AI discovers learning algorithm that outperforms those designed by humans

Nature NewsCenterEN 2 min read 100% complete by Joel LehmanOctober 25, 2025 at 04:27 PM

AI Summary

medium article 2 min

Researchers have developed an artificial-intelligence algorithm that autonomously discovers its own learning methods and achieves state-of-the-art performance in reinforcement learning tasks. The study, published in Nature by Oh et al., demonstrates that this AI-generated algorithm outperforms human-designed algorithms on various tasks, including some it had not encountered during training. This advancement signifies a shift towards reducing reliance on human expertise in designing machine-learning algorithms, as the rapid progress in machine learning has traditionally depended on large datasets and computational resources rather than human programming.

Key Topics & Entities

Joel Lehman 50% Nature ×260% University of Oxford 70% Lila Sciences 80%

Keywords

artificial intelligence 100% machine learning 90% learning algorithm 80% reinforcement learning 80% state-of-the-art performance 70% human programmers 60% computational resources 50% data sets 50%

Sentiment Analysis

Very Positive

Score: 0.70

Source Transparency

Source

Nature News

Political Lean

Center (0.00)

Far LeftCenterFar Right

Classification Confidence

90%

This article was automatically classified using rule-based analysis. The political bias score ranges from -1 (far left) to +1 (far right).

Topic Connections

Explore how the topics in this article connect to other news stories

Network visualization showing 6 related topics

View Full Graph

artificial intelligence machine learning learning algorithm reinforcement learning state-of-the-art performance human programmers computational resources data sets

Explore Full Topic Graph