undefined | Better HN

0 pointsgavmor1mo ago0 comments

Take a look at https://chatjimmy.ai/ -- it's running against Taalas' "hardcore" silicon model, ie a dedicated, ASIC-like chip.

0 comments

3 comments · 2 top-level

bikelang1mo ago· 1 in thread

Wow - actually pretty astonishing how fast their inference is. So fast it feels fake?

qingcharles1mo ago

Yeah, when you find fast inference like that it almost feels like the answer arrives before you hit return. Now imagine it running locally with no server round-trip.

jiggawatts1mo ago

Sure it's fast, but it's at ChatGPT 2.0 levels of intelligence.

j / k navigate · click thread line to collapse