F052 AI in healthcare 6/6: What if AI gets out of control? (Bart De Witte)
In the final episode of a series about AI in healthcare, Bart De Witte discusses data privacy, the future of AI models in healthcare. If we decide to let monetization of healthcare data get out of control, a dystopian future lies ahead.
When social networks came into existence, it wasn’t immediately obvious what could be the negative consequences of connecting people. Brexit and the US elections showed the power of data gathering and the influence algorithms can have on our thinking. “This is why we need to be careful who we trust with our data,” says Bart De Witte, a digital health expert with longterm experience in executive functions with global vendors such as SAP and IBM, and the founder of HIPPO AI - an NGO that focusses on Open Sourcing Medical AI.
His personal disappointment regarding data management from large enterprises happened when he purchased a 23andme genetic test kit over a decade ago. He felt his and other customers data was being exploited when the company changed its business model and started selling data to third-party providers.
Patients should not be able to sell their data
Healthcare blockchain projects popularized the idea that patients should own and be able to sell their medical data. Bart De Witte warns this could lead to monopolization of data, and therefore, data should become a public good. “When Facebook bought Whatsapp, it became clear that data has value. If you look at healthcare data, you can ask yourself, what is healthcare data in the first place? Each dataset defines people, human life. As a society, we need to ask ourselves if we want to capitalize on human life? Do we want to allow people to sell their data? We don’t allow people to sell their organs. With data sales, slowly, each molecule in your body could have a capital value, which is a dystopian direction. I believe data should be free. In the future, when knowledge will be dependant on data, privatization could occur in knowledge access, which is why data should be a public good.”
If we start putting price tags on our data, it could quickly happen that data would get accumulate in the hands of those with buying power, hindering research opportunities of public research institutions.
Open data and data standards
He also clarifies what open data is or is not: “Open data does not mean we open our data to everyone. But open datasets do accelerate AI development and make it accessible to all - African innovator or Danish innovator,” explains Bart De Witte.
He is critical of the current state of medical data structures and availability: “The problem is that even if in theory same standards are used across countries, each country is developing different specifics of standards. Additionally, every doctor has his own practice; we have guidelines but lack of standardization in terms of data. Furthermore, genomic data is a special category because it cannot be anonymized since genetic traits are linked to an individual.” Clinical data is under a lot of challenges, but machine learning could overcome the issue of data heterogeneity, says De Witte. Not everything needs to be standardized, however. Machine learning has the additional potential to solve the problem of data centralization. With federated learning (born at the intersection of on-device AI, blockchain, and edge computing/IoT), data could be analyzed in silos, and only learnings would be stored centrally.
Follow less, DO more
In the end, the discussion resolves around societal values. Europe is often criticized as rigid due to tough legislations, which are at the moment, hindering data aggregation and slowing down scientific research. Regardless, says De Witte, we should not sell our values because we would fall behind the rest of the world. “Privacy is something our predecessors fought for. Loss of privacy brings loss of freedom and problematic consequences as Germany has learned in the past.”
The problem in this discussion is that older people, which are in power, are much less aware and cautious about data privacy in real life, compared to younger generations. To change the future, we need to act today. “Looking at the future as utopian or dystopian is a passive approach. I believe in Do-topia - we need to DO things in order to influence the future. Technology in the end, is not doing anything. It is the people behind the technology that give technology purposes. Europe tends to be behind is adoption due to our eternal skepticism. In the end, we use technology developed by the rest of the world and miss the opportunity to shape our future. We should be more doers than adopters.”
Hear the full discussion below or in iTunes.
Some questions addressed:
- End of 2017, Stephen Hawking said AI could be the worst event for humankind. a few months later Elon Musk said AI is more dangerous than nuclear weapon. Where do you stand on the question of the danger of AI?
- You are the founder of HIPPO AI, an NGO that focusses on Open Sourcing Medical AI. What is and what isn’t open data?
- Could we bridge interoperability with open data standards?
- How can data standardization in healthcare be improved?
- 10 years ago there was less awareness about where data gathering is going. Today data is the new oil. Should we profit out of our data?
- What about patients, with high healthcare costs - wouldn’t it be fair for them to be able to monetize their data and lower their healthcare costs?
- Privacy debates: does privacy hinder innovation in healthcare? Will Europe fall behind in development?