New insights into training dynamics of deep classifiers
MIT researchers uncover the structural properties and dynamics of deep classifiers, offering novel explanations for optimization, generalization, and approximation in deep networks.
Computer science and technology
Large language models are biased. Can logic help save them?
MIT researchers trained logic-aware language models to reduce harmful stereotypes like gender and racial biases.
Robot armies duke it out in Battlecode’s epic on-screen battles
The long-running programming competition encourages skills and friendships that last a lifetime.
Integrating humans with AI in structural design
A process that seeks feedback from human specialists proves more effective at optimization than automated systems working alone.
Solving a machine-learning mystery
A new study shows how large language models like GPT-3 can learn a new task from just a few examples, without the need for any new training data.
Automating the math for decision-making under uncertainty
A new tool brings the benefits of AI programming to a much broader class of problems.