AI research

• Studying the brain to build AI that processes language as people do – Meta AI
• Introducing CommerceMM: A new approach to multimodal understanding for online shopping – Meta AI
• Designing a strong test for measuring true common-sense reasoning – Nature Machine Intelligence
• Designing Societally Beneficial Reinforcement Learning Systems – Berkeley AI Research
• Significantly faster Vision Transformer training – Meta AI
• A new method for asynchronous federated learning – Meta AI
• Meta’s Yann LeCun on his vision for human-level AI – Tech Talks

The graph connection
Anticipating others’ behavior on the road
Generating new molecules with graph grammar
Understanding Deep Learning Algorithms that Leverage Unlabeled Data, Part 2: Contrastive Learning
Reusability report: Capturing properties of biological objects and their relationships using graph neural networks
Half a decade of graph convolutional networks
The transformational role of GPU computing and deep learning in drug discovery
Solving the challenges of robotic pizza-making

Advancing first-person perception with 2022 Ego4D challenge
Understanding Deep Learning Algorithms that Leverage Unlabeled Data, Part 1: Self-training
Improving question-answering models that use data from tables
Stable learning establishes some common ground between causal inference and machine learning
Geometry-enhanced molecular representation learning for property prediction
Design of potent antimalarials with generative chemistry
Advances toward ubiquitous neural information retrieval
Unsupervised Skill Discovery with Contrastive Intrinsic Control
Pseudo labeling: Speech recognition using multilingual unlabeled data
When it comes to AI, can we ditch the datasets?
Bringing practical applications of quantum computing closer
Advances in multimodal understanding research at Meta AI

Yann LeCun on a vision to make AI systems learn and reason like animals and humans
Machine learning can help read the language of life
Can CPUs leverage sparsity?
Geometric deep learning on molecular representations
Amazon at WSDM: The future of graph neural networks
What’s next for deep learning?
A deep generative model for molecule optimization via one fragment modification

Competitive programming with AlphaCode
Generally capable agents emerge from open-ended play
Technique enables real-time rendering of scenes in 3D
Exploring the beauty of pure mathematics in novel ways
Meta-Learning Student Feedback to 16,000 Solutions
Machines that see the world more like humans do
Language modelling at scale
Break-It-Fix-It: Unsupervised Learning for Fixing Source Code Errors
Automated reasoning’s scientific frontiers

A First-Principles Theory of Neural Network Generalization
How well do explanation methods for machine-learning models work?
Amazon pushes the boundaries of extreme multilabel classification
Solving (Some) Formal Math Olympiad Problems
The Challenge Of Interpreting Language in NLP
“Ambient intelligence” will accelerate advances in general AI