A couple of weeks ago I was privileged to attend this year’s excellent virtual Tri-Livery Roundtable, along with 200 other interested people. Given our reliance on technology this year, the topic seemed appropriate. So much of our daily considerations around Artificial Intelligence seem to hinge on the luxuries of a talking fridge that can top up your Tesco order or applications in media and entertainment. Unusually, this was an exploration of how AI has the capacity to deliver economic and social good. As a Stationer, Past Master Marketor and Chair of the Financial Services Group of Livery Companies, of which the Information Technologists are members, you could not have found a more interested listener.
The definition of Artificial Intelligence
We hear so much about how artificial intelligence (AI) will change the way we work in the future; how it will cause structural unemployment amongst repetitive professional jobs and of those in low skilled areas. However, the discussion panel, expertly chaired by Marketor Roz Morris, explored the reality that AI is not only already well established in our lives but is more often than not a force for good.
Often, it seems that the definition of AI becomes muddied by lumping it together with other forms of technology that do not really constitute artificial intelligence. So, to begin the debate, Maxine Ricketts, Chair of ‘AI4C’ created by the WCIT to help charities, gave us a simple working definition:
’Any computer system that performs tasks previously performed by humans’ usually by using a big data reference table.’
Charity support, cancer diagnosis and tackling crime
Ms Ricketts told us that charities are already triaging incoming calls using chatbots to answer simple questions and focus enquiries, getting to deeper issues quicker and amassing data faster than humans.
Meanwhile, Dr Christina Messiou, the distinguished radiological researcher, gave a presentation that showed how AI is driving forward innovative imaging for the benefit of oncology patients. She showed how Myeloma patients (bone marrow cancer) are benefitting from whole body MRI imaging to accurately detect the disease. The process is non-invasive, with no injections or radiation and reduces the time from 30 to 2.5 minutes.
Dr Messiou reminded us however, of the significant shortage of both qualified people and investment, with only 1/10 NHS treatment centres currently able to deliver this analysis. This emerging form of ‘Precision Medicine’, using algorithms to measure disease beyond that possible by human experience, will need a digital ready workforce – informaticians, data scientists and medically qualified programmers.
Jonathan Sinclair, of Bristol Myers Squibb, told us how organised crime is ever more active. That means that AI-driven cyber security is required to deliver the scale, speed and high levels of accuracy necessary to defend our networks and drive the global information security programmes now required.
Ben Gancz a former Met Police/NCA detective who specialised in child protection, which is both psychologically demanding and repetitive work for humans, then told us about the ‘human-in-the-loop’ AI systems and automatic image classifiers that he has developed to detect indecent images of children. They have the benefit of not suffering fatigue, of getting ‘accustomed’ to images or suffering psychological harm, but can automatically look at millions of images in support of the ultimate human decision.
Raising questions about AI
As is often the case, the question and answer session which followed the presentations, was revealing. AI, more accurately referred to as ‘machine learning’ or ‘scale computing’, given that machines are not ‘conscious’ (except in science fiction) will unquestionably deliver improvements and efficiencies in our lives, but naturally it does raise some profound societal questions.
The use of AI will require the digital ‘upskilling’ of existing workforces, or their reallocation towards more added value roles rather than repetitive tasks. A lot of junior staff roles will disappear over time. However, many new job titles will emerge, especially those related to the training and validating of the algorithms used. Of the need to avoid data riddled with unconscious biases, detecting and removing obtrusive data and moving some current ‘black box’ solutions to more ‘white box’ transparent ones.
Then there are the issues surrounding the collation of mass data points whilst ensuring personal privacy by anonymising the data. Legislation, even in local geographies, is always slow to catch up, and who will regulate this emerging global phenomenon is as yet unclear.
As ever, we were reminded that the key issue with any computer system will remain the need to clearly define the problem we are actually trying to solve. Ben Gancz concluded the evening by reassuring us that all the research still shows that ‘people do like speaking to another human being’, and so we can view AI as something that will change the way we work, and hopefully help support more individuals in the most important and possibly challenging areas of their lives, but not obliterate the need for human skills either.