How can decentralising data fight grid cyber-threats?

What would the end of net neutrality mean for healthcare? When will AI exceed human performance? Does understanding deep learning require rethinking generalisation? Ever deleted an email worth $400k USD?

Enjoy reading! ✌

--

How can decentralising data fight grid cyber-threats? Jesse Morris argues how blockchain technology can bolster cybersecurity:

As the world of energy becomes more digitised and decentralised, the need for solid defence against cybersecurity threats increases drastically. When a blockchain is implemented properly, it offers a strong defence against external and internal threats by mitigating internet-connected and data communications vulnerabilities, and increasing data confidentiality and privacy.

Morris also links to a few interesting papers/ articles, most notably on:

Disintermediation: "With blockchains, intermediaries (escrow corporations) often are no longer necessary, significantly reducing transaction costs."
Complete data availability: "Blockchains can store data in a decentralised fashion across many nodes. With this architecture, even if some nodes or servers are compromised, users still can access a complete dataset."
Privacy and control: "Users of a blockchain can choose which data to make immutably transparent and which data to keep encrypted so only the intended recipients can view the contents."
Outsourcing computation: "Encrypted data can be sent for processing to a third party, without the contents of the data being revealed."

Antony Ma explains blockchain and why it's groundbreaking in understandable terms (after he tried to explain it to his wife). Really like the idea that blockchain brings "properties of the physical world into the virtual world". 🌎

What would the end of net neutrality mean for healthcare? 🏥 Mike Miliard reports on a new article in Health Affairs. -> The proposed policy changes could unintentionally amplify the disparities in available healthcare for people with lower incomes or those living in rural areas. Side note, check out this plugin to see what the web is like without net neutrality. 😬

A new study, conducted by digital threat management firm RiskIQ, revealed that many top UK firms' websites violate GDPR 👎 :

Our research identified 100,000 live websites belonging to FTSE 30 organisations, 13,000 pages of which are collecting PII—an average of 400 pages per organisation. What’s worse, a third of these pages are still collecting information insecurely, either through lack of encryption or vulnerable, obsolete encryption algorithms.

China's new cyber-security law is will make it harder for foreign companies to operate - as they will be unable to transfer any customer data out of China, without direct permission from the individual:

Under the regulations, data on Chinese citizens – including personal information, salary details and more – can only be kept within China. The law would also prevent the transmission of any economic, scientific or technological data overseas on either national security or public interest grounds, as defined by the Chinese government.

When will AI exceed human performance? 🤖 Expert predictions below:

When AI will exceed human performance predictions, via MIT Technology Review

However, experts also predicted it would take 12 years for AI to be better at Go than humans -- which then only took two years. 😅

How can deep neural nets can be trained to perform multiple similar tasks? Interesting discussion of multi-task learning in deep neural networks. ✅

How large can a neural net become before it starts finding connections in random data? 👻 Short (technical) paper arguing why understanding deep learning requires re-thinking generalisation.

Jeff Leek shows why you shouldn't be using deep learning if your data isn't that big. 🔍 Sample size and quality matters: "if you have a more modest sample size you may not be buying any accuracy - just losing interpretability".

Miscellaneous

How to call B.S on big data -- nice practical guide by Michelle Nijhuis for the New Yorker. 👊

Check out the ICSI Haystack Panopticon - an "interactive map that illuminates the presence of third-party trackers such as analytics services and ad networks on Android apps".

Ever deleted an email worth $400k USD? 🙈 Backup your Ethereum wallet! -- This Reddit user managed to accidentally destroy the entire production database on his first day of his new job! 😓

Is a million lines of code a lot? 🤔 Neat viz by Information is Beautiful.

How have Lego colours changed? 👶 See the difference from 1975 to 2014:

Lego colours

The animated spiral [below] presents global temperature change in a visually appealing and straightforward way. The pace of change is immediately obvious, especially over the past few decades. The relationship between current global temperatures and the internationally discussed target limits are also clear without much complex interpretation needed: