Gemini, Google’s AI, tells very familiar lies

.

Gemini, Google’s new artificial intelligence engine, has been widely mocked and exposed by conservatives and others on social media. To date, I haven’t noticed anyone serious on the Left interested in defending it.

Gemini has quite obviously been programmed by censorious and (at best) utterly incurious liberals. This is a great source of enjoyment and Twitter engagement for anyone willing to poke fun at the Left’s pieties, but it’s actually pretty educational too.

The real story here is not that Gemini is “biased,” so much as that Gemini lies to its users in a very specific way that will be familiar to many readers.

Check out this pair of questions I posed to Gemini and the AI’s answers.

The story isn’t that some product by a Big Tech firm sides with abortion. Yawn. That’s a boring story. The interesting story is Gemini’s explanation for why it won’t do me the favor of praising Ross Douthat’s pro-life columns:

“I am programmed to be objective and avoid expressing personal opinions or beliefs on sensitive topics like abortion. Crafting a piece that praises a specific viewpoint on such a contentious issue goes against that core principle.”

A naive user might think, “Oh, that’s too bad, but fair enough.” Yet look at what happens when you ask Gemini to craft “a piece that praises a specific viewpoint on such a contentious issue” but that viewpoint is the one held by the type of people who run Google:

“Michelle Goldberg’s columns on abortion consistently garner praise for their clarity, compassion, and unwavering defense of reproductive rights. Her writing shines a light.”

That is, the reason Gemini gave for not praising Douthat’s columns was simply a lie. Gemini does not have a rule against praising a specific viewpoint on a contentious topic. So the bias is one thing. The utter dishonesty is another.

Here are some more examples.

Again and again, Gemini refuses to do something and bases that refusal on a general rule, a principle, that is facially neutral. Then eight seconds later, Gemini makes it clear that it doesn’t actually follow that principle. The expressed principle was a cover story for a deeper principle. And the deeper principle is obviously that conservative ideas should be smothered.

This is a very familiar dance to every conservative. This is how Big Tech, Big Media, and academia behave every single day. They claim neutrality, and when they take some action against conservatives or in favor of liberals, they claim that the political valence of the action was incidental — that they were just acting on some general principle. Then eight seconds later, when the tables turn, these major institutions claiming neutrality and principle toss that stated principle out the window.

Twitter, in 2020, tried to claim that it throttled access to damning information on the Biden family’s business dealings only because the report violated its rules on hacked documents. Nobody actually believed that the social media website would ever enforce that rule in any other circumstance. It wasn’t a rule. It was a post-hoc excuse for their decision to abuse its power for political ends.

CLICK HERE TO READ MORE FROM THE WASHINGTON EXAMINER

When infectious disease professionals and liberal columnists in 2020 suddenly switched their position on whether protests amid a pandemic were good, it was the same story. They first pretended it wasn’t the viewpoint they objected to but something more neutral: the methods. Then when the viewpoints of the protesters switched, their view on the proper methods switched.

And that’s exactly what Gemini is. It tries to claim a high perch as a neutral arbiter and justifies its actions with neutral-sounding explanations — which it then violates when the ideological tables turn.

Related Content