Artificial intelligence

Read about interesting models and framings for AI safety here: https://docs.google.com/document/d/145yJBoNTYHOJ_FMOO2hO-x2KnJQT45hxhtd0I84HVLE/edit

https://blog.openai.com/debate/

https://maliciousaireport.com

OpenAI - Concrete AI safety problems paper

AI ethics

See here for David Dustch comment (linked to specific time): https://vimeo.com/22099396#t=2758s

https://maliciousaireport.com/

https://medium.com/@francois.chollet/the-impossibility-of-intelligence-explosion-5be4a9eda6ec – comments here: https://www.facebook.com/guillermovalleperez/posts/10156139287091223

Excellent article not only on what may be the most tangible current AI risk, but more importantly, its possible solution. The choice forks the future into a potential dystopia or a more humane and enritched society.

https://medium.com/@francois.chollet/what-worries-me-about-ai-ed9df072b704

https://aiimpacts.org/likelihood-of-discontinuous-progress-around-the-development-of-agi/

Adversarial examples

What is the simplest way to attack a model. Justin Glimer. Security of ML model is about test error basically. Defend against attackers trying random stuff to fool model. VIDEO