AI Could Be ‘Poison Apple,’ Experts Warn

Generative artificial intelligence (AI) could be a “poison apple,” according to a joint report by experts including University of Tokyo Prof. Fujio Toriumi and Keio University Prof. Tatsuhiko Yamamoto.

Released Monday, the report is the second such paper compiled by the experts since January 2022.

The 2022 report stressed the importance of “information health,” referring to internet users’ capacity to resist misinformation by consuming content from a diverse range of sources.

The latest report includes issues associated with generative AI and warns that generated text could be disseminated without knowledge about the source material used to produce the content.

As generative AI tools can produce seemingly plausible responses to queries, the technology could be a “poison apple” that adversely affects our “information health,” according to the experts.

The report also refers to the roles that should be played by telecommunications companies and advertising agencies, and touches on education and IT literacy.

The experts proposed the introduction of a system to measure “information health” and the development of IT literacy educational materials, among other recommendations.

The 2022 document highlighted issues such as fake news and “filter bubbles,” in which internet users encounter only information that reinforces their own beliefs.

The latest report involved 16 co-authors, including University of Tokyo Prof. Kuniyoshi Sakai, who specializes in the neuroscience of language. Toriumi specializes in computational social science and Yamamoto is an expert on constitutional law.

“We’d like to promote the concept of ‘information health’ not only in Japan but also around the world,” Yamamoto said.