2 ms·
> also don't consider using word processing software with grammar and spelling correction "cheating", even though they use fundamentally the same processes behi
by chii 4y ago
> also don't consider using word processing software with grammar and spelling correction "cheating", even though they use fundamentally the same processes behind the scene and differ only in the user interface, really.
i think it's a stretch to claim that spelling/grammar correction and large language models are similar or fundamentally the same.
that's like saying the horse-drawn carriage is fundamentally the same as a plane. Sure, they both move people around, but the differences are much larger than the similarities.
- jaggederest 4y agoIt's a little bit more like saying an F1 car is fundamentally the same as a Model T. Spellcheck was one of the first language models (albeit miniscule) in widespread use. You can draw a direct line from bloom filters for spellcheck in resource constrained environments to predictive text and thence on to LLMs.
- drw85 4y agoI think this is more semantics than useful discussion at this point. Spellcheckers for the msot part act on what you made. ChatGPT in this case is used to generate stuff, so you don't have to make it. That's a huge difference and the first analogy is more fitting than the Model T one. Spellcheckers are more like wheelbarrows, they don't move anything by themselves. ChatGPT is like an airplane with a broken autopilot.