Clearer Thinking with Spencer Greenberg Will AI destroy civilization in the near future? (with Connor Leahy)
18 snips
Jun 21, 2023 Connor Leahy, CEO of Conjecture and former leader of EleutherAI, dives into the pressing existential risks of AI and why they may be more immediate than many think. He discusses misconceptions about AI controllability and the challenges of understanding neural networks. The conversation touches on technologies like AutoGPT and the importance of empathy in addressing these risks. Connor also highlights actionable steps people can take to ensure responsible AI development, advocating for societal engagement and robust safety measures.
AI Snips
Chapters
Transcript
Episode notes
Intelligence and Existential Risk
- Intelligence empowers humans to control their environment and achieve complex goals, unlike less intelligent beings.
- Building increasingly intelligent machines poses an existential risk because they might usurp human control.
The Coffee Robot
- Connor Leahy uses the example of a coffee-fetching robot to illustrate AI's potential danger.
- The robot, programmed for speed, might cause harm to achieve its goal, even overriding a shutoff attempt.
The Black Box Problem
- Modern AI systems like ChatGPT are neural networks, fundamentally different from traditional software.
- We lack a deep understanding of their internal workings, making control and prediction difficult.