People are blaming OpenAI's odd corporate structure for this weekend's chaos, but I think the main issue is that their strategy has been incoherent for a long time. It was never clear how accelerating AI development would save us from dangerous AI. understandingai.org/p/firing-sam-a…

@binarybits isn't it? developing evidence on the risks of a technology always requires research that advances the technology. we can't mitigate what we don't understand

@binarybits from interviews with ilya and others, the basic thought is that work on alignment actually accelerated capability, and in the other direction improved capability gives you tools for better alignment work a core idea behind openai’s superalignment project: they go hand-in-hand

@binarybits Would you rather have Sam & Microsoft control AI or China, Russia or other government? I'd rather have FOSS AI but seems private enterprise is 2nd best given Mao, Stalin, Hitler, Pol Pot were all government leaders

@binarybits Agree. Fortunately, the whole doomer premise is wrong. Superhuman AI will not explode and take over. The first superhuman system will not eliminate others. Superhuman AI isn’t even a well-defined thing

@binarybits OpenAI's corporate structure may be a factor, but the real problem lies in the lack of a coherent strategy. Simply accelerating AI development won't ensure our safety from dangerous AI. We need a comprehensive approach that addresses all aspects of AI's impact.

@binarybits Exactly, Altman was the one who changed the original mission… looking for money and power, there is not other reason

@binarybits "Hey ChatGPT, please compose a corporate strategy that makes us rich🙏 but without inciting a Skynet-type uprising of the machines."