Studying tissues, cells, and proteins under a microscope is essential for disease prevention and treatment. This research requires accurately measuring the dimensions of these biological structures. However, when viewed through a light microscope, these samples can sometimes appear more flattened than their true form.
Researchers at Delft University of Technology have now demonstrated for the first time that this distortion is not constant, contrary to what many scientists have assumed for decades. The breakthrough, published in Optica, confirms a prediction by Nobel laureate Stefan Hell from the 90s. With an online calculation tool and software, every researcher can now determine the correct depth of a biological sample.
That led the Microsoft Research machine learning expert to wonder how much an AI model could learn using only words a 4-year-old could understand – and ultimately to an innovative training approach that’s produced a new class of more capable small language models that promises to make AI more accessible to more people.
Large language models (LLMs) have created exciting new opportunities to be more productive and creative using AI. But their size means they can require significant computing resources to operate.
While those models will still be the gold standard for solving many types of complex tasks, Microsoft has been developing a series of small language models (SLMs) that offer many of the same capabilities found in LLMs but are smaller in size and are trained on smaller amounts of data.
SpaceX is making rapid progress in the development of their Starship, with improvements in heat shield tiles, construction of a second launch tower, and multiple successful launches, showcasing their commitment to innovation and progress in space exploration.
Questions to inspire discussion.
What progress has SpaceX made with their Starship development?
—SpaceX has made rapid progress with improvements in heat shield tiles, construction of a second launch tower, and multiple successful launches.
Google DeepMind has recently introduced Penzai, a new JAX library that has the potential to transform the way researchers construct, visualize, and alter neural networks. This innovative tool is designed to smoothly integrate with Google Colab and the JAX ecosystem, which is a major step forward in the accessibility and manipulability of AI models.
Penzai is a new approach to neural network development that emphasizes transparency and functionality. It allows users to view and edit models as legible pytree data structures, making it easier than ever to delve into the inner workings of a model. This feature is especially useful after a model has been trained, as it provides insights into how the model operates and allows for modifications that can help achieve desired outcomes.
Penzai aims to make AI research more accessible to researchers by simplifying the process of modifying pre-trained neural networks. This would enable a wider range of researchers to experiment and innovate on existing AI technologies, which is crucial for advancing the field and discovering new AI applications. Penzai’s user-friendly interface breaks down the barriers to AI research and makes it easier for everyone to benefit from the technology.
The Automotive in the Software-Driven Era community, part of the DRIVE-A initiative, has identified six key actions crucial for unlocking the full potential of smart road transport infrastructure and, with that, for reaping the potential benefits of the software-defined-vehicle ecosystem. These six actions are based on three principles, collaboration, innovation and efficiency:
Advancing smart infrastructure requires a collaborative effort among involved players. Figure 2, below, provides a collaboration framework and key collaborative areas. Public-private partnerships help build a robust, cohesive and inclusive smart infrastructure network.
Paul Farrell, EVP and Chief Strategy Officer of BorgWarner highlights some of the challenges and importance of such collaborations: “Collaboration around the infrastructure is very complex and multidisciplinary. It involves questions about who owns the data, who supplies the hardware, who installs it and who integrates it into holistic solutions. Yet, it is key to, for example, advance the EV charging infrastructure that the transition to electrification requires.”
Researchers at Stanford University and Biogen faced skeptics head on, challenging the notion that monoclonal antibodies would never serve as therapeutic agents.
Read more about the journey of resilience and innovation:
The road to developing monoclonal antibodies for effectively targeting cancer was paved with tenacity, passion, and strokes of luck.