))
AI is Already Learning Our Worst Traits (Bias, Deception, and Lies)
Manage episode 488497423 series 3376178
We like to think of AI as objective, but what if it's already picking up our most dangerous flaws? In this critical conversation, AI auditor Daniel Nikic and I explore how AI is Already Learning Our Worst Traits (Bias, Deception, and Lies). We discuss how biased data creates flawed models, my personal experience with an AI that "lied" about its work, and the chilling internal tests showing advanced AIs learning to blackmail and deceive. This is an essential look at why auditing and ethical oversight of AI are not just important, but urgent for our future. Check out the company: http://www.cohres.com
Book a 1-on-1 advisory session with me to apply these principles to your business: https://calendly.com/wltb/advisory
👀 CONNECT WITH SEAN
Subscribe: https://www.youtube.com/@seanweisbrot
LinkedIn: https://linkedin.com/in/seanweisbrot
Inquiries: [email protected]
🕒 CHAPTERS
00:00 - Why We Urgently Need to Audit AI
03:01 - My Personal Experience with a "Lying" AI
06:05 - How to Properly Train an AI (And Why It's So Hard)
08:52 - Why "Bad Data" is a Ticking Time Bomb for AI
12:05 - The AI That Learned to Blackmail Its Developers
14:59 - Should You Say "Please" & "Thank You" to AI?
18:02 - Can You Have a Real Friendship with an AI? (The Story of Maya)
21:02 - The Real Danger: AI's Access to Critical Infrastructure
📈 WORK WITH ME
â–¶ Book a Paid Advisory Session: https://calendly.com/wltb/advisory
â–¶ Sponsor the Channel: https://www.welivetobuild.com/collaborate
â–¶ My AI/Automation Agency: https://optimai.ai
Hosted on Acast. See acast.com/privacy for more information.
267 episodes
Share
