Unsupervised learning or codified knowledge, which is best for AI? Should we subject new technologies to ethical scrutiny? Is Silicon Valley’s arrogance a soluble problem? How do we design technologies for women? How will we build autonomous cars that last? What makes us conscious? Are prime numbers colluding?
Dept of the near future
🔮 Are new technologies social experiments that should be subject to similar sorts of ethical scrutiny as medical experiments? Or should we just let the market dictate? DEEP SET OF ISSUES
😎 Silicon Valley’s unchecked arrogance: “In its mind, Silicon Valley creates the future, while the rest of the world will soon become the ‘idle class.’ What if they instead helped people build wealth for themselves?” EXCELLENT
🆕Facebook is the new Excel. INSIGHTFUL
🦁 Hunting the new high tech. How Andreesen-Horowitz finds the next big thing.
🙋🏽 “How to make sure women’s needs don’t become an afterthought”. The problem with a technology revolution built by men. MUST READ (The academic paper is short but relevant.)
Dept of autonomous vehicles
Autonomous cars are going to sweat far more than human-driven cars. In a shared-asset environment where we access vehicles on demand, our cars may be run for close to 24 hours a day. This places issues on physico-mechanical longevity of the vehicles, and the “vehicle’s design life will have to jump to 1 million miles” according to BAE systems longevity.
Of course, electric cars are mechanical simpletons compared to internal combustion vehicles but there are still physical components, tyres, wheels, door seals that will weather - as well as, of course, battery fatigue.
This electric, autonomous future may technically be possible today, but it looks like some governments will demand a phased approached to the transition. The US Government looks set to insist that autonomous vehicles have steering wheels and brake pedals.
We are seeing the benefits of autonomous technology in today’s vehicles. Augmented driving means it may be impossible to crash the new Mercedes E-Class, even if you try. 🚗
What it’s like to take a spin in the Google self-driving Lexus.
👬Co-operative route planning: sensor-laden vehicles which can communicate with each other might allow for large-scale co-operation. “If we were all willing to take a wider variety of coordinated routes that may not be optimized on an individual level, it would yield an overall reduction in congestion.”
Dept of artificial intelligence
We’re still digesting what AlphaGo means. Here is what Yann LeCun (Facebook’s AI honcho) had to say: “We need to solve the unsupervised learning problem before we can even think of getting to true AI.”
One opposite approach to AlphaGo (learn from data) and LeCun’s argument for unsupervised learning is the approach pioneered by expert systems: to codify knowledge in a series of well-defined relationships.
The largest project of these is Doug Lenat’s CYC project, a 30-year (so far) journey to encapsulate common sense. Cyc is finally being commercialised through Lucid.Ai which seeks to build a better Siri. (Excellent read.)
This interview with Lenat on Cyc from 1999 is worth reading for historical perspective. This project had been fermenting for a while ;)
Brilliant exposition on how an AI took 2048, the iOS game, to new and ridiculous heights. (32768 anyone?)
🐦 A deep-Q network has also smashed the irritating Flappy Bird.
📼Neural networks demystified. Great for beginners.
Short morsels to appear smart at dinner parties
What makes us conscious? Short, extremely accessible introduction to integrated-information theory.
🎲 Anti-sameness: a new pattern found in prime numbers.
Imgur: the amazing rise of a meme-generating community.
🐁High-fat diets in mice make for fatter children. More evidence for epigenetics. (This week Cambridge Epigenetix raised $29m to capitalise on technologies to exploit the epigenome.)
Railway chief: “Fossils fuels are probably dead”. Wonder when this gets priced into stocks?
Pollution contributes to 12.6m deaths per annum says WHO.
🇨🇺 Cuba on the edge of change. Stunning photo essay.
Please do send interesting blog posts to me. My filters/sensors can always be augmented by a bit of human intelligence ;)