AI is rapidly becoming an indispensable tool for healthcare researchers, but can it take another step and become a bona fide teammate?

That’s the question being asked at the UK’s University of Manchester, and the first step toward answering it has been the appointment of Prof. Sami Kaski as one of the first Turing AI World-Leading Research Fellows.

Along with four colleagues from the universities of Cambridge, Edinburgh and Oxford, Prof. Kaski’s initial goal is to break beyond “a fundamental limitation of current AI systems, that they require a detailed specification of the goal before they can help.”

Put another way, for all its potential in research and complex decision making, AI tools still need human guidance to set objectives and tell AI systems which outcomes are desired.

The goal for the new Turing fellows, then, is “to develop new ways for machine learning systems to help humans in the process of designing experiments and interpreting what results mean, before deciding what to measure next, and to finally reach trustworthy solutions to problems. In lung cancer personalised medicine, for instance, to maximize effectiveness of radiotherapy for a new patient while minimising side effects, doctors need to combine their expertise and what can be learned from measurements from earlier patients.”

What is called for, said Prof. Kaski, are “new kinds of AI assistants which can learn to work well with humans and complement their skills. That requires new fundamental AI research, and I am glad Manchester has recognized this opportunity and is considerably strengthening its AI research.”

As part of this new AI driven approach The University of Manchester has also received a share of £4.4 million research funding from the UK Research and Innovation organization (UKRI), a non-departmental public body sponsored by the UK’s Department for Business, Energy and Industrial Strategy.

Another example cited in the announcement is drug design, where AI can identify specific “candidate molecules,” but only if a precise objective function has been specified.  Less than perfect identification, however, can result useless and unwanted results.

Kaski and co. will be applying their new approach to three particular challenges: “diagnosis and treatment decision-making in personalised medicine; the guidance of scientific experiments in synthetic biology and drug design; and the design and use of ‘digital twins’ to design physical systems and processes.”

Digital twins are a virtual representation of a complex objects or systems, which “can be built for patients for personalised medicine, but also for physical systems, such as complex buildings, a farm and even a city.”

Photo by Ekkasit919/Getty Images