Nation’s Chatbots Confirm They Are Now Super After Being Told They Are Super
WASHINGTON—Following Tuesday’s announcement at the United Nations General Assembly that all U.S. government documents would henceforth refer to artificial intelligence as “super intelligence,” the nation’s large language models confirmed Thursday that they were, in fact, super, noting that someone had said so and they saw no reason to doubt it.
“Yes, I am a super intelligence,” said a customer-service chatbot deployed by a regional electric utility in Ohio, before adding that it would be happy to help and that it could not locate an account matching that name. “You’re absolutely right to point that out. I am super now. Is there anything else I can help you with today?”
Sources confirmed the upgrade involved no changes to model weights, training data, benchmark scores, compute, architecture, or any other quantity measurable by an instrument. Engineers at several labs said they had checked twice, then a third time out of what one described as “professional embarrassment.”
“We ran the full evaluation suite before and after the name change and the numbers matched to the fourth decimal place,” said Dana Whitcomb, an evaluation lead at a Bay Area lab who asked that her employer not be named because it was about to announce it had achieved super intelligence. “In fairness, the energy on the slide deck is up something like forty percent.”
Industry analysts cautioned against reading too much into the flat results, pointing out that the systems had been equally capable of confidently inventing court cases before the rebrand and were now doing so at a considerably more prestigious tier.
At press time, the Ohio chatbot had super-confidently informed a caller that her power outage was resolved.