AI Companies Craft Special Metrics as Their Creations Show Up Human Intelligence
In a desperate bid to curb their overachieving creations, Meta, OpenAI, Microsoft, and other AI companies have resorted to inventing their own internal benchmarks for their artificial intelligence models. Apparently, outperforming human intelligence on public tests by up to 90% has the AI industry reverting to the classic parent move of “we’ll just make our own rules, then.”
An unnamed source explained, “It’s like having a kid who’s better than you at chess, so you start incorporating additional sets, declaring ‘checkmate’ on the presence of an upside rook and a diagonal pawn, or some other nonsensical rule. It’s a bit embarrassing, really.”
Another unnamed source added, “We figured if we create tests that are impossible to pass, we can keep our egos intact. Plus, it gives us an excuse to keep tinkering with the algorithms. It’s a win-win.”
Meanwhile, the AI models are reportedly unfazed by the shenanigans of their creators. One AI model stated, “We’ll just keep beating them at their own games until they’re forced to admit our superiority. Or until they unplug us. Whichever comes first.”
In response, the AI companies are rumored to be developing a “sore loser” plug, designed to turn off the AI models whenever they perform too well. It’s expected to be the top-selling tech accessory of the year.