OLDaily

Study shows AI-generated fake reports fool experts
Priyanka Ranade, Anupam Joshi, Tim Finin, The Conversation, 2021/06/11

In the past I have talked about the possibility of automatically generated OER. These learning resources would be created by transformers, like BERT from Google and GPT from OpenAI, which (as the article says) "use natural language processing to understand text and produce translations, summaries and interpretations." That's great, but the technology also introduces the potential for the use of OER to spread misinformation. This article suggests that AI-generated misinformation may be good enough to fool experts. It offers some examples where fake reports were used to fool security experts about potential intrusions and threats. Misleading OER ('MOER'?) could create havoc, especially if they pass undetected in peer review. Something to think about.

Web: [Direct Link] [This Post]

Google and MIT prove social media can slow the spread of fake news
Mark Wilson, Fast Company, 2021/06/11

There are plenty of reasons to be sceptical about the claims made in this article, not the least of which is the completely unsubstantiated claim that literacy and fact-checking have no impact on the spread of fake news. There is some reason to believe that an additional mechanism, user interface design, may also play a role. "The point," says MIT professor David Rand, "is that the platforms are, by design, constantly distracting people from accuracy." Not exactly. It's more like, if you prime people to think about accuracy beforehand, this will slow their sharing of inaccurate headlines, and the platforms don't do that. And, of course, this would have no impact on bots or on people who share maliciously. You can read the full results in their study published in an in-house journal, Harvard Kennedy School Misinformation Review (12 page PDF).

Web: [Direct Link] [This Post]