Reinforcement-learning lets algorithms improve themselves faster than humans can.
Evidence
AlphaGo Zero learned only to imitate itself. It used its predictions to make better decisions, and then learned to predict those better decisions in turn. It began by making random predictions and random moves. Seventy-two hours later, it was the strongest Go player the world had ever seen. — Brian Christian, The Alignment Problem, 1057 (
#expands, via linked claim)
Given that there is significant bias in U.S. criminal justice2 and almost no interventions have successfully curtailed it, we think the time is right to introduce technology as a way to correct for human error. For some years now, it has been known that big data and AI often outperform human judgment alone (Berk, 2008; Lazer et al., 2009). — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1171 (
#qualifies, via linked claim)
Categorization
- corroborated: 2 independent passages support this claim
- XREF: connects to: Networked machine learning compounds individual errors into population-level knowledge faster than humans can. — Shoshana Zuboff, The Age of Surveillance Capitalism, 1319; Self-imitation can bootstrap a machine to superhuman skill — Brian Christian, The Alignment Problem, 1057; Big data and AI outperform human judgment in high-stakes decisions. — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law: A Definitive Look at…, 1171 (+3 more)
Related
- Self-imitation can bootstrap a machine to superhuman skill (similarity 0.82)
- Networked machine learning compounds individual errors into population-level knowledge faster than humans can. (similarity 0.81)
- Big data and AI outperform human judgment in high-stakes decisions. (similarity 0.81)
- Reinforcement-learning lets algorithms reprogram themselves using internal performance and reward subroutines. (similarity 0.80)
Broader Connections
- Self-imitation can bootstrap a machine to superhuman skill
#expands(0.85) — Brian Christian, The Alignment Problem, 1057 - Big data and AI outperform human judgment in high-stakes decisions.
#qualifies(0.75) — Joseph Avery, Jim Sidanius, Joel Cooper, Gregory Davis, Amanda N., Bias in the Law, 1171