PR by Xuan-Son Nguyen for `llama.cpp`: > This PR provides a big jump in speed for WASM by leveraging SIMD instructions for `qX_K_q8_K` and `qX_0_q8_0` dot product functions. > > …
I’ve had similar conversations discussing some of your points in this thread (albeit at a much much higher level because I’m basically an analyst) with C-suites at my company, and what truly terrifies me is that they do not give a fuck as long as they get a passable finished product. And the bar for passable is now in hell.
My company doesn’t even have any software engineers because my manager convinced his boss (a c-suite) that we could do all our coding with ChatGPT scripts. Surprise surprise, now we have a ton of busted code in our GitHub because apparently copy and pasting from the LLM of the day is what passes.
I’ve had similar conversations discussing some of your points in this thread (albeit at a much much higher level because I’m basically an analyst) with C-suites at my company, and what truly terrifies me is that they do not give a fuck as long as they get a passable finished product. And the bar for passable is now in hell.
My company doesn’t even have any software engineers because my manager convinced his boss (a c-suite) that we could do all our coding with ChatGPT scripts. Surprise surprise, now we have a ton of busted code in our GitHub because apparently copy and pasting from the LLM of the day is what passes.