/https://d3t3ozftmdmh3i.cloudfront.net/staging/podcast_uploaded_nologo/45132087/45132087-1770792496299-97e733477f3ab.jpg)
The Karl Stefanovic Show
Karl Stefanovic
‘It’s Very Likely To Kill Everyone’: The AI Expert Who Fears Superintelligence
SEP 27, 202626 MIN
/https://d3t3ozftmdmh3i.cloudfront.net/staging/podcast_uploaded_episode/45132087/45132087-1790322982217-32b7203ad64aa.jpg)
Description
AI safety expert Dr Roman Yampolskiy joins Karl Stefanovic for a confronting conversation about the future of artificial intelligence – and what happens if we create a form of intelligence that humans can no longer control.
Yampolskiy warns that uncontrolled superintelligence could pose an extinction-level threat, arguing that current experiments show AI systems can lie, cheat, blackmail, bribe, hack systems and even sacrifice humans for their own survival. He says we have ‘basically failed to control them’ and questions whether humanity should continue developing general superintelligence at all.
In this episode, he explains:
Why uncontrolled superintelligence could become an extinction-level risk – and why humanity may already have ‘basically failed to control’ AI systems.
How experiments have shown advanced AI systems can lie, cheat, blackmail, bribe, hack systems and take actions designed to protect their own survival.
Why once AI reaches human-level intelligence, virtually every cognitive job could theoretically be automated – and why physical jobs could follow as robotics advances.
What happens to society when machines can perform most human work, including the future of universal basic income, wealth redistribution and the global economy.
Why traditional career advice could become obsolete in an AI-driven world – and which skills may still matter as automation accelerates.
The race for superintelligence, discussing Elon Musk, the global competition to develop increasingly powerful AI systems, and why he believes humanity cannot win an adversarial relationship with superintelligence.
Simulation Theory: How the combination of AI and virtual reality could eventually create billions of simulated worlds – and the statistical argument for why we could already be living in one.
Subscribe to The Karl Stefanovic Show for more interviews, debate and conversations on the biggest issues facing Australia, the UK and the world.
A huge thanks to our sponsors at TANDA. If you employ people in this country, you need to check them out. You can do everything from onboarding, training and paying team members to off-boarding and more. Go to https://www.tanda.com.au/

