Looking to the future
AA: What could LLMs like ChatGPT mean for the future of AI and data, and in particular, how do they affect the data science scene – including the field of NLP?
NVO: ChatGPT is taking the internet and the AI community by storm but, like all previous announcements made with a lot of marketing hype, one should be a little wary of some of the grandiloquent claims. Yes, LLMs and the different versions of ChatGPT (and their siblings) have shown positive and somewhat surprising results, but the reality is much more nuanced than what the official communiqués and the tide of comments would have us believe.
ChatGPT generates answers that are riddled with nonsense. What some generously call “hallucinations” are – from my perspective – inherent flaws, even though some can and have been fixed.
This is the same “good old” debate that probably can’t be settled unless we can show that this approach would lead...