Why Tool AIs Want to Be Agent AIs · Gwern.net
AIs limited to pure computation (Tool AIs) supporting humans, will be less intelligent, efficient, and economically valuable than more autonomous reinforcement-learning AIs (Agent AIs) who act on their own and meta-learn, because all problems are reinforcement-learning problems.
--- title: "Why Tool AIs Want to Be Agent AIs" description: "AIs limited to pure computation (Tool AIs) supporting humans, will be less intelligent, efficient, and economically valuable than more autonomous reinforcement-learning AIs (Agent AIs) who act on their own and meta-learn, because all problems are reinforcement-learning problems." created: 2016-09-07 modified: 2018-08-28 status: finished confidence: likely importance: 9 css-extension: dropcaps-kanzlei ... > Autonomous AI systems (Agent AIs) trained using [reinforcement learning](!W) can do harm when they take wrong actions, especially
saved by
- Blake Elias
- Anson Yu
- abinaya dinesh
- Ratan Kaliani
- Aryan Naik
- Benedict Neo
- Nhật Nguyễn Thế
- Aaron Pham
- Lydia Nottingham
- Sudarsh K
- Timothy Kostolansky
- Adi Mehta
related reading
- Guardian Angels: LLM Personalization for Productivity and Security · Gwern.netgwern.net
- Evolution as Backstop for Reinforcement Learning · Gwern.netgwern.net
- Reward is not the optimization target — LessWronglesswrong.com
- Simulators — LessWronglesswrong.com
- Using Artificial Intelligence to Augment Human Intelligencedistill.pub
- The behavioral selection model for predicting AI motivations — LessWronglesswrong.com
- Embedded Agency (full-text version) — LessWronglesswrong.com
- Without specific countermeasures, the easiest path to transformative AI likely leads to AI takeover — LessWronglesswrong.com
- Reward Is Not the Optimization Targetturntrout.com
- Questions about the Future of AI - by Dwarkesh Pateldwarkesh.com
- Building better AI tools | Hacker Newsnews.ycombinator.com
- How training-gamers might function (and win)blog.redwoodresearch.org