Areas of innovation

Core AI

Core AI relates to the fundamental technologies and principles underlying artificial intelligence. These technologies form building blocks from which sophisticated AI systems can be built. When it comes to driving forwards innovation in AI technologies, an important factor is the refinement of core AI components such as deep learning architectures and improved neural network models, to enhance efficiency, accuracy, and adaptability over a diverse range of real-world applications.  

Bioinformatics

Healthcare is changing beyond recognition, in part driven by the fact most of us have powerful computers in our pockets, on our wrists, and in our homes. Digital health apps are being developed to leverage this capacity, to make patient monitoring, assessment, cheaper, and more efficient.

Modern AI is very good at picking out patterns in images, some of which even escape the human eye. One of the many areas where this has shown promise is in image analysis (PET, radio, MRI,) for diagnostics and prognostics purposes, e.g. to identify and monitor the growth of tumours from scans, to identify signs of diseases (e.g. nodules caused by diseases such as pneumonia in chest x-rays), etc.

The availability of large amounts of omics data (typically along the central dogma, i.e. genomics, transcriptomics, proteomics and combinations thereof) has opened new possibilities for diagnostics and prognostics, enabling patient stratification (the identification of subgroups of patients, such as subgroups that are likely or unlikely to benefit from specific therapies or subgroups with different prognosis, etc.), increasing diagnostic accuracy, reducing the need for invasive tests for e.g. cancer and pre-natal tests, etc. This underlines much of the precision medicine revolution.

Digital health

AI provides significant time and therefore cost savings across a number of medical imaging and diagnostic technologies. The capability of AI to carry out repetitive and large volumes of analysis of data can be utilised to and improve the accuracy and speed of diagnosis. For example, AI is being used to detect and highlight tissue abnormalities in radiology scans allowing radiologists to prioritise their time and focus on critical patient scans. Even simple blood tests can be better processed using AI where machine learning algorithms can be used to predict diseases and medical conditions with far greater accuracy than traditional methods.

AI coupled with telehealth is driving a shift in diagnostic testing currently carried in the clinical setting into the patient’s home with AI facilitating the early detection of data anomalies leading to earlier diagnosis of medical conditions with an associated improved patient outcome. For example, diagnosis of epilepsy in a clinical setting is typically time-consuming and costly. However, connected, wearable devices can be used to generate data for AI analysis which can be used to detect and predict epileptic seizures and to monitor response to treatment.

Cheminformatics

Cheminformatics (or chemoinformatics) is an emerging term for the use of computational methods and information technology to analyse, manage, and interpret chemical data.  

In the context of drug discovery, cheminformatics involves the analysis of chemical structures, properties and activities of compounds to identify potential drug candidates. It utilises computational tools to predict the bioactivity of molecules, understand structure-activity relationships, and streamline the process of designing new pharmaceuticals.  

In the field of material science, cheminformatics accelerates the materials discovery process, providing valuable insights, optimising properties, and facilitating the design of materials with tailored functionalities for specific applications.  

Robotics

Advances in AI are driving developments within the field of robotics in a variety of ways, contributing to the evolution of more intelligent, more capable robotic systems and impacting industries ranging from manufacturing to healthcare and beyond.  

Machine learning enables robots to improve their performance over time, making them adaptable to new tasks and scenarios. For example, through deep learning algorithms, robots can achieve advanced functionalities such as object recognition, allowing them to identify and interact with objects within their environment with a greater precision.  

Another important AI application for robotics relates to path planning and navigation. AI-powered algorithms enable robots to analyse their surroundings, select optimal paths, and navigate through complex environments. This can be particularly advantageous in relation to logistics and manufacturing uses, where robots need to move efficiently and safely, often within dynamic spaces.  

A final example relates to AI enhancement of a robot’s ability to interpret and respond to sensory input. Robots can process information from various sensors to understand their surroundings. Artificial intelligence facilitates improved decision making in real time, enabling robots to interact more effectively with the world around them.  

Generative AI

Generative AI is a class of artificial intelligence involving the creation of new content based on patterns learned from pre-existing data.  

In the field of natural language processing, generative AI can be utilised for a range of tasks including: translation, summarising material, and even text generation. Models such as OpenAI’s GPT (Generative Pre-trained Transformer) demonstrate the power of large scale generative models. These models are trained on large volumes of diverse textual data, giving rise to the ability to generate coherent and contextually relevant responses.