Why You Don't Need To Share Personal Data To Benefit From AI
Source: Matthew Russell
When New Scientist published a leader that Britain’s National Health Service (NHS) had quietly shared the healthcare data of 1.6 million patients with DeepMind, Google’s London-based artificial intelligence (AI) division, there were complaints across the nation’s newspapers.
While the information was ostensibly shared to build an app that would help hospital staff monitor kidney disease, the journal reported that the agreement went much further. The notion of a tech behemoth poking around the most private of personal data, without direct consent, concerned patient and privacy groups. The unwillingness of DeepMind or the NHS Trust involved to discuss its plans for using the data — anonymized yet intimate details including patient location, visitor information, pathology reports, HIV status, past drug overdoses and abortions — didn’t help matters.
Among clinicians who analyze large datasets in the hopes of curing disease and improving care, this was a non-story. Seeking insights and trends in anonymized data is common practice and leads to vital medical discoveries. Applying AI to this task promises to extend and accelerate the benefits. Recognizing this, New Scientist commented: “This tension between privacy and progress is a critical issue for modern society. Powerful technology companies can tell us valuable things, but only if we give them control of our data.”
I think this is a false and risky assumption. As a computer scientist who focuses on data mining and works as the CTO of a cognitive computing company, I have no lack of enthusiasm for the possibilities of AI and data analytics. However, as a former military officer, I also understand why it’s important, often essential, to keep certain data private.
The Data-Capture Business Model
The data-capture business model — best exemplified by Google and Facebook, and embraced by countless internet-era businesses – has become an article of faith. Doubts about relinquishing our data are neutralized by the huge public appetite for “free” services provided by technology companies in exchange.
For decades, the vast majority of big data went unused because it was unreadable using conventional computers. Thanks to techniques like machine learning, deep learning and cognitive computing, we have dramatically expanded our ability to turn data into knowledge and achieve what might be termed “big understanding.” There is a wide acceptance that the strongest models are based on the largest data sources. Put simply: the more data, the better the model.
| }
|