A Canadian legislator read an apparent line generated by a large language model (LLM) during a floor speech last month, sparking social media attention this week. Bill Oliver, a Progressive Conservative Party member of the legislative assembly of New Brunswick, included a line in his speech that appeared to be an LLM-generated alternative phrasing of a legislative section. The line, which read like a natural legislative speech rather than a series of short points, was noted for its hallmarks of an LLM response to a prompt. The odd reading went largely unnoticed at the time, but video of the remarks began circulating on platforms like Reddit and Threads earlier this week.
The incident has drawn mainstream attention in Canada, with outlets like the Canadian Broadcasting Corporation and The Toronto Star commenting on the event. The Toronto Star described it as a sign of a growing societal divide between elites who delegate duties to AI and the masses who find this practice objectionable. Oliver is not the first politician to read a speech prepared by someone else, but the public nature of his apparent AI-generated non-sequitur represents a new level of embarrassment in AI use.
The incident highlights a broader trend of professionals whose AI use has been revealed in mortifying ways. Lawyers, authors, journalists, and academics have all faced issues when their LLM use was exposed, often due to hallucinated errors in their work. A Duke University study found that workers tend to hide their AI use because colleagues view it as lazy or replaceable.
Source: arstechnica
