Month: October 2017

Scaling Surveillance

China is building the world’s most powerful facial recognition system to identify any one of its 1.3b citizens within 3 seconds. The goal is for the system to able to match someone’s face to their ID photo with 90% accuracy. However, the project is encountering “many difficulties” due to the technical limits of facial recognition and the sheer size of the database involved. Some totally unrelated people in China have faces so alike that even their parents cannot tell them apart.

Cynefin

Things we believe to be true that ain’t necessarily so
There are underlying relationships between cause and effect in human interactions and markets, which are capable of discovery and empirical validation. Under this assumption, an understanding of causal links in past behavior enables us to define best practice for future behavior. There is a right or ideal way of doing things.
Faced with a choice between one or more alternatives, human actors will make a rational decision. Under this assumption, individual and collective behavior can be managed by manipulation of pain or pleasure outcomes, and through education to make those consequences evident.
The acquisition of capability indicates an intention to use that capability. “We accept that we do things by accident, but assume that others do things deliberately.”

Spyscape

Adjaye Associates has unveiled designs for SPYSCAPE, a new museum and interactive experience that illuminates the world of espionage from historical secret intelligence to modern day hacking through a collection of rare artifacts, exhilarating storytelling and immersive personalized experiences. The space will use architecture as a key element of the museum experience. Inspired by the spaces occupied by the world’s most significant spy organizations, the building interiors will resemble a small town, with a variety of spaces unfolding beneath a vaulted canopy. Circulation will lead visitors through a wide range of vantage points and perspectives, playing with perceptions and drawing you into the individual pavilions.

Google Maps planets

20 years ago, the spacecraft Cassini launched from Cape Canaveral on a journey to uncover the secrets of Saturn and its many moons. During its mission, Cassini recorded and sent 500k pictures back to Earth, allowing scientists to reconstruct these distant worlds in unprecedented detail. Now you can visit these places—along with many other planets and moons—in Google Maps right from your computer. For extra fun, try zooming out from the Earth until you’re in space!

ML vulnerabilities

Identifying vulnerabilities in the ML model supply chain

we show that maliciously trained convolutional neural networks are easily backdoored; the resulting “BadNets” have state-of-the-art performance on regular inputs but misbehave on carefully crafted attacker-chosen inputs. Further, BadNets are stealthy, .i.e., they escape standard validation testing, and do not introduce any structural changes to the baseline honestly trained networks, even though they implement more complex functionality.

Material Abundance

is a world where basic needs are provided for and we can print anything a utopia? I posit: No. But it’s no dystopia either. Is a system that ensures nobody goes hungry and that no one person or corporation could manipulate the market beyond its natural elasticity worth the loss of what many of us believe is a fundamental inalienable right? Only the future will tell.

Police and systemic storytelling

police spend a large part of their time distributing crime to the sorts of people who seem likely to be criminals — the poor and marginal — and the prediction is prophetic: these people turn out to be criminals as soon as they are stopped and forced to turn out the contents of their pockets or glove boxes. Leave them alone, and most would never be “criminal” at all.

TFX

TFX also includes a suite of data transformations supporting feature wrangling. As an example, TFX can generate feature-to-integer mappings, known as vocabularies. It’s easy to mess things up when transformations differ in subtle ways between training and serving. TFX automatically exports any data transformations as part of the trained model to help avoid these issues.