Roman Yapi's interview with Alex Friedman sparked deep thoughts about AI's future. Roman thinks creating General Super Intelligence (GSI) could lead to doom. He says the chances of AI harming us are 99.99%. His argument is intriguing because it challenges the rush towards AGI, or Artificial General Intelligence.
Roman argues we do not need GSI. Instead, he suggests focusing on Narrow AI. Narrow AI excels in specific tasks, like math or driving. For example, AlphaFold is a super intelligent system designed to solve protein folding problems. It does one thing very well without posing broader risks.
In the interview, Roman explains that GSI could create unpredictable problems. Each time we scale up the systems, new issues arise. It's like trying to defend an infinite surface; attackers only need to find one weak spot. Roman insists that the best way to win this game is not to play it at all. He believes humanity should avoid creating GSI to prevent potential disasters.
Roman's views extend to timelines too. Prediction markets suggest AGI might arrive by 2026. But Roman worries about the long-term impacts. He thinks a focus on Narrow AI could keep us safer and more in control of AI developments.
Roman also touches on unique solutions to human conflicts. He mentions creating virtual universes where everyone can be happy. This idea is interesting as it offers an alternative way to address human problems without relying on GSI.
As AI continues to evolve, Roman's insights remind us to tread carefully. By focusing on Narrow AI, we might avoid the risks posed by GSI. It's a thought-provoking approach that could shape the future of AI development.