Study: AI models fail to reproduce human judgements about rule violations
Models trained using common data-collection techniques judge rule violations more harshly than humans would, researchers report.
Using reflections to see the world from new points of view
A new computer vision system turns any shiny object into a camera of sorts, enabling an observer to see around corners or beyond obstructions.
Training machines to learn more like humans do
Researchers identify a property that helps computer vision models learn to represent the visual world in a more stable, predictable way.
Researchers create a tool for accurately simulating complex systems
The system they developed eliminates a source of bias in simulations, leading to improved algorithms that can boost the performance of applications.
AI system can generate novel proteins that meet structural design targets
These tunable proteins could be used to create new materials with specific mechanical properties, like toughness or flexibility.
A method for designing neural networks optimally suited for certain tasks
With the right building blocks, machine-learning models can more accurately perform tasks like fraud detection or spam filtering.
Learning to grow machine-learning models
New LiGO technique accelerates training of large machine-learning models, reducing the monetary and environmental cost of developing AI applications.
New method accelerates data retrieval in huge databases
Researchers used machine learning to build faster and more efficient hash functions, which are a key component of databases.
Efficient technique improves machine-learning models’ reliability
The method enables a model to determine its confidence in a prediction, while using no additional data and far fewer computing resources than other methods.
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.