Personalised and personal
“If you can apply AI to healthcare, you can make it work in any sector,” observes Jeroen Maas as an opening to the roundtable ‘Think like a public/private partnership’.
At the Amsterdam Economic Board, Maas has a strategic role in driving innovation and business development in the health sector. He works with leading CEOs, scientists and policymakers to promote data science and AI with the aim to gain two healthy years for every inhabitant of the Amsterdam region by 2025.
Applying AI to any industry is challenging. You need to tackle a whole range of issues, such as ethics, privacy concerns, tech and storage challenges, etcetera.
But healthcare comes with an extra burden: it’s an arena where public and private interests intertwine. So how do you inspire both parties to collaborate in a mutually beneficial way to improve individual patient care and outcomes?
Then there’s yet another layer: health data is also a deeply personal issue. How do we share our bodies?
A smarter workplace or a dumber future?
The roundtable was part of the international event AI@Work, which took place on 5 and 6 March 2020 at Circl in Amsterdam. Organised by Reshaping Work and KIN Center for Digital Innovation, it brought together computer scientists, ethicists, academics, policy makers and business leaders to confront the question: “Are we headed to a smarter workplace, or a dumber future?”
Keynote speakers Marleen Huysman (professor at VU Amsterdam and head of KIN) and Marc Burger (country head at Capgemini Invent) – a public/private duo to be sure – both emphasised how AI will not erase the net number of jobs but create different ones. In fact, as Huysman observed, it’s more likely that there will be a skill shortage.
Another immediate challenge is how to implement more auditing to make sure that AI does not result in profiling, bias or discrimination. “It’s about developing strategies and a robust system of checks and balances that fits our ethical standards,” says Burger.
Stay flexible. Keep learning.
Smart technologies will certainly change team dynamics, observes Huysman. And different stakeholders will need to embrace different strategies. For example, AI developers will have to up their social skills and embrace collaborating with domain experts. Managers will need to focus on making the right organisational changes – and keeping these up to date. And policymakers will need to make sure an infrastructure is in place that promotes tech skills and the idea of life-long learning.
According to Burger, as demand decreases for ‘routine jobs’ such as data entry, retail, telephone service lines and the distribution of goods, the demand will increase for ‘non-routine jobs’ such as entrepreneurs, architects, doctors, designers, builders and baristas.
Regardless, it’s a good idea for everyone to become comfortable dealing with complexity and ambiguity – i.e. the changes will keep changing.
More time for the patient
Various studies back the observations of Huysman and Burger. During the presentation of ‘Manoeuvring uncertainty and urgency in working with AI: the case of radiology’, researcher Bomi Kim noted how radiologists are not suffering any loss of professional identity because AI is now surpassing them in making more accurate and efficient diagnoses.
In fact, when they receive support from management and policymakers, radiologists are even happier since it means they have more time for more complex tasks: namely, providing more personalised patient care and communication.
An inspiring ecosystem
The conference also hosted an expo in collaboration with StartupAmsterdam and Smart Health Amsterdam. These two City of Amsterdam initiatives selected the five most prominent local AI startups: BrainCreators, aigent, Syntho, Clear and Skinvision.
Syntho confronts one of the great challenges of applying data science to the clinical setting. “By creating fully anonymous synthetic datasets, with the same statistical properties and structures as the original dataset, we enable companies and healthcare institutes to use and share this sensitive data without privacy concerns – and thereby help boost innovation,” says cofounder Simon Brouwer.
And Brouwer sees innovation flourishing in Amsterdam: “There’s great symbiosis going on. Healthcare institutes such as Amsterdam UMC and others are really frontrunners when it comes to collaborating with AI startups. We are all strengthening each other.”
Relieving an overstretched system
Loes van Egmond is SkinVision’s business development director and she heartily agrees with Brouwer on the healthiness of the health ecosystem. Her company is seeking to redesign the skin cancer care pathway with an app that allows people to self-check their skin for early signs of skin cancer – to help relieve an already overstretched healthcare system. To scale, they must collaborate with a variety of partners, including care facilities and insurance companies.
“The big challenge for AI in the life sciences is not to refine the algorithms, but to implement them in healthcare,” says Van Egmond. “And Amsterdam is a real hub for AI and healthcare, combining the right talent with top notch institutes and an international network.”
Public/private – and privacy
Meanwhile back at the ‘Think like a public/private partnership’ roundtable, the challenges around applying data science and AI to a clinical setting continue to pile up. Nuance is everywhere.
For instance, the privacy issues that arise from collecting data on age, sex and postal code (since this gives a clue to one’s socio-economic background – a very important factor for health). However with this info alone, you are already getting very close to a patient’s identity.
Then there’s all these other conflicting technologies being created: for example, algorithms already exist that can recreate a person’s face from segmentised MRI scans.
Know your rights
Meanwhile, Europe is trying to develop policies that make personal data shareable while keeping the ownership with the individual – in contrast to China where personal data is state-owned, and to the US where it’s company-owned.
While European experts generally agree that data should never be company owned, we do need the companies to help pay for the infrastructure to keep the data safe and shareable.
But then what happens to the data when a Dutch medtech company gets sold to a US company, which happens to 40% of all such companies?
Indeed, it’s complicated. Meanwhile, the attendees at AI@Work all seem strangely optimistic.
“Maybe it’s just about starting with something simple,” says Maas. “Perhaps we can start with institutionalising reverse Miranda rights: ‘any data you share will not be used against you.’”