It’s easy to marvel at the technical wizardry behind breakthroughs such as AlphaFold.
Pioneering crystallographer Helen Berman helped to set up the massive collection of protein structures that underpins the Nobel-prize-winning tool’s success.
Boston Dynamics and Toyota Research Institute (TRI) Wednesday revealed plans to bring AI-based robotic intelligence to the electric Atlas humanoid robot. The collaboration will leverage the work that TRI has done around large behavior models (LBMs), which operate along similar lines as the more familiar large language models (LLMs) behind platforms like ChatGPT.
Last September, TechCrunch paid a visit to TRI’s Bay Area campus for a closer look at the institute’s work on robot learning. In research revealed at last year’s Disrupt conference, institute head Gill Pratt explained how the lab has been able to get robots to 90% accuracy when performing household tasks like flipping pancakes through overnight training.
“In machine learning, up until quite recently there was a tradeoff, where it works, but you need millions of training cases,” Pratt explained at the time. “When you’re doing physical things, you don’t have time for that many, and the machine will break down before you get to 10,000. Now it seems that we need dozens. The reason for the dozens is that we need to have some diversity in the training cases. But in some cases, it’s less.”
Boohoo Group Plc Chief Executive Officer John Lyttle is stepping down as the struggling online retailer begins a strategic review that could potentially lead to a breakup of the company.
Would you like to see more applications for Neuralink in the future? Share your thoughts in the comments.
Elon Musk’s brain technology startup, Neuralink, reported that its implant is functioning well in a second trial patient, identified as Alex. This implant is designed to help paralyzed patients control digital devices through thought alone. Unlike the first patient, Noland Arbaugh, who experienced thread retraction issues post-surgery, Alex has not faced similar problems. Neuralink implemented new measures to prevent such complications, including reducing brain motion during surgery. Both patients have been able to use the implant to perform tasks like playing video games, browsing the internet, and even designing 3D objects.
A team of physicists from the universities of Amsterdam, Princeton and Oxford have shown that extremely light particles known as axions may occur in large clouds around neutron stars. These axions could form an explanation for the elusive dark matter that cosmologists search for—and moreover, they might not be too difficult to observe.
The head of Singapore’s biggest lender said only about half of the banking industry has made sufficient progress in transforming their businesses to embrace digitalization and artificial intelligence.
Humanity could use a nuclear bomb to deflect a massive, life-threatening asteroid hurtling towards Earth in the future, according to scientists who tested the theory in the laboratory by blasting X-rays at a marble-sized ‘mock asteroid’
Women worldwide could see better treatment with new AI technology, which enables better detection of damaged cells and more precisely predicts the risk of getting breast cancer, shows new research from the University of Copenhagen.
Breast cancer is one of the most common types of cancer. In 2022, the disease caused 670,000 deaths worldwide. Now, a new study from the University of Copenhagen shows that AI can help women with improved treatment by scanning for irregular-looking cells to give better risk assessment.
The study, published in The Lancet Digital Health, found that the AI technology was far better at predicting the risk of cancer than current clinical benchmarks for breast cancer risk assessment.
Exposure to certain pollutants, like fine particles (PM2.5) and nitrogen oxides (NOx), during pregnancy and childhood is associated with differences in the microstructure of the brain´s white matter, and some of these effects persist throughout adolescence. These are the main conclusions of a study led by the Barcelona Institute for Global Health (ISGlobal), a centre supported by “la Caixa” Foundation. The findings, published in Environmental Research, highlight the importance of addressing air pollution as a public health issue, particularly for pregnant women and children.
An increasing amount of evidence suggests that air pollution affects neurodevelopment in children. Recent studies using imaging techniques have looked at the impact of air pollutants on the brain’s white matter, which plays a crucial role in connecting different brain regions. However, these studies were limited in that they only looked at one timepoint and did not follow the participants throughout childhood.
“Following participants throughout childhood and including two neuroimaging assessments for each child would shed new light on whether the effects of air pollution on white matter persist, attenuate, or worsen,” says ISGlobal researcher Mònica Guxens. And that is what she and her team did.
Multimodal AI models are powerful tools capable of both understanding and generating visual content. However, existing approaches often use a single visual encoder for both tasks, which leads to suboptimal performance due to the fundamentally different requirements of understanding and generation. Understanding requires high-level semantic abstraction, while generation focuses on local details and global consistency. This mismatch results in conflicts that limit the overall efficiency and accuracy of the model.
Researchers from DeepSeek-AI, the University of Hong Kong, and Peking University propose Janus, a novel autoregressive framework that unifies multimodal understanding and generation by employing two distinct visual encoding pathways. Unlike prior models that use a single encoder, Janus introduces a specialized pathway for each task, both of which are processed through a unified transformer. This unique design alleviates conflicts inherent in prior models and provides enhanced flexibility, enabling different encoding methods that best suit each modality. The name “Janus” aptly represents this duality, much like the Roman god, with two faces representing transitions and coexistence.
The architecture of Janus consists of two main components: an Understanding Encoder and a Generation Encoder, each tasked with handling multimodal inputs differently. For multimodal understanding, Janus uses a high-dimensional semantic feature extraction approach through SigLIP, transforming the features into a sequence compatible with the language model. For visual generation, Janus utilizes a VQ tokenizer that converts visual data into discrete representations, enabling detailed image synthesis. Both tasks are processed by a shared transformer, enabling the model to operate in an autoregressive fashion. This approach allows the model to decouple the requirements of each visual task, simplifying implementation and improving scalability.
