New York City
Companies That Own The Future Of Human Knowledge Plead Ignorance Under Oath
By BSN Staff, . BSN Network. Satire.
NEW YORK, NY. Representatives from four of the world's most powerful artificial intelligence companies sat before the New York City Council on Monday under oath, having collectively trained on approximately 90 percent of the written record of human civilization, and declined to answer whether their products might destroy it.
OpenAI, Anthropic, Meta and Google sent representatives to City Hall in response to New York's push to regulate the AI industry, following warnings from inside whistleblowers that the companies would not provide clear answers to basic safety questions. The companies did not provide clear answers to basic safety questions. The whistleblowers have not said whether they feel vindicated, but sources close to their facial expressions suggest yes.
Central among the unanswered questions: what is the probability of a catastrophic outcome caused by AI? Lawmakers asked it four separate ways. The representatives answered it zero separate ways, cycling instead through a rotation of "it's a complex space," "we take safety extremely seriously," and one answer that lasted six minutes and contained the word "framework" eleven times.
"We are deeply committed to responsible innovation and to working collaboratively with regulators to ensure that the deployment of these technologies reflects the values of the communities they serve," said one of the four representatives, in a statement that could have been generated by the product they were there to explain.
Council members pressed on specifics. Could the companies put a number on the risk? No. Could they define the threshold at which an AI system becomes dangerous? Not at this time. Had any of them read the whistleblower testimony submitted in advance? That question was met with the kind of pause that costs $400 an hour in a different room.
Meta's representative noted that the company has an internal AI safety team, which the council appeared to receive as reassuring until a follow-up question established that the team's findings are not shared externally, are not subject to independent review, and are described internally as "ongoing." Meta has been conducting ongoing AI safety work since 2016, which is also the year it concluded that engagement algorithms were fine.
Anthropig, whose entire brand identity rests on being the safety-conscious AI company, sent a representative who explained that safety is "a journey, not a destination," a framing that works equally well for a tech company and for a person who has never once arrived anywhere.
Google's representative explained that the company operates under a set of internal AI principles established in 2018. When asked how those principles had performed so far, the representative said the question was outside the scope of today's hearing.
By the second hour, several council members had begun asking questions directly into the table.
The hearing produced no binding commitments, no numerical risk estimates, no agreed definitions of the word "safe," and one pamphlet. The pamphlet described the city's regulatory ambitions. It was printed on paper, which none of the four companies had anything to do with, and which has so far caused zero catastrophic outcomes.
Legislation is expected to follow. The companies have indicated they look forward to continued dialogue.
The story we are making fun of: https://www.amny.com/news/ai-giants-nyc-council-whistleblower-warnings/