3 ms·
> almost all of what us old-timers called NLP is just crushed these days For this to be true for most production service use cases, LLMs would need to be at le
by bugglebeetle 3y ago
> almost all of what us old-timers called NLP is just crushed these days
For this to be true for most production service use cases, LLMs would need to be at least ~10X faster. I generally agree they can be quite good at these tasks, but the performance is not there to do them on large datasets.