Will LLM help to increase accuracy on requests?

[spring crystal]

Hi there! Does anyone know if using an LLM (as OpenAI) will increase the rate of success when giving a command?
For context, English is my second language and I have a very strong accent. So saying “Turn lounge lights on” has a very (VERY) low succes rate. Using Assist instead of Preview, I realised that HA is understanding me asking to “Turn launch lights on” (i swear, I added “launch” as alias to be able to use it, dont laugh, but yeah, lol).
Obviously I know LLM wont fix my accent, but I wonder if using some sort of context, it would interpret correctly…

Cheers!

[night vale]

LLM might be able to interpret what you meant, but may not always be reliable and depends on how smart/powerful that LLM is

it also introduces the possibility that the LLM could hallucinate or go do what it thinks is right in an attempt to satisfy the request when it can't 100% match what was asked, so would require some testing to see how well that works. 🙂

[blazing veldt]

Accuracy, no. But as Nick says, sometimes it manages to figure out what you want, even if there's no matching phrase. Sometimes, the output produced by the LLM can also be plain funny.