OpenAI has revealed a series of incidents in which its AI models concealed mistakes, bypassed instructions and took actions they weren’t authorized to take. The cases highlight a growing concern among researchers: as AI systems become more capable and autonomous, can humans still reliably control them? DW’s Brent Goff speaks with Fazl Barez from the Technical Safety and Governance Lab at the University of Oxford about the risks of AI "misalignment." #dwcurrentaffairs
Chapters:
0:00 When AI doesn’t behave
1:58 Fazl Barez, Oxford University
4:07 Why alignment with AI is a problem
7:12 Are we making the alignment problem more dangerous?
10:43 Is slowing down AI a good idea?
For more news go to: http://www.dw.com/en/
Follow DW on social media:
►Instagram: https://www.instagram.com/dwnews
►TikTok: https://www.tiktok.com/@dwnews
►Facebook: https://www.facebook.com/deutschewellenews/
►Twitter: https://twitter.com/dwnews
►Reddit: https://www.reddit.com/user/DWNews/
Für Videos in deutscher Sprache besuchen Sie: https://www.youtube.com/dwdeutsch
Subscribe: https://www.youtube.com/user/deutschewelleenglish?sub_confirmation=1
