I tengah cuba ikat binatang LLM.
Hari ni, I tengah sibuk study "agent harnesses".
Kalau you biasa pakai ChatGPT, bayangkan:
- LLMs (large language models) ni macam 'otak' pandai.
- But LLMs are 'non-deterministic'. Susah nak predict. So you bagi dia freedom? Memang merewang.
- So kita kena 'ikat' dia. Bagi structure, tools, loops, goals, constraints. (Macam pasang leash jugak la.)
Anyway, I'm dreaming big: nak tambat T-Rex.
...Tapi for now, maybe I cuba Triceratops dulu.
Kalau you biasa pakai ChatGPT, bayangkan:
- LLMs (large language models) ni macam 'otak' pandai.
- But LLMs are 'non-deterministic'. Susah nak predict. So you bagi dia freedom? Memang merewang.
- So kita kena 'ikat' dia. Bagi structure, tools, loops, goals, constraints. (Macam pasang leash jugak la.)
Anyway, I'm dreaming big: nak tambat T-Rex.
...Tapi for now, maybe I cuba Triceratops dulu.