallucinations A LOT. Alternatively, follow up with "please double check..."