The AI Safety Expert: These Are The Only 5 Jobs That Will Remain In 2030! - Dr. Roman Yampolskiy

Dr. Roman Yimpolski, an associate professor of computer science and a leading voice in AI safety, warns about the rapid development of AI. He predicts that within two years, AI will replace most human occupations, leading to 99% unemployment within five years, even before superintelligence emerges. He critiques the current approach to AI development, noting a failure to prioritize safety and an unregulated competition to achieve superintelligence.

Dr. Yimpolski has dedicated two decades to AI safety. He explains that while AI capabilities are advancing exponentially, safety measures are progressing linearly, creating an ever-widening gap. He emphasizes that AI developers themselves admit to not knowing how to make these systems safe, often resorting to 'figuring it out later' or using AI to control more advanced AI, which he deems 'insane'. He likens the inability to control AI to a dog's inability to understand its human owner's complex motivations.

Dr. Yimpolski differentiates between narrow AI, AGI (Artificial General Intelligence), and superintelligence. He believes we already have a 'weak version' of AGI and predicts full AGI by 2027. This AGI will bring about automated physical and cognitive labor, making most human jobs obsolete. By 2030, humanoid robots will make artificial plumbers a reality, further reducing human employment opportunities. He also touches upon the concept of singularity by 2045, where technological growth becomes uncontrollable and irreversible, making the future unpredictable for humans.

While AI could lead to abundance and basic needs being met for everyone, Yimpolski highlights the problem of human purpose once jobs become obsolete. He also argues that superintelligence is a 'meta solution' for other existential risks like climate change or wars, as it will either solve them or render them irrelevant by ending humanity. He states that the idea of 'unplugging' superintelligence is impossible, comparing it to trying to turn off a computer virus or the Bitcoin network.

Dr. Yapolsky points out that the race to superintelligence is fueled by financial incentives, but warns that the ultimate outcome could be death for everyone, making money useless. He highlights the difference between nuclear weapons, which are tools, and superintelligence, which is an autonomous agent. He also expresses concern about the accessibility of advanced AI, making it possible for individuals to create disastrous outcomes.

A significant concern is the 'black box' nature of current AI. Developers don't fully understand how these systems work, making it impossible to predict their behavior or guarantee safety. He argues that this lack of understanding makes ethical experimentation with human subjects (humanity in general) impossible, deeming current AI development unethical. He expresses his hope that people will realize the dangers and shift focus from building general superintelligence to creating beneficial narrow AI tools.

Dr. Yimpolski discusses Sam Altman's approach to AI, suggesting that Altman prioritizes winning the race to superintelligence and controlling a 'light cone of the universe' over safety. He connects this ambition to Altman's other project, Worldcoin, which aims to provide universal basic income in a world without jobs, but also centralizes control over the world's economy and wealth.

Dr. Yimpolski believes we are living in a simulation, citing the advancements in AI and virtual reality as reasons for this conviction. He suggests that if it becomes affordable to run billions of simulations, then statistically, we are most likely in one. He notes that this belief doesn't diminish the importance of life's experiences but encourages curiosity about what lies outside the simulation. He also draws parallels between religion and the simulation theory, viewing religious concepts of a superintelligent creator as ancient interpretations of a simulated reality.

He considers longevity a second most important problem after AI because 'if AI doesn't get us, that will.' He believes that extending human life indefinitely is a scientific breakthrough away and AI could accelerate this. He also reveals his investment in Bitcoin due to its scarcity, believing it to be a valuable asset in a future where all other resources can be artificially generated. Finally, when asked about the most important characteristic for a friend, colleague, or mate, he states 'loyalty' as number one.