DeepMind : between fine-tuning and disruption in the Healthcare industry

Imagine a world where doctors and nurses can have tasks taken off their day-to-day through data analytics and automation. A scenario where scans, tests results, follow-ups and notes are not only automatized but also given something the healthcare world desperately lacks today: structure.

One can identify many reasons for why the healthcare industry seems ripe for disruption. Beyond the outdated tech and oft-times byzantine processes, organizations across different countries suffer from one main constraint: keeping the upward-set costs structure under control. In most developed countries, a consistently-aging population and dwindling budgets threaten to bankrupt systems dating back to post-WW2. While many struggle with this challenge, an agreed-upon low hanging fruit lies in streamlining the current workflows. It is worth noting that, while the media lend hype to the claims of the tech sector to : find the best treatments for cancer (e.g. IBM Watson) or customize treatments for each patient’s DNA (e.g. Human Longevity Inc.); the main players are not distracted by the fact that the most considerable gains are to be reaped from enhancing the very existing system we have today. In the words of Moos Suleyman, founder of DeepMind: “There is no other area where we invest so much money in technology and get so little back”.

Imagine a world where doctors and nurses can have tasks taken off their day-to-day through data analytics and automation. A scenario where scans, tests results, follow-ups and notes are not only automatized but also given something the healthcare world desperately lacks today: structure.

Enter DeepMind, Google’s digital brain arm. For its next challenge, DeepMind has boldly decided to move from games[1]  to healthcare. As Bloomberg stated recently, “while DeepMind’s research on Go may be years away from yielding practical applications, its health-care work is affecting people’s lives today through projects with the U.K.’s National Health Service”[2]

Having built long-lasting partnership with UK healthcare organizations, DeepMind rolled out an app called Streams. Through its app, and by centralizing and analyzing patients’ data, DeepMind is able to provide healthcare providers with a sense of where patient’s condition is headed. For instance, data suggesting patient X is suffering from a deterioration, prompting preventive action. Furthermore, the hospitals where it is being tested stand to gain from fast-tracked triage. Indeed, all primary and “routine work” of reviewing tests and scans can be performed by an algorithm rather than a human, leaving doctors and nurses more time to focus on patients and more pressing matters. What’s more, DeepMind recently announced it is working on a digital ledger system – akin to blockchain technology – that would give NHS hospitals a tamper-proof audit trail of who has accessed patient data.

This might alleviate some of the pressures and concerns surrounding the whole initiative. As it stands, tech companies today are no longer seen as benign entities when it comes to handling critical data, let alone in the context of patient data. As Bloomberg stated: “DeepMind’s maiden voyage into the field has […] run smack into an iceberg of privacy and ethical concerns—and the resulting controversy has threatened to sink its ambitions of using AI to transform health care.” While Streams does not specifically use AI, it is a fair assumption that the Google arm is aiming for greater use of it in the future. The NHS-provided data is then seen as training ground for future healthcare-work intelligent algorithms.

As of today, who really benefits most from this deal lays in the eye of the beholder. Data does allow DeepMind – some would claim Google in general, though questionable from a legal standpoint – to improve its technical and commercial viability. The healthcare providers, on the other hand, enjoy seldom-encountered process-enhancing innovations whose costs are technically supported by DeepMind.

After a year-long investigation, U.K. regulators ruled that a London Hospital had illegally provided DeepMind access to 1.6m patient records going back 5 years. Although DeepMind itself was not directly impacted by this ruling, this begged the question of where its business model was headed? Are the regulators, the healthcare providers and the patients all on board with a tech company having access to health records? Tomorrow shall tell.

[1] :
[2] :
[3] :


Nielsen: What people watch, what people buy


AutoGrid is Turning the World’s Energy Data into Power

Student comments on DeepMind : between fine-tuning and disruption in the Healthcare industry

  1. Thank you for your insightful post! I am keenly interested in healthcare and DeepMind has always been a poster child for using AI in healthcare, especially after its high profile acquisition by Google. Your post correctly states how much AI and machine learning can add value to providers and patients, and how companies like DeepMind are building such capabilities. While I am super hopeful that such companies succeed, I have found that those companies with more limited a scope of what they’re solving are more likely to succeed. For example, BenevolentAI (another UK company that uses AI for drug discovery) has achieved much more traction in making drug R&D cheaper. Science 37, which leverages AI and ML for clinical trials only, has had a lot of traction with multiple pharma clients to lower clinical trial costs. Flatiron Health, which leverages data analytics to solve for cancer research, just got sold for $2B to Roche and has a powerful database that multiple healthcare players can leverage. In this ecosystem, DeepMind seems to be trying to solve for everything or rather, they seem to attempting to use what they have learned in previous AI applications to healthcare. I doubt if such a broad approach can succeed – it is like trying to find a needle in a haystack. I realize that this is another one of Google’s forays into healthcare but I wonder it the resources of DeepMind are being spent wisely. Nonetheless, I truly hope DeepMind succeeds in applying AI to healthcare – the world desperately needs it!

  2. It is very interesting. I do wonder though, whether Deepmind could have been used for more targeted approach whereby personal data was not collected on the patients, but as you mentioned the necessary data for profile were collected to help improve diagnosis of patients, and learn about how to treat patients better. You could then use machine learning to improve the accuracy of diagnosis, and maybe one day artificial intelligence to prevent mistakes from happening or predict the probability of certain conditions to develop.
    I also wonder how different Watson for Doctors and DeepMind Health are different. Do you have any idea on that?

  3. Given that DeepMind have focused on high predictability machine learning with huge datasets, I wonder if they are best placed to solve healthcare problems. DeepMind algorithms may make accurate predictions, but it will be almost impossible to prove causality, like putting the data into a black box and getting an answer. This is a particularly trixky issue in a high accountability industry like healthcare.

Leave a comment