The powers and perils of AI in transforming human society and public health
By Henry Li | Twitter: @henrylsl
In his epic Also Sprach Zarathustra, the philosopher Nietzsche posits a three-step journey of transformation and transcendence of the human spirit: first as a resilient camel that travails, then as a ruthless lion that conquers, and finally as an innocent child that creates. This process of metamorphosis can be applied to understand the co-advancement of humans and machines, and the policies that govern the constantly evolving interactions between human and artificial intelligence (AI).
In this light, the powers and perils of AI in transforming human society were discussed at AI and Public Health, part of the Innovation and the Welfare State public lecture series from the UCL Institute for Innovation and Public Purpose (IIPP) and the British Library. The core concern — as discussed by IIPP deputy director Rainer Kattel, AI expert and IIPP visiting professor Ian Hogarth and public health expert Carol Sinclair — was one of public value. Capitalism is good at learning, but it creates inequality when left to its own device. With AI becoming an increasingly effective means to automate learning, how can society better shape AI for public value?
The recent rise of AI has been meteoric in capturing the ambition of the field and the imagination of the public, but the workman-like practical basis of the technology — machine learning — is more modest in funtionality and less glamorous in character. Essentially, it is a process in which a computer learns to discriminate patterns in a narrowly defined task. As the foundation of AI, machine learning is akin to a heavily loaded camel, bearing the weight of extremely intensive tasks and of exceptionally high expectation.
But the camel is ready to morph into a lion, a technology that has the potential to revolutionise health. Describing the phenomenal progress made in the field over the past two decades, Hogarth observed that machine learning is ready for prime time. Computer vision has now surpassed human experts in the accuracy of diagnosis in some areas. Brain signal translation leads to breakthrough in recovering lost physical capacities. Deep learning, a form of enhanced learning based on artificial neural networks, changes the ways medical and pharmaceutical innovations are developed and applied. Transfer learning, a process that make use of general datasets to create learning that can be applied to specific problems, greatly speeds up the rate of innovation.
The exponential potential for machine learning and AI in health extends well beyond the realm of medicine. Drawing on her experience in developing the AI mission for the Mission-Oriented Innovation and Industrial Strategy (MOIIS), Sinclair expected AI to be central to informing service planning, public health policy and shaping behavioural change. But AI per se is not the solution to health system challenges, which are ultimately characterised by the difficult balance between steep demand and constraints in resources. Here, AI can augment human decision making for better resource allocation and outcome delivery. After all, healthcare is delivered by humans for humans.
But is the rise of AI’s capability met by commensurate policies that can maximise and protect public value creation, unleashing its lionlike potential at the same time as reining in its claws and fangs? These policies are especially critical because of the dominance of the private sector in the field. Once a powerhouse that created monumental public goods projects like the UK Biobank and the Human Genome Project, the public sector has now been far surpassed by the private: Hogarth’s State of AI Report 2019 shows that most of the nearly $30bn going into AI every year are from private sources.
The panel recognised that the current trajectory of AI policies does not match that of the technology and is wanting in three broad areas: data transparency, public return on public investment, and mission-oriented industrial strategy.
AI is an Apollonian mode of reasoning bounded by its algorithms and the dataset it can draw on. However, citizens are not privy to the contents of the black box, nor its inherent incompleteness and biases. Better governance and regulation to ensure AI promote equality rather than reinforcing existing social biases at the population level strongly rely on increasing transparency, citizen engagement and stakeholder consultation.
Public return on public investment
It is essential for public sector leaders to put public return at the heart of AI implementation. Sinclair noted that too often the collective nature of value creation in the use of AI is overlooked. Navigating commercial competitiveness and collaborative spaces in AI for all actors to deliver public return requires the public sector to take lead in ascertaining where values lie, creating a multidisciplinary environment for all the necessary expertise (including clinicians, lawyers and IT experts), and identifying efficient ways to obtain or procure their service.
Mission-oriented industrial strategy
Hogarth and Sinclair agreed that without a wholistic mission-oriented AI strategy aimed at solving societal problems, political and financial commitments would be piecemeal and ineffectual. Hogarth contended that UK has all the potential to be a front runner; however, existing thinking in AI policies lacks both the supportive capacity of a camel and the visionary ambition of a lion. Political leadership should try to understand the macro challenges to mobilise resources for AI as a strategic industry, rather than dwell on micro decision making such as the number of AI PhDs to fund each year.
AI, like all revolutionary technologies before it, will play a crucial role in humankind’s transformation from Nietzche’s enduring camel, conquering lion to innocent child — an unaffected and inspired spirit eager to play and experiment. In many respects, the technology is ready, but not the political will. Creating the political capital for change requires raising public awareness and interests. Hogarth and Sinclair called for the public to question the ownership of data, challenge how money is managed and assigned in the public sector and create popular pressure for politicians to act. What would it mean for us as a species and our society when AI and its policies eventually play and dance to the tunes of public value?
Henry is a Research Associate in Health Innovation and Policy Engagement at IIPP.
Sign up to the UCL Institute for Innovation and Public Purpose’s mailing list to hear about our latest research, news and events. You can also follow us on Twitter: @IIPP_UCL.