Who WOULD you want to create the solution? Cathy O'Niel described an algorithm as "bias codified as math" in Weapons of Math Destruction. "Bias" as a neutral but honest term - "A preference or an inclination, especially one that inhibits impartial judgment." Training data IS bias. It favors one answer over another, that's the point. The problem is in the codification. As a reminder: Pornhub was utterly bereft of Paris Hilton (civil tort, good lawyers) and utterly awash in Traci Lords (literal child pornography, no enforcement). Everyone got freeze peaches about how they couldn't fap it to revenge porn anymore and nobody gave much of a shit what that did to people. It was a real problem but it wasn't until someone threatened the money that Pornhub lifted a finger. Cathy O'Niel's solution to the bias in algorithms is to make them OPEN which is something that simply can't be done with LLMs. So we're cool with OpenAI telling the incels to dickmax? Are we cool with OpenAI telling the White Nationalists to kikepurge? What about ISIS? what's OpenAI's obligation to anyone targeted by ISIS?