I, like so many others have been very impressed by ChatGPT. I'm a software engineer who has some understanding of deep learning, but my day to day work is far outside of the AI field. So I'm not up to date on the current state of chat bots, speech synthesis, etc. I'm wondering how big of a step forward or breakthrough ChatGPT is compared to the state of the art from a year or a few months ago?
It's a breakthrough maybe the way iPhone was a breakthrough (but to a far lesser degree IMO). It's not doing anything previously inconceivable, it's just doing it in a more polished way with some more hype. Of course that can make all the difference. From a research perspective though it doesn't mean much
I think it's about as big a breakthrough as that "wormhole" they made with a quantum computer. That is, 99% hype.<p>It somehow gets more empathy than many humans but the answers are frequently convincing bullshit. If it replaces the likes of Donald Trump, L. Ron Hubbard and Sam Bankman-Fried it may be a very good thing. It might be able to run Twitter better than Elon Musk.
Remember that ChatGPT is only as good as its data, and the data it has is the "wisdom of crowds" sourced from public sentiment on the internet. If you ask it something? The answer you get will always be the "common-sense" answer that a room full of people would also come up with. It lacks outside-the-box thinking.