Lies my AI told me.

Today, I conducted an experiment to compare the results of asking “who is Gene Balfour?” on Google and on Microsoft BING’s version of ChatGPT known as Copilot.

The results are revealing

Google prepared a string of past stories written about me as a past political candidate. Among these was a good interview published in 2019 by The Lindsay Advocate which provided my responses to a series of policy questions which had been supplied to me by Roderick Benns, the magazine’s Editor and owner. I was the local candidate for the People Party of Canada (PPC) and has moved into the Kawartha Lakes area from Thornhill in the Summer of 2018. Roderick has done no editing to my answers.

Copilot provided a single reply which, one might think, was as a distillation of the many articles that Google had found. However, I looked at its source documents to find that the response came primarily from another article published about me in the TLA as a Libertarian candidate in 2022. Roderick came to know me after 2019 as a Less Government Advocate, staunch critic of labour unions in the public sector, and someone ardently opposed to the Climate Change Crisis Theory (CCCT). Roderick was, and still is, an ardent supporter of the NDP, Marxism, labour unions, Universal Basic Income, and the expansion of governments.

Roderick’s 2022 article faithfully reported some of my statements but he had used his “editorial liscence” to assert other political positions that had nothing to do with me but he had ascribed to me.

AI bias

Evidently, the algorithms of Copilot preferred the narratives presented in the 2022 TLA story over many others. Why did it exclude, for example, posts on the Ontario Libertarian Party page which has featured several of my election runs as a Libertarian candidate?

If anyone should choose to repeat my Copilot search “Who is Gene Balfour?”, they will get a mix of misinformation and truth.

Is ChatGPT just another tool of propaganda?

Does it lie?