News & Views

A classic question in cognitive science is whether learning requires innate, domain-specific inductive biases to solve visual tasks. A recent study trained machine-learning systems on the first-person visual experiences of children to show that visual knowledge can be learned in the absence of innate inductive biases about objects or space.

AI tools such as ChatGPT can provide responses to queries on any topic, but can such large language models accurately ‘write’ molecules as output to our specification? Results now show that models trained on general text can be tweaked with small amounts of chemical data to predict molecular properties, or to design molecules based on a target feature.

Recent work has demonstrated important parallels between human visual representations and those found in deep neural networks. A new study comparing functional MRI data to deep neural network models highlights factors that may determine this similarity.

The implementation of particle-tracking techniques with deep neural networks is a promising way to determine particle motion within complex flow structures. A graph neural network-enhanced method enables accurate particle tracking by significantly reducing the number of lost trajectories.

New research reveals a duality between neural network weights and neuron activities that enables a geometric decomposition of the generalization gap. The framework provides a way to interpret the effects of regularization schemes such as stochastic gradient descent and dropout on generalization — and to improve upon these methods.

A framework for training artificial neural networks in physical space allows neuroscientists to build networks that look and function like real brains.

Recommender systems are a predominant feature of online platforms and one of the most widespread applications of artificial intelligence. A new model captures information dynamics driven by algorithmic recommendations and offers ways to ensure that users are exposed to diverse content and information.

New algorithms allow researchers to decode words the brain is hearing with a non-invasive method, outside the scalp.

Efficient quantum-control protocols are required to utilize the full power of quantum computers. A new reinforcement learning approach can realize efficient, robust control of quantum many-body states, promising a practical advance in harnessing present-day quantum technologies.

A ‘programming’-like approach provides a one-step algorithm to find network parameters for recurrent neural networks that can model complex dynamical systems.

An in vitro biological system of cultured brain cells has learned to play Pong. This feat opens up an avenue towards the convergence of biological and machine intelligence.

A new geometric deep learning method can reconstruct cellular and subcellular trajectories and characterize mobility in microscopic imaging, for a broad range of challenging scenarios.

There is a continuing demand for high-quality, large-scale annotated datasets in medical imaging supported by machine learning. A new study investigates the importance of what type of instructions crowdsourced annotators receive.

Predicting whether T cell receptors bind to specific peptides is a challenging problem because most binding examples in the training data involve only a few peptides. A new approach uses meta-learning to improve predictions for binding to peptides for which no or little binding data exists.

Predicting RNA degradation is a fundamental task in designing RNA-based therapeutic agents. Dual crowdsourcing efforts for dataset creation and machine learning were organized to learn biological rules and strategies for predicting RNA stability.

Machine translation of languages can now automatically detect different cell types from single-cell transcriptomic data. Such a feat opens the prospect of dissecting complex clinical samples such as heterogenous tumours at scale.

Antibodies are an essential class of therapeutics but low breadth or off-target binding are major concerns for antibody–drug efficiency and safety. To predict which targets an antibody can neutralize, a machine learning pipeline based on an adaptive graph convolutional network architecture is proposed that learns the binding landscape of antibodies to multiple mutated viruses at the same time.

Tool use is one of the defining traits of human cognition that sets our species apart from other animals. A novel computational framework may enable robots to use tools as intelligently as humans do.

In animals, both body and neural control have co-evolved to be adaptable to the environment. While a newborn foal learns quickly how to use its legs, traditional robotic approaches require careful engineering and calibration for stable walking robots. Bio-inspired robotics aims to bridge this gap.

Designing viable molecular candidates is pivotal to devising low-cost and sustainable storage systems. A reinforcement learning framework has been developed that can identify stable candidates for redox flow batteries in the large search space of organic radicals.