[khaki54]

I know it's unrealistic, but I wish this were a local appliance I could buy. Maybe plug into Home Assistant too.

Sending all my data off to two external services might be a bridge too far for me... Then again, I don't trust Microsoft or Google either.

[theolivenbaum]

We're actually exploring that with our app, if you've a GPU large enough you can run a llama2 based model locally at almost chatgpt speeds... If you're interested in trying drop me a message at rafael (at) curiosity.ai

[khaki54]

Thank you!

[_boffin_]

How much would you pay for an at home appliance? His responsive would it need to be?

[khaki54]

I would pay between 5 and 10k USD depending on how close it was to Iron Man's Jarvis. I'd even pay an annual subscription for updates.

Would need to be about as fast as ChatGPT seems to be for the conversation piece and perhaps utilize some kind of optimistic response while it does things asynchronously

[jychang]

1. Not enough to make it profitable in 2023

2. Consumers want 0 latency. Every millisecond counts. Ask google about how they optimize search result latency