Can Democracy Survive Artificial Intelligence?
Scientific American asks this question after surveying the current state off affairs where pervasive computing has ushered in a level of total surveillance that George Orwell wrote about as fiction. The article points out the example of Singapore as the canonical pathway implementation of the digital state of the future where nudge and mass scale customisation of news atomises individuals into their own filter bubble. Cut off from a sense of unifying polity the author suggests we are groping towards totalitarianism aided and abetted by digital technology:
personalized information builds a “filter bubble” around us, a kind of digital prison for our thinking. How could creativity and thinking “out of the box” be possible under such conditions? Ultimately, a centralized system of technocratic behavioral and social control using a super-intelligent information system would result in a new form of dictatorship. Therefore, the top-down controlled society, which comes under the banner of “liberal paternalism,” is in principle nothing else than a totalitarian regime with a rosy cover.
The message is that this brave new world coming whether we like it or not and a better alternative is still possible if we act collectively by asserting our values:
Big data, artificial intelligence, cybernetics and behavioral economics are shaping our society—for better or worse. If such widespread technologies are not compatible with our society’s core values, sooner or later they will cause extensive damage. They could lead to an automated society with totalitarian features. In the worst case, a centralized artificial intelligence would control what we know, what we think and how we act. We are at the historic moment, where we have to decide on the right path—a path that allows us all to benefit from the digital revolution.
Interesting look at the application of AI in the fashion industry through the guise of a small luxury lingerie brand called Cosabella. Unlike many other boutique outfits operating in that space, their CEO was prepared to allow an outside consultancy build models with their raw customer data with impressive results:
Cosabella is using artificial intelligence and machine learning to track customer behavior, high- and low-performing products, and popular silhouettes and color patterns to predict what new categories and pieces will sell.
Bob the Builder too could be obsolete within a few years:
A San Francisco-based 3D-printing startup, Apis Cor, has come up with an exceedingly affordable solution for building new houses. It can 3D-print concrete walls for a small house in under 24 hours.
Netflix’s ‘Dynamic Optimizer’ is an interesting new way of improving the quality of playback in really challenging low bandwidth environments like for example mobile devices in Africa. The image below shows the before and after improvement on a 100kbps link. Dynamic Optimizer arose because Netflix are ‘allergic to rebuffering’:
It’s using artificial intelligence techniques to analyze each shot in a video and compress it without affecting the image quality, thus reducing the amount of data it uses. The new encoding method is aimed at the growing contingent of viewers in emerging economies who watch video on phones and tablets.
Interesting Google Research write-up of the unique on-device Machine Learning solution they developed for Android Wear 2.0 which uses a projection embedding technique optimized for low-memory constraints of a typical disconnected smartwatch:
We first use a fast, efficient mechanism to group similar incoming messages and project them to similar (“nearby”) bit vector representations. While there are several ways to perform this projection step, such as using word embeddings or encoder networks, we employ a modified version of locality sensitive hashing (LSH) to reduce dimension from millions of unique words to a short, fixed-length sequence of bits. This allows us to compute a projection for an incoming message very fast, on-the-fly, with a small memory footprint on the device since we do not need to store the incoming messages, word embeddings, or even the full model used for training. … Next, our system takes the incoming message along with its projections and jointly trains a “message projection model” that learns to predict likely replies using our semi-supervised graph learning framework. The graph learning framework enables training a robust model by combining semantic relationships from multiple sources — message/reply interactions, word/phrase similarity, semantic cluster information — learning useful projection operations that can be mapped to good reply predictions.
Cool name and a back to the future product. Twenty years on the ground-breaking Psion 5 PDA is reprised in the guise of the Linux-based Gemini put together by the slightly mysterious London-based Planet Computers group. Turns out they include some old Psion luminaries including industrial designer Martin Riddiford of Therefore:
Good LinkedIn post on Agile Quality Assurance (QA) and Testing. It’s all about embedding:
In Cross Functional Teams the responsibility of testing belongs to the whole team, not just a QA. QA’s are there to share their specialist knowledge to help the team be better at assuring quality.
Vox on a sadness data analysis of Radiohead’s mighty oeuvre courtesy of some sentiment analysis of their lyrics and Spotify metadata:
The takeaway? Radiohead has been sad for a while — but maybe not as depressing as its cultural reputation might indicate. Its low point right up until last year was its mid-course, the 2001 album Amnesiac, which scored a 38 out of 100 overall on the “gloom index.” Afterward, the band steadily climbed toward the light before plunging sharply again with last year’s A Moon Shaped Pool.