Meta AI and Wikimedia Foundation Build an ML-Powered, Citation-Checking Bot (digitaltrends.com) 17
Digital Trends reports:
Working with the Wikimedia Foundation, Meta AI (that's the AI research and development research lab for the social media giant) has developed what it claims is the first machine learning model able to automatically scan hundreds of thousands of citations at once to check if they support the corresponding claims....
"I think we were driven by curiosity at the end of the day," Fabio Petroni, research tech lead manager for the FAIR (Fundamental AI Research) team of Meta AI, told Digital Trends. "We wanted to see what was the limit of this technology. We were absolutely not sure if [this AI] could do anything meaningful in this context. No one had ever tried to do something similar [before]."
Trained using a dataset consisting of 4 million Wikipedia citations, Meta's new tool is able to effectively analyze the information linked to a citation and then cross-reference it with the supporting evidence.... Just as impressive as the ability to spot fraudulent citations, however, is the tool's potential for suggesting better references. Deployed as a production model, this tool could helpfully suggest references that would best illustrate a certain point. While Petroni balks at it being likened to a factual spellcheck, flagging errors and suggesting improvements, that's an easy way to think about what it might do.
"I think we were driven by curiosity at the end of the day," Fabio Petroni, research tech lead manager for the FAIR (Fundamental AI Research) team of Meta AI, told Digital Trends. "We wanted to see what was the limit of this technology. We were absolutely not sure if [this AI] could do anything meaningful in this context. No one had ever tried to do something similar [before]."
Trained using a dataset consisting of 4 million Wikipedia citations, Meta's new tool is able to effectively analyze the information linked to a citation and then cross-reference it with the supporting evidence.... Just as impressive as the ability to spot fraudulent citations, however, is the tool's potential for suggesting better references. Deployed as a production model, this tool could helpfully suggest references that would best illustrate a certain point. While Petroni balks at it being likened to a factual spellcheck, flagging errors and suggesting improvements, that's an easy way to think about what it might do.
Citogenesis (Score:4, Insightful)
Seems like something like this might have difficulty dealing with Citogenesis (circular citing, media articles with facts originating from Wikipedia, then they get cited)
Re:Citogenesis (Score:5, Insightful)
It's a bit worse that that. A system like this will undoubtedly support something like "proof texting", an unethical practice by which you start with the conclusion you want and then find sources to support it. It's common among undergrads who don't know how to properly write a research paper ... and on Wikipedia.
Re: (Score:2, Insightful)
Did you realize that you are describing how the Republican Party operates these days? Of course their version of "facts" include ravings that at one point would have led to incarceration in a mental ward.
Re: (Score:2)
All fanatics operate like that. That is one of the core reasons why they cannot get anything right: They do not look at reality.
Re: (Score:2)
Checker or enforcer (Score:3)
Checking references is one of the major banes of writing an academic paper. The idea of an extra level of support that will mitigate the prospect of getting one wrong is attractive. However if the system enforces rules, banning certain sources and imposing certain interpretations on material, then we will have a problem...
Re:Checker or enforcer (Score:4, Funny)
However if the system enforces rules, banning certain sources and imposing certain interpretations on material, then we will have a problem...
How can you possibly imagine it would be otherwise?
Re: (Score:2)
Which is funny (Score:2)
DALL-E banned the word "shooting" (Score:2)
Re: (Score:2)
References with reproducible results? (Score:2)
You might argue that the reason a lot of current results especially in ML aren't reproducible, is because they are based on earlier results that also aren't reproducible. So a very useful additional feature would be to flag the quality of references.