🔮 The data monopolies; bikes vs cars; Schmidhuber on smart machines; can AI save science; personal space, hipsters & ketamine++ #112
|Azeem Azhar||May 7, 2017|
How should we act on data monopolies? Are bikes the future of transport? Is human intelligence limiting scientific progress? How can we explain AI’s decisions? Ketamine as an anti-depressant, bearded hipsters, understanding personal space.
**Feel inspired this week :) **
🚀 Forwarded this from a friend? Sign up to Exponential View here.
Dept of the near future
📡 “The world’s most valuable resource is no longer oil, but data”, argues The Economist, and “if governments don’t want a data economy dominated by a few giants, they will need to act soon.” We have raised the issue of data monopolies several times in EV, good that it is getting mainstream attention. The Economist’s detailed briefing on how regulators should act is a MUST READ.
🎙️ Brilliant interview with Jürgen Schmidhuber, the less well-known pioneer of many breakthrough AI technologies. “Our best protection [from powerful AI] will be their lack of interest in us, because most species’ biggest enemy is their own kind. They will pay about as much attention to us as we do to ants.” EXCELLENT
🍃 By 2030, self-driving electric vehicles will represent 95% of all miles driven, argues a new report on sustainable EVs. (See also this week not even oil companies can afford to stay silent about EVs as new oil development hits a 70 year low. And, India’s power minister, Piyush Goyal announces the ambition that "not a single petrol or diesel car” will be sold in India within 13 years.)
🚣 Could artificial intelligence save science from the limits of the human mind? Going back to 17th-century England and Francis Bacon might give us some clues. THOUGHT PROVOKING
🚴 Are bikes the future of transportation? Horace Dediu argues that “bikes have a tremendous disruptive advantage over cars. Bikes will eat cars.” (China is the leading market, so much so that bike-sharing is causing city authorities problems. The FT has some useful data on this. 450k sharing bikes in Shanghai alone.)
Dept of Artificial Intelligence
“Who’s responsible when something goes wrong?” is the core question of AI’s black box problem. Jill Schwiep gives a great overview of the current state of explainable AI. See also: Nvidia is developing systems that show, in real-time, what self-driving vehicles “see”.
Melinda Gates and Fei Fei Li unite in an attempt to “liberate AI from guys with hoodies”. The main concern: “If we don’t get women and people of colo[u]r at the table — real technologists doing the real work — we will bias systems”.
Machine learning has such massive computational demands that new chip architectures will almost certainly be needed. Cade Metz briefly reviews what Facebook, Nvidia, Google and Intel are trying.
ARM designed an image signal processor for autonomous vehicles, and the technology promises to have a big impact in mobile camera technology.
Inequality of computing resources strangles academic research. Libby Kinsey: “ […] what avenues remain open for compute-limited academic contribution? […] (I also wondered whether this will soon be a moot question, when GAFA et al have finished recruiting all the academics!)”.
New startup Lyrebird uses machine learning to generate AI speech from any voice (including those of politicians.) Not perfect. Yet…
Per The Economist story above, data is the rocket fuel for artificial intelligence. While the GAFAs and others seek to build monopolies on key datasets, unlike oil where any given barrel of oil is a perfect substitute for any other given barrel, data is not as fungible. True, the GAFAs have very interesting datasets about personal preferences and behaviour (and the reach of humans to leverage them on) but there are many types of data they currently don’t yet have.
It is heartening to some of these being released for developers to experiment on. This exciting dataset of LiDAR data from Washington DC is now available on AWS, as a public dataset. (h/t Rodolfo Rosini).
Third-party developer scraped 40,000 pictures off Tinder to create a facial dataset for AI experiments. Company portends investigation.
“Liar, liar, pants on fire!” The largest publicly available dataset for fake news detection covers one decade, and 12.8K manually labelled short statements.
Google tweaks its algorithm and changes rules for its “raters” in an attempt to suppress fake news. (See also Lyrebird above.)
Former Google Ventures, now GV, leading a $10 million investment round into Abundant Robotics, the company that built a ‘self-driving car’ that picks apples. Robots work around-the-clock, can handle fragile fruits, and keep the costs low. Video here.
Short morsels to appear smart at dinner parties
🥇 Scientists surgically remove HIV DNA from live animals using CRISPR/Cas9.
Using graphene to create multi-purpose thermoacoustic speakers. COOL
First large-scale study reinforces ketamine’s effectiveness as an antidepressant.
😞 Leaked documents show Facebook helping advertisers target teens who feel “worthless” and “need confidence boost”.
Looking into data: Is Facebook tougher on female engineers? ♀️
🍼 Early life stress alters genes responsible for controlling bodies’ response to stress later in life.
According to 1200 years long record-keeping, Japanese cherry blossoms are emerging increasingly early. Climate change? 🌡️ 🌸
More farm mechanisation in the ‘40s has led to more intense rainfall in the US Midwest.
📚 Paper books reclaim market share, as ebooks falter. (Also read my personal reflection on why paper books trump Kindles, includes some science.)
How do you actually configure a new passenger jet? It is fiendishly complicated. 👍
🏠 Personal space has a lot to do with where you grew up, a new international study finds.
A new episode of the Exponential View podcast is out — Scott Santens and I discuss universal basic income: the concept, reality, and opposition. Give it a listen, and let me know what you think via Twitter.
Azeem (also on Twitter as @azeem)