rajatrocks
Joined 18 years ago
Joined 18 years ago
It's giving people a tool to get more information, hopefully reputable, on a "fact".
The tool I built enables people to check a statement against the knowledge of Google Search and then decide for themselves.
Is Google Search 100% accurate? No, it's based on information from people, and people lie, manipulate, fudge and deceive. Information changes, facts change.
But I believe that it's a better alternative than blindly accepting or rejecting everything that you read.
disinformation = intentionally misstating the facts in an effort to mislead ex: "January 6th was a day of peace and love"
fact-check : to verify the factual accuracy of
"Our fact-check work is supported in part by a grant from Meta."
1) 100% agreed. I would never release a tool like this that just used an LLM. But this tool uses Grounding with Google Search. So in this case Gemini uses Google Search as a tool to try to verify whether the statement is true or not. If it cannot ground its response, it returns a message saying that it can't fact check the statement - it will never return an LLM's "opinion". If it can ground it, it returns an answer with citations and links to Google searches.
2) That may be true. But even I have found it useful as I'm reading comments on YouTube posts and in Facebook groups to verify whether what people are saying is true or not. Even people I agree with spread (intentionally or not) mis- and disinformation since it's so easy to do.
3) Per the first point, since it's grounded in Google Search it is giving very real-time responses. Try it with the headlines on any news site.
Previously grounding was $35 per 1,000 queries.
I first tried out something like this in the early 1990s at VPL Research - one of the originators of the VR industry. They had a visual programming language called Body Electric for controlling how things behaved in VR. Some details here: https://news.ycombinator.com/item?id=22788773
And more recently, Google showed PromptChainer for wiring together LLM calls and Javascript: https://www.cs.cmu.edu/~sherryw/assets/pubs/2022-promptchain...