Dr. Yimpolsky asserts that controlling a superintelligent agent is not just a difficult engineering challenge, but a fundamentally impossible problem. He argues that the gap between AI's exponential capability growth and the linear progress in safety research is insurmountable, making any attempt to build AGI an uncontrollable experiment.
Yimpolsky predicts that AGI will be capable of automating most cognitive and physical labor by 2027-2030, leading to unemployment levels approaching 99%. He warns that society and governments are completely unprepared for this rapid economic and social disruption.
The development of an agent smarter than all humans is framed as the single most significant threat to humanity's existence. Yimpolsky argues that such an entity would be unpredictable and uncontrollable, making its creation a gamble with 8 billion lives.
Yimpolsky criticizes the leaders of major AI labs, particularly OpenAI's Sam Altman, for prioritizing the race to AGI over safety. He suggests their motivations are driven by a desire for legacy, power, and profit, leading them to ignore dire warnings from the safety community.
Yimpolsky expresses near-certainty that we are living in a computer simulation. He bases this on the argument that any civilization capable of creating realistic simulations would run billions of them, making it statistically inevitable that we are in a simulated reality rather than the base reality.
Keep pulling the thread on Dr. Roman Yampolskiy.