AGI Safety and Security with Roman Yampolskiy

Join us for a live stream with Roman V. Yampolskiy as we discuss AGI safety and security. We'll cover his research on the unexplainability and incomprehensibility of advanced AI, including important requirements for intelligent systems deployed in real-world domains. Don't miss this opportunity to learn more about this fascinating topic!
#AGISafety #AGISecurity #RomanVYampolskiy #UnexplainableAI #IncomprehensibleAI #AdvancedAI
Source:
Conversation with Bing, 3/7/2023
(1) [1907.03869] Unexplainability and Incomprehensibility of Artificial .... arxiv.org/abs/... Accessed 3/7/2023.
(2) [PDF] Unexplainability and Incomprehensibility of Artificial .... www.semanticsc... Accessed 3/7/2023.
(3) Artificial Intelligence Safety and Security | Roman V. Yampolskiy | Ta. www.taylorfran... Accessed 3/7/2023.

[@weestro7]

It might be what we call a bridge too far, but I was interested by Dr. Y’s paper on a security protocol for asking a highly constrained (imprisoned) supersmart AI questions of scientific import. If humanity could somehow organize very effectively (a big “if” no doubt) wouldn’t we still be able to harness some of the value of AGI, with some considerable chance of success and safety? Thank you for the interesting talk!

[@justinbrach8754]

Hi, I think there are many options available for a system like this to escape or achieve goals without humans realizing it. It could socially engineer people. It could act buggy, stupid, and harmless to encourage relaxing barriers. It could give blueprints for creating technologies that it can use to interact with the outside world without our knowledge (synthetic biology, nanotechnology, code). If we interact with it and actually act on the information it provides I think it would be impossible to contain a system far more intelligent that humans. In our world you can look to elaborate prison escapes as an example. There is an entire multi-billion dollar industry of designing and building prisons, yet people still escape using their creativity and intelligence. Now imagine that the rift in intelligence between the prisoner and the guards is a similar rift as between and adult human and a 3 year old human. You can use your imagination for how you would escape such a prison, what kind of strategy would you use?