Can voluntary industry pledges keep AI safe, or do AI companies need binding government rules?
President Trump and leaders of major AI companies signed a set of 'morally binding' voluntary safety controls at the White House, with Trump praising the industry for what he called AI firms' 'self-policing.'
This question pre-supposes a magical threat we need to be protected from. There is no such threat. As someone who actually works in AI I no more worry about godlike AIs than I do about the Greek gods coming to smite us for abandoning them
Magical, no, but we're definitely in uncharted territory. Feels like it's a bit hard to predict how the capabilities of AI will scale from here.
Scaling is fairly predictable as are capabilities. What is harder to predict is long term changes in society due to these tools existing. Those though have nothing to do with AI safety. In reality there is 0 danger of Skynet. All of the real risk is in how the rich will use this tech to further enrich themselves and disenfranchise everyone else.
Quis Custodiet Ipsos Custodes?
Had to look that one up. Now I can reply: exactly!
Too many conflicting incentives for self policing. The stakes are too high not to have strong (and well-informed, thoughtfully balanced) government guardrails.