[senttoschool]

>What might be holding them back?

Probably because LLMs hallucinate and often answer with the wrong information. Can you imagine Siri spitting out wrong information, causing some accident or health risk? Apple's pristine PR image would never allow something like that. Quite frankly, I'm surprised that Google and Microsoft are willing to take that risk.

I totally get why OpenAI said "f it, let's just ignore hallucination and accuracy" that only affects 5% of queries. OpenAI does not have an established worldwide brand.

But could you imagine the news stories if Apple released ChatGPT?

[0asa]

I can imagine conversing with a voice assistant if it could process complex language queries, rather than the current restricted tasks (e.g., setting alarms, playing music, checking the weather, turning on lights).

Research in this area is flourishing, so it is certain that LLMs will improve. The popularity of LLMs is partly due to the fact that it has become natural to interact with an "intelligent" computer program. ChatGPT makes this natural, and combining a smart speaker with LLMs would make it even more so.

Enabling a "closed" conversation mode initially, similar to what the free version of ChatGPT does, would be harmless unless people take language model-generated words as absolute truth. This mode could be used to learn new things while cooking and save long conversations to a notes app for later reference. Wouldn't this be an attractive feature for buyers?