ArchiTECHt Daily: Finding the right balance on data privacy

1. Read this story about a strange data-sharing deal between Google DeepMind and a London hospital. (
ARCHITECHT
ArchiTECHt Daily: Finding the right balance on data privacy
By ARCHITECHT • Issue #39
1. Read this story about a strange data-sharing deal between Google DeepMind and a London hospital. (Or read the full paper it’s based on, which actually is more like a piece of investigative journalism than a research paper.)
2. Think about what a mutually beneficial arrangement that adequately ensures privacy would actually look like.
There are lots of funky things in the original arrangement between the two organizations, where Royal Free NHS in London gave DeepMind access to patient data, supposedly as part of a program to improve care for patients with kidney problems. The deal was, according to the paper’s authors, vague, opaque and peculiar in that in prohibited DeepMind—a deep learning company—from using data for machine learning.
DeepMind, for what it’s worth, has actually blogged about the agreement a couple of times in the past month:
I think the bigger issue here is how we strike the right balance between privacy, innovation and the law. It’s easy to pick a side on opposite ends of the privacy-innovation spectrum, but the reality is that there’s a lot of space in between that we probably haven’t adequately investigated. This includes the very basic question of where data is actually safer from prying eyes (I might argue on Google servers).
A functional public-private partnership on issues like health care could be remarkably beneficial to everyone involved—patients included—but it’s going to require governments moving beyond some preconceived notions about privacy, and companies putting in the extra work to ensure they’re putting privacy first.
For a couple more stories from yesterday generally related to this topic, check out:

Sponsor: Datos IO
Around the web: Artificial intelligence
This is probably fine for rental and homeowners insurance. I would caution against automating too much for other types of insurance, though.
This is a good write from WIRED about research into how AI systems can communicate with each other in their own, purposeful manner, rather than trying to recreate human communication. Read the full paper here.
The results of this study are pretty remarkable—large annual gains, even during market turmoil. I would imagine there are quite a few folks already making a killing with approaches like this.
phys.org  •  Share
And they want access to systems from Google and IBM when they’re available—Google says in 5 years.
Sponsor: Marshal.io
Around the web: All things data
For certain workloads, at least. Watch this 3-minute video of Berkeley’s Michael Jordan explaining the Ray project, or check out its website.
How does a 400-500 percent increase in throughput sound? Or a 60 percent reduction in latency?
Gartner’s Merv Adrian looks into which Hadoop ecosystem projects have support from leading distributions; Kafka was the last one to garner broad support. The new move is to integrate pet projects.
This post from Cloudera turns into a marketing pitch, but it begins with a good look at the open source tools data scientists use, and why they use them.
Did you enjoy this issue?
ARCHITECHT
The most interesting news, analysis, blog posts and research in cloud computing, artificial intelligence and software engineering. Delivered daily to your inbox. Curated by Derrick Harris. Check out the Architecht site at https://architecht.io
Carefully curated by ARCHITECHT with Revue. If you were forwarded this newsletter and you like it, you can subscribe here. If you don't want these updates anymore, please unsubscribe here.