Engineering and Technology
Explore Engineering and Technology
GenBench: Mapping out the Landscape of Generalization Research
From ChatGPT to Google Gemini, large language models are now increasingly important in our everyday lives. These models are part of the field of natural language processing – or ‘NLP’ – which studies how machines understand and generate human language. Most NLP systems are built using machine learning, and vast amounts of language data are used as training material. Afterwards, a successfully trained model should be able to handle new scenarios. This ability is called ‘generalization’. For large language models that generalize well, a conversation about a topic it hasn’t been trained on, such as new scientific discoveries, should not be a problem.
Professor Erik Folven | Gaining Tighter Control Over Artificial Spin Ices for Technological Applications
The energy demands required to operate our increasingly connected society is unsustainable. Data-centres, the Internet of Things, and Artificial Intelligence all create unsustainable power demands. As such, new energy-efficient technologies are needed to fulfil humanity’s computing needs into the future. Spin-based technologies could greatly reduce the energy requirements of computing. One reason that such technologies are more energy efficient is that far less energy is wasted as heat. Additionally, computation and memory can both occur within the same spin-based device. One promising spin-based technology is artificial spin ice (ASI).
Dr Martin Haesemeyer | A Deep-learning Framework that Links Brain Activity with Behavior
Neuroscientists have been trying to uncover the relationships between brain activity and behavior for decades. Identifying these links could shed new light on the functions of different brain regions, while also highlighting possible therapeutic targets for psychological disorders. In recent years, researchers have gathered a vast amount of brain activity recordings alongside data describing the behavior of animals or humans while these recordings were collected. Recent advances in deep-learning algorithms have now opened new possibilities for analyzing this wide pool of data.
Dr Elif Miskioğlu – Dr Kaela Martin – Dr Adam Carberry | Fostering Intuition in Engineering Students to Solve 21st Century Challenges
Experienced engineers typically have advanced technical knowledge and unique skillsets. Many also develop impressive intuition through years of experience, which helps them to devise solutions to complex real-world problems. Cultivating such intuition in engineering students could better equip them to tackle humanity’s increasingly complex challenges. Before we can design classroom interventions that foster intuition in prospective engineers, we need methods that can reliably assess intuition. Using such methods, the effectiveness of a given intervention could be measured by assessing students’ intuition before and after they take part.
Damian Nowak – Adam Bachorz – Professor Marcin Hoffmann | Using Machine Learning to Discover New Medicines
Many researchers today are dedicated to the discovery of new medicines. Over the past few decades, their tireless efforts have culminated in a database of around 100 million known drug molecules. This value may already sound vast, but by current estimates, the true number of small drug-like molecules could actually range anywhere between 1023 – already more than the number of grains of sand on Earth – and 1060 – comparable to the number of atoms in an entire galaxy. With existing approaches, researchers ultimately need to test the medical potential of these molecules individually, taking up vast amounts of time and computing power.
Amit Kumar | A Circular Economy for Electric Vehicle Batteries
Electric vehicles – or EVs – are a cornerstone for curbing greenhouse gas emissions. They’re far better for the environment than their fossil-fuel-burning counterparts, but there’s one major catch: the toxic materials found in their batteries still present a serious environmental risk. So, how can we address this? By embracing a circular economy for EV batteries! In a circular economy, used batteries are repurposed for a second life, or are recycled to recover strategic raw materials, which can be processed further to manufacture new batteries. But before we get there, there are a number of hurdles to overcome.
Dr Alvin Orbaek White | Advancing Energy Technologies with Ultralong Carbon Nanotubes
Carbon nanotubes are a unique family of molecules that look like tiny carbon tunnels with honeycomb walls. They are typically around 1 nanometre in diameter and several nanometres long. In principle, carbon nanotubes are much more efficient at conducting electricity than metals, especially since they are so lightweight and have high-temperature stability compared to metals. So far, however, their full potential has been limited by the techniques used to manufacture them.
Dr Aaron Tallman | Assessing Uncertainties When Measuring 3D Printed Metal Parts
The ability to 3D print metal parts presents exciting opportunities to simplify the designs of many advanced technologies, and improve their performance. However, on microscopic scales, printed metals can have defects that cause their mechanical properties to vary unpredictably, lowering the quality of final products. To assess these variations, researchers use a technique named profilometry-based indentation plastometry, or PIP. This technique involves pressing a hard tip into a material on a flat surface, and then scanning a probe across the crater to measure the shape left behind.
Dr Yujeong Bae | Advancing Scanning Tunnelling Microscopy for Quantum Information Processing
The ability to manipulate single atoms and molecules would transform how we store and process digital information. This can be achieved using a cutting-edge technique named scanning tunnelling microscopy. Scanning tunnelling microscopes (STMs) are powerful imaging devices, which operate by holding a sharp metal tip less than one nanometre above a conducting sample. Through the effects of quantum tunnelling, electrons can pass through the tiny vacuum gap between the tip and the sample surface.
Dr Thomas Shaffer – Dr Tariq Rahman | pneuRIP: An Innovative New Technology to Monitor Children’s Breathing
Respiratory inductive plethysmography (RIP) is a technique used to monitor a patient’s breathing patterns. It is used for diagnosing and monitoring children with lung disease, and assessing the effectiveness of therapies. Using bands placed around the torso, the technique measures volume changes in the abdomen and ribcage. These measurements are then translated into useful indicators of lung health. Dr Thomas Shaffer and Dr Tariq Rahman at Nemours Children’s Hospital in Delaware have developed a new technology called pneuRIPTM, which allows continuous, real-time monitoring using RIP.
Professor Jozina de Graaf | Improving Outcomes Following Lower Limb Amputation
After the amputation of a lower limb, amputees can learn to walk with an artificial replacement for that limb known as a prosthesis. This can be challenging, however, due to the loss of somatosensory information such as the perception of touch and pressure coming from the foot. For the majority of amputees, their lost limb can still be perceived through a phenomenon known as phantom limb, in which a painless tingling or a warm sensation is often felt where the limb used to be.
Sabine Grüner-Lempart – Julian Eckert | Removing Industrial Pollution with Bacteria
Harmful chemicals are commonplace in many different industries. Volatile organic compounds – or ‘VOCs’ – represent one such type of chemicals, which are particularly prevalent in industries that require spraying of paints and coatings. Unfortunately, VOCs can readily evaporate into the air, potentially harming people’s health through inhalation. Some VOCs are also environmental pollutants and can even contribute to climate change.
Stay Up To Date With SciTube