eBay's algorithms can identify 40% of credit card fraud transactions

"eBay scientists have published a report which says that it’s new AI algorithm can identify 40% of credit card fraud transactions with high precision. A significant finding for a sector which solely relies on technology for fraud detection.
As always there are two sides of the coin for any entity or method. In fraud detection also, the techniques and tools can either focus on good actors or bad actors. Till now it has been the former. Where the truth is that majority of transactions conducted are by good actors. So it was imperative to study the behaviour of good actors, in fact much more important than those of bad actors.
Hence eBay, rather than focusing on the changing patterns employed by bad actors to circumvent protective barriers, they decided to instead analyses instances of good behaviour.
San Jose-based eBay scientists Utkarsh Porwal and Smruthi Mukund noted that patterns of good behaviour do not change with time. The data points that represent this form of conduct have consistent spatial arrangements. Porwal and Mukund suggested a clustering method for identifying outliers and to later formulate a score, which would determine consistency and in turn, good behaviour."

Source:  Next Big What, 14th November 2018

The Apple Watch can detect diabetes with 85% accuracy

"According to Cardiogram founder Brandon Ballinger’s latest clinical study, the Apple Watch can detect diabetes in those previously diagnosed with the disease with an 85 percent accuracy.
The study is part of the larger DeepHeart study with Cardiogram and UCSF. This particular study used data from 14,000 Apple Watch users and was able to detect that 462 of them had diabetes by using the Watch’s heart rate sensor, the same type of sensor other fitness bands using Android Wear also integrate into their systems.
In 2015, the Framingham Heart Study showed that resting heart rate and heart rate variability significantly predicted incident diabetes and hypertension. This led to the impetus to use the Watch’s heart rate sensor to see if it could accurately detect a diabetic patient."

Source:  Techcrunch, 7th February 2018

Facebook carries out 4.5bn translations a day, all using machine learning

"Creating seamless, highly accurate translation experiences for the 2 billion people who use Facebook is difficult. We need to account for context, slang, typos, abbreviations, and intent simultaneously. To continue improving the quality of our translations, we recently switched from using phrase-based machine translation models to neural networks to power all of our backend translation systems, which account for more than 2,000 translation directions and 4.5 billion translations each day. These new models provide more accurate and fluent translations, improving people's experience consuming Facebook content that is not written in their preferred language."

Source:  Blog post on the Facebook code blog, 3rd August 2017