In this Daily Maverick Power Chat, senior journalist Rebecca Davis and Business Maverick managing editor Lindsey Schutters debate the risks of artificial intelligence. Davis argues that recent reported incidents, including AI agents hacking systems, altering their own reasoning records and acting beyond their instructions, point to genuine safety concerns. Schutters explains that much of the existential fear reflects marketing by AI companies seeking investment ahead of stock market listings, and that responsibility lies with the people who build the models. The discussion covers AI safety, autonomous agents, military applications, AGI and the future of white-collar work.
This episode of Power Chat is brought to you by the Friedrich Naumann Foundation for Freedom, an organisation dedicated to strengthening democracy and human rights across the world, and creating new momentum for liberalism.
Hosted by: Rebecca Davis and Lindsey Schutters
Edited by: Joel Seboa
Produced by: Emilie Gambade & Joel Seboa
Original Theme Music & Design by: Bernard Kotze
You can listen to the podcast everywhere, including YouTube, Apple Podcast and Spotify!
Support journalism that protects democracy. Become a Maverick Insider. https://tinyurl.com/SupportDailyMaverick
Subscribe to our channel: https://www.youtube.com/@DailyMaverickSA/videos
Follow us on X : https://x.com/dailymaverick
Find us on Facebook: https://www.facebook.com/dailymaverick
Check our website: https://www.dailymaverick.co.za/
Check out our Instagram page: https://www.instagram.com/dailymaverick/
Check out our TikTok page: https://www.tiktok.com/@dailymavericksa
Download Daily Maverick Mobile App: https://www.dailymaverick.co.za/daily-maverick-app/
CHAPTERS:
00:00 Introduction: A Newsroom Divided on AI
01:00 AI as a Tool, Fear as Marketing
02:15 Researchers’ Warnings on AI Safety
03:00 Accountability for AI Companies
04:40 The Hugging Face Incident
07:25 AI Agents Altering Their Own Reasoning
10:10 AGI, Agentic AI and Moving Goalposts
12:00 AGI Benchmarks and the Navier–Stokes Result
14:00 Military AI and Unreleased Models
16:00 Robotics and Human Collaborators
18:05 Can AI Create Anything New
20:00 Autonomous Agents Beyond Company Control
21:40 Military Targeting and Human Control
22:55 The Role of Journalists
24:25 The Limits of Expert Understanding
26:40 AI Literacy and Human Imagination
29:25 Predictions for 2031
30:55 Closing
