Superintelligence; programming our cars to kill; algorithmic education; machine learning; the future of entrepreneurs++ #32
What will super-intelligence value; machine learning & feature optimisation; creativity in AI; decisioning kills in autonomous vehicles; entrepreneurs & the future of capitalism ; genetically modified beagles; apes with small balls.
Share this (help get to 2,500 subscribers!): Tweet this!
Referred by a friend? Sign up.
Dept of the near future
👳🏽 What will super-intelligences value? And can we engineer those values? Excellent
🚓 Why self-driving cars must be programmed to kill: the Trolley-experiment needs be encoded. (Wonder how many ethicists Tesla or BMW have on staff?) Must read
😜The must-read Michael Wolff deck on media and media consumption. (e.g. Multi-tasking has increased the effective length of the day to 31 hours.) Invest time to read
🔥 Climate change may reduce global incomes by 23% within 90 years; with the burden of the reduction being borne by the poorest nations widening inequality. **Please read. **(More accessible version of same research here).
🔮 Information technologies are becoming ubiquitous and innovation with them will become ubiquitous but small argues Jerry Neumann Reposting from last week because not enough of you read it - it’s really good!
🎓 The algorithmic future of education: Audrey Watters cautions against relying on AI and algorithms to engender serendipity and curiousity in education. Long, thought provoking
😛The disruptive idiots from Silicon Valley: Theranos & technology Darwinism.
Dept of machine learning
No grand thesis this week, but a handful of interesting readings looking at some of the building blocks of machine learning.
In the wild, machine learning is still an early stage technology. Google, the most successful firm built on machine learnin, so far, is rethinking its entire approach. (More, not less, machine learning.)
* Deep feature synthesis seems like a promising approach to using machine learning methods to enhance the development of machine learning models.
* EV subscriber, Prof Maggie Boden: “Artificial intelligence has an Achilles’ heel. It can’t decide what’s relevant.”
* Exploring the space of adversarial images: turns out radical misclassification of images using neural-network-based classifiers could be a bigger problem than we first banana.
* Ben Goetzel: Is Google Deepmind close to achieving AGI? tl;dr: No.
* Here are mechanisms by which AI might recursively improve itself.
* What happens when you show DeepDream real neurons? (Missed this from July, sorry.)
* Older still: nice taxonomy of machine learning approaches.
Dept of future of capitalism
“Entrepreneurs are redesigning the basic building block of capitalism” argues my alma mater, The Economist. Startups eschew public ownership, in favour of largely employee (and founder) owned structures. Financing comes not from CFOs and balance sheets but from venture partnerships with marginally longer time horizons.
👶Indeed, while the VC/entrepreneur dynamic is a young one, 43% of all US firms founded since 1979 were back ed by venture, and these firms account for 82% of all the R&D spending of that post-1979 cohort. VCs + entrepreneurs = faster growing companies, investing more heavily in R&D. Excellent Stanford report. (ht @brenthoberman)
One element of the VC/entrepreneur dynamic is the much heavier emphasis on employee ownership. But some, such as Tikhon Bernstam, believe founders are still too parsimonious, and need to more generously share ownership of their firms with employees.
The emphasis on growth and jam tomorrow, and the increasingly availability of private capital, can create bad behaviours among notionally-hot startups. VC Fred Wilson identifies that many of these corporations are guilty of negative gross margins as they try to capture market share. In other words, they are fundamentally unsustainable. Great read.
There is too much to say on this subject in a short post, and there are many questions to ask about these youthful, fast-growing countries and the dimensions of their behaviour.
😦An example, is the rise of quantifying worker behaviour through tracking, as explored in this academic paper. Quantified “workers have internalized the imperative to perform, a subjectification process as we become observing, entrepreneurial subjects and observed, objectified labouring bodies.” Will quantified workers be further subordinated to capitalism at the cost of their own well being? Highly recommended.
Dept of climate change
🔥Burn baby burn. 2015 is the hottest year every recorded. (Ever.) And it looks increasingly unlikely we can keep climate change to below 2 degrees C.
CO2’s new normal: 400ppm ht @revkin (Looks like exponential hockey stick here.)
Dalai Lama calls for the protection of Tibet’s glacier population.
“Hurricane Patricia is exactly the kind of terrifying storm we can expect to see more frequently in the decades to come” In fact, at one point Patricia became more powerful than the theoretical maximum for hurricanes.
(Scott Kelly’s amazing photos of Patricia from Space)
Small morsels
Tesla’s autopilot is augmented driving not autonomous driving. These drivers should know better.
🐘 Magic Leap is building an augmented reality computing platform. (With obligatory excellent demo video.)
🎏 Stream processing is an increasingly important technology in a world of never-ending torrents of selfies and telemetry from IOT devices. Here is a nice primer on it.
The staggering impact of screwed up IT systems. Really fun explora-graphic.
🐒 The louder the primate, the smaller the balls. May also apply to homo sapiens driving Ferraris/Lambos.
💥An excellent guide to product development through experimentation written by Twitter.
Einstein won’t be happy: physicists demonstrate spooky action at a distance.
Also physicists demonstrate Zeno’s paradox: a watched pot never boils.
🐶 These cute beagles are the first CRISPR-gene-edited mammals. They also have double the muscle mass of a typical beagle thanks to a gene knockout: good for “hunting, police (military) applications”. (China, natch.)
The necessary illusion of self: “Brain and body are inextricably linked—and work in concert to create our sense of self.” Recommended
What you wrote
Theranos’ problem is not its technology. It is that it has neither a business model nor consumer proposition writes EV reader Frank David.
EV reader? Written something interesting, let me know.
End notes
Last month’s Exponential Dinner on Artificial Intelligence with Maggie Boden was a good shout. The NPS of the event was 100 - so we’re looking to recreate that for future dinners.
It’s short notice, but I’d love to hold a dinner on November 23rd (in London). I’ll send out a note next week, but in the meantime exploring what the themes might be (willing to look at Mars, for example, as a thought). Drop me suggestions.
This week is too long, I know. Sorry. Shorter next week.