What is AI?- Current and Future Relevance to Human and Veterinary Biomedical Sciences

DataHub Director Kevin Wells Presents:

Introduction

Artificial Intelligence (AI) is revolutionizing various fields of human endeavour, including biomedical sciences, through its unique ability to ingest, analyse, unlock insight and assist in decision-making. AI refers to the development of computer systems capable of performing tasks that typically require human intelligence, such as learning from data, recognizing patterns, and making decisions. The key difference with AI compared to a standard piece of software is that the solutions AI generates are not pre-programmed. In essence, AI produces outputs that are the result of codifying a set of rules needed to solve a problem (e.g. perceiving the key features to solve a clinical case, and having reasoning capability to deduce the correct diagnosis), but the actual specific, particular problem (e.g. a specific dog presented in a practice) has not been seen before.

Types of AI

AI can be broadly classified into two major categories based on capabilities and scope:

General AI
Also known as Strong AI, represents the theoretical concept of AI systems that possess human-like intelligence and can autonomously perform ‘any’ intellectual task that a human can. Often referred to as AGI, or Artificial General Intelligence, there are arguments about how exactly to measure success with any AGI test, akin to discussions around Turing’s famous Turing Test. However, Transformer based ChatBot such as GPT-4 are edging ever closer to a performance that is likely to be accepted as AGI.

Machine Learning, Deep Learning and Transformers

Machine Learning (ML) is a subset of AI that focuses on enabling machines to learn from data and improve performance without being explicitly programmed. ML algorithms can be trained on large datasets to recognize patterns, make predictions, and adapt to new information, making them invaluable tools in biomedical research and healthcare.

Machine learning has made great strides in recent years in achieving human-level and better-than-human performance in a number of different tasks. This has been facilitated by the development of deep learning – essentially networks of layered perceptrons, the fundamental building blocks of artificial neural networks. Whilst deep learning networks are not new their recent successes have largely been due to the ability to train on large datasets, and the availability of low-cost GPU (General Processor Unit, previously known as Graphics Processor Unit) cards, the latter thanks to the development of computer games, and its demand for ever increasing graphics performance.

However, it was the Transformer architecture, with its in-built attention mechanism that has shown so much promise in the last two years. This has produced phenomenal human-like conversational ChatBots, as well as in other domains (e.g. Vision transformers for images/video analysis).

Relevance & Applications of aI in human and veterinary biomedical sciences

In the context of biomedical sciences, AI has the potential to impact on all areas of biomedical activity, from fundamental drug discovery, diagnosis, prescription and personalized medicine. A key point is that the way machines can detect and comprehend patterns is often different from the way humans perceive a problem and its solution. This allows machines to potentially make new discoveries, and moreover to absorb such vast amounts of information that would be difficult for a human to comprehend.

Similarly, veterinary medicine stands to benefit from AI technologies, with applications ranging from disease diagnosis and treatment planning to genetic screening and population health management, mirroring many of the opportunities for patient benefit seen in human healthcare.

Within disease diagnosis and prognosis, AI can be used to analyse diagnostic images (e.g. planar 2D X-rays, MRI, and CT scans), to assist in the early detection and diagnosis of diseases). Predictive models based on AI can also estimate disease progression in oncology with analysis for grading tumours in digital pathology and prognosis.

In the near future we might expect the arrival AI-powered platforms that can analyse vast amounts of multi-modality data to advance precision medicine spanning patient medical history, lifestyle, breed, as well as multi-omics data. This will facilitate targeted therapies tailored to each patient’s unique needs and so optimise curative outcome.

However, a major question for all professions impacted by AI surrounds the potential for replacement of human professionals with synthetic/robotic alternatives. Whilst fully automated robotic practices may be many decades away, it is clear that the veterinary profession will need to adapt to the opportunities presented and the new ways of working that will be presented.

Those who do not embrace the coming change will be replaced by those that do.

Conclusions

As AI continues to demonstrate increasing impact on human and veterinary biomedical sciences, we can expect the pace of innovation to increase, as well as the opportunity for enhanced therapeutic outcomes. From disease diagnosis to personalized medicine in veterinary care, AI-driven innovations hold the promise of improving outcomes, advancing knowledge discovery, and ultimately enhancing the well-being of both human and animal subjects. And whilst AI is unlikely to replace veterinary professionals, it is clear that those who do not embrace the coming change will be replaced by those that do.