3 ms·
I've seen this. I can get quite a bit of decent responses from LLMs about Python. But using it for Elixir produces more shady/dated answers. And Jetpack Compose
by travisgriggs 3y ago
I've seen this. I can get quite a bit of decent responses from LLMs about Python. But using it for Elixir produces more shady/dated answers. And Jetpack Compose, it just completely gets weird (and mostly worthless, being either too synthetically approximated or dated).
It really illustrates (to me) that LLMs are "Plausibility Machines". They produce plausible answers. Plausibility is informed by repeated observability over large sets. We asses something as plausible when it matches the majority of our observations. That's how the parlor trick of seeming to "tell the truth" is pulled off. The venn overlap of "plausible" and "true" is pretty high. But when "truth" is in flux, you start to see the feet of the wizard behind the curtain.
- School-Cotton 3y agoWhich LLMs have you tried?
- sho 3y agoThat's weird. Which ones are you trying? I get great elixir help out of CGPT4. Sure, it needs close supervision - you can't just trust it and you need to actually understand everything it suggests, and it loves to "forget" things you told it just a moment ago about how you can't use a guard clause there, or whatever - but in terms of just getting a decent starting point, man, it's been a game-changer. The $20 I send OpenAI each month is the most no-brainer money I spend. It was definitely pretty bad with liveview for a while, while that was under heavy development and API changes, but it's had a data update that mostly fixes that, and if you specify the version it's helpful there too.
- AlchemistCamp 3y agoI’ve found that ChatGTP’s usefulness for Elixir improved dramatically after the updates around dev day. I’d say it’s still uniquely good with Python, but I find it roughly as useful with Elixir as it is with some other very popular languages.