[IshKebab]

I agree. LLMs have a ton of unrealised applications in business. Imagine training one on your company wiki and chat history.

Barely any companies have done that yet because of legal and security concerns and because it isn't easy to do yet, but that will change.

It's not going to be long before someone makes and end-to-end speech to speech model. A single model that incorporates speech recognition, LLM and speech synthesis. In fact I'm really surprised it hasn't happened already because it's such an obvious thing to try. That's going to blow people's minds.