I want to share some reactions that I had related to the article -“Why big Pharma sees a remedy in data and AI” by Sarah Neville published on January 26th, 2020 in the Financial Times- taking into account that it is a subject with important complexity on various fronts, from regulations to biology.
(source: image backlink to original publication)
1) I believe that the great challenge of the time regarding the discipline of Big Data falls on DATA. The DATA + AI combo is the holy grail for solving problems from autonomous vehicles to health where I see DATA as the differentiated asset while AI is “commoditizing”.
In the field of health, it is becoming increasingly easy to collect data from people through ubiquitous technological devices in their possession (smartphones, wearables, etc.). This is very conducive to the proliferation of various companies that, leveraging on this, can exploit access to democratized data that previously was only accessible by a few. Therefore, today more than ever there are more companies that can answer questions related to our well-being.
2) Regarding DATA in particular, we can think of two important dimensions: a) The collection history or duration of the observations and b) the number of attributes measured per unit of time. Regarding health, there are very important historical studies that were primarily focused on point a). For example, the famous "Framingham Heart Study" study that started more than 70 years ago and continues to analyze cardiovascular risk.
DATA from that study has been fundamental in crucial discoveries; for example: the relationship of smoking to the increased risk in cardiac pathologies. Nowadays, according to remark 1) mentioned above, we are in a position where we could also incorporate dimension b) into the picture. However, it is important to keep in mind that it will be necessary to accumulate historical data that will only be produced over time. Therefore, in my opinion we are going through the very initial phase where we can collect immense amounts of attributes and with a temporal resolution of milliseconds, but it will require some non-negligible time (keep in mind that the first important discoveries of the Framingham study only came 12 years after it started) until we will be able to inject those variables into different AI algorithms to answer the still unresolved questions on human health.
3) It strikes me that regulatory issues such as HIPPA are not mentioned in the article. Only, through a linked note, there is a reference made to a “federated learning” processes that has more to do with how various companies can “exchange” the data among themselves without running the risk of losing their own business assets and generating knowledge from the collaboration. I think the topic on regulations is going to be critical and I still don't see it as present in the public opinion as it is about other types of data (for example: FB data, telco data, etc.).
4) On the other hand, beyond the fact that in many aspects the arguments related to reducing the "lag to market" have their support, I believe that on health issues, the "precautionary principle" and the associated regulations that ensure the risks and harms that new drugs can cause are not going to be easy to avoid. Moreover, I do not know whether it is something that we want to take lightly as a society without falling into a paralysis side or a vision of extreme benevolence from nature.
These issues in my opinion are going to slow down the outputs that can be generated from Big Data in the pharmaceutical industry, and coupled with this I think it will be a great technological, moral, and business challenge to avoid the utilization of health in discriminatory practices, differential coverage prices or access to certain drugs.
5) Finally, I believe that there are certain spaces from the non-invasive health topics such as studies within the list of DSM-5 disorders (eating disorders, sleep disorders) that can be very beneficial for society and probably easier to work without major drawbacks as the ones discussed before.
Comments