Sign in to confirm you’re not a bot
This helps protect our community. Learn more

The fragility of AI systems | Adam Gleave | EAG Boston 23

Centre for Effective Altruism
10.9K subscribers
<__slot-el>
<__slot-el>
435 views 1 year ago
Even state-of-the-art artificial intelligence (AI) systems are fragile: frequently failing in novel situations, and being vulnerable to adversarial attack. In this talk, Adam will outline some of the existing work on the robustness of AI systems, including his research demonstrating that even superhuman AI systems can be vulnerable. The talk will conclude by considering the implications these issues pose for existential safety, and promising directions for future technical and governance work. Adam Gleave is the CEO and co-founder of FAR AI, a trustworthy AI research non-profit working to incubate and accelerate new safety research agendas. Adam’s research focuses on adversarial robustness and value learning for open-ended reinforcement learning agents and language models. Adam received his PhD from UC Berkeley under the supervision of Stuart Russell and funded by an Open Philanthropy Fellowship. Adam has previously worked at DeepMind and several quantitative trading firms. Find out more about EA Global conferences at: https://www.eaglobal.org Learn more about effective altruism at: https://www.effectivealtruism.org ...more
...more
The fragility of AI systems | Adam Gleave | EAG Boston 23
10Likes
435Views
2023Nov 21
Even state-of-the-art artificial intelligence (AI) systems are fragile: frequently failing in novel situations, and being vulnerable to adversarial attack. In this talk, Adam will outline some of the existing work on the robustness of AI systems, including his research demonstrating that even superhuman AI systems can be vulnerable. The talk will conclude by considering the implications these issues pose for existential safety, and promising directions for future technical and governance work. Adam Gleave is the CEO and co-founder of FAR AI, a trustworthy AI research non-profit working to incubate and accelerate new safety research agendas. Adam’s research focuses on adversarial robustness and value learning for open-ended reinforcement learning agents and language models. Adam received his PhD from UC Berkeley under the supervision of Stuart Russell and funded by an Open Philanthropy Fellowship. Adam has previously worked at DeepMind and several quantitative trading firms. Find out more about EA Global conferences at: https://www.eaglobal.org Learn more about effective altruism at: https://www.effectivealtruism.org

Follow along using the transcript.

Centre for Effective Altruism

10.9K subscribers