Jump in the discussion.

No email address required.

As much accuracy as I would expect from Bard :marseyxd:

Jump in the discussion.

No email address required.

I've seen a funny article from theverge comparing the gpt4, bing and bard yesterday and the conclusion cracked me up a little

https://www.theverge.com/2023/3/24/23653377/ai-chatbots-comparison-bard-bing-chatgpt-gpt-4

If you’re looking to accomplish verbal tasks, whether creative writing or inductive reasoning, then try ChatGPT (and in particular, but not necessarily, GPT-4). If you’re looking for a chatbot to use as an interface with the web, to find sources and answer questions you might otherwise have turned to Google for, then head over to Bing. And if you are shorting Google’s stock and want to reassure yourself you’ve made the right choice, try Bard. :marseyxd:

Jump in the discussion.

No email address required.

I finally got chatgpt yesterday and thought of trying to get it to help in my research.

Fricker was just making up papers, and when I told it the papers don't actually exist, it apologized and repeated the same paper titles along with made up author names.

Completely useless.

Jump in the discussion.

No email address required.

Interesting on this issue in particular: https://twitter.com/dsmerdon/status/1618816703923912704

tl;dr it's a language model and doesn't understand the request much further than "ok let's reply with something that would naturally follow such a request in a conversation" with the result being things that sound like they could be papers.

Jump in the discussion.

No email address required.

To find sources and answer questions you might otherwise have turned to Google for, head over to Bing.

Jump in the discussion.

No email address required.

I get great results but that's because I know how to wrangle it. I've been wrangling :marseycowboy: these LLMs since AI Dungeon's first release in December 2019 :boomer:

You have to talk to them in a specific way and avoid polluting the token history with shit that's incorrect. If you see something factually wrong and it mentions it more than once, just make a new chat and try again. In superior interfaces you could just edit the response but ChatGPT was made by r-slurs that don't know what UX even stands for

Jump in the discussion.

No email address required.

Link copied to clipboard
Action successful!
Error, please refresh the page and try again.