🔮 Facebook's bet on AI; taxing robots (or not); the changing nature of employment; outsmarting Intel; heroin in Philly; digital contraception; hoarding cobalt++ #102
Facebook’s AI ambitions. VR isn’t a fad. Why we shouldn’t tax robots. The changing nature of employment. Using cryptocurrency to align incentives. In praise of simple heuristics. Digital contraception. Heroin in Philadelphia.
Hope this sparks great conversations!
❤️ Love this? Please share it via Twitter, on LinkedIn, by Facebook.
🚀 Forwarded this from a friend? Sign-up to Exponential View here.
Dept of the near future
💡 Stephen Levy on how Facebook is infusing artificial intelligence into its products. What struck me is Facebook’s conviction that building decent consumer products now requires the predictive capabilities of AI. LONG READ
Ⓜ Great analysis on how Facebook is scaling down the ambitions of its AI assistant, Facebook M. M was only able to service 30% of requests without human intervention.
🌈 Is VR a fad? Early numbers suggest the adoption curve is not far off that of the iPhone.
👩💻 Self-employment in the UK is growing significantly faster than other advanced economies. The growth has been in high-paid sectors on one hand and precarious ones on the other. Does it represent a remaking of labour markets? Full PDF | Brief summary. (Also, report suggests 250k UK civil servants could be replaced by automation by 2032.)
🇺🇸 David Brooks argues the US in this century is broken. “The 21st century is looking much nastier and bumpier: rising ethnic nationalism, falling faith in democracy, a dissolving world order.” And no, it isn’t robots that are killing the growth. See also this fascinating extract from a Goldman Sachs report on the problems in the US labour market and this grim prognosis that only catastrophe has reliably reduced income inequality.
💸 “As the true cost of Bitcoin transactions rises, utility at the margin falls, and the platform’s fundamental value as a tool for human economic interaction declines alongside” argues Eric Vorhees.
Dept of robots & artificial intelligence
🤖 Should we tax robots, as Bill Gates has suggested? The Economist makes a persuasive argument that we shouldn’t. The rationale? Automation doesn’t seem to be the cause of the reduction of the returns on labour, rather it is the increase in the power of certain firms (by network effects, superior cultures or government protections, like IP.) MUST READ.
pix2pix - demo of deep-learning image translation
I drew the stylish image of a backpack with the help of the implementation of pix2pix and image-to-image translation. Have a play yourself. (Won’t work on iPhone.)
💵 Numerai, which is backed by a number of subscribers to this newsletter, is a startup building financial trading models by ensembling the best algorithms from a community of data scientists. Numerai had to overcome an incentive problem. How do you get people to share algorithms that generate alpha when participants in a market are usually adversaries? Numerai proposes a novel solution where traders back their confidence in their algorithms by stakes in a new cryptocurrency. (More detailed PDF here.)
Performant algorithms lead to excess payouts, poor performers lose their stake. The model might align incentives for co-operation on the platform. Will Numerai’s approach work? No choice but to wait and find out.
How AI has wrong-footed Intel (and its long-held bet on generalised CPUs)
A major US hospital ended its collaboration with IBM Watson, the firm’s AI unit (inflated expectations?)
AI smashed humans in Super Smash Bros. “After a couple of hours the AI was good enough to beat the in-game AI, and after a couple of weeks it could beat the top-ranking humans.“
Solid list of many the narrow tasks where AI outperforms humans
DeepCoder learns to write code by stealing from other programs
Here is where jobs will be lost when trucks drive themselves
GM plans to test thousands of self-driving cars in American cities next year
🎧 New EV podcast: Intimate lives with robots with Kate Devlin
Small morsels to appear smart at dinner parties
💡 Simple heuristics often outperform complex rules. Great profile of Gerd Gigerenzer
Inside Philadelphia’s heroin epidemic
Mexico’s sugar tax is succeeding in reducing sugar consumption
💏 Elina Berglund’s contraceptive algorithm is as effective as the pill.
How Beepi failed. Analysis of the implosion of a once-hot car marketplace
🔋 Speculators are hoarding cobalt (needed in Li-Ion batteries)
Six-legged robot uses two legs to move faster
The SHA-1 cryptographic function has been broken. EXCELLENT read
🚀 Organic compounds found on Ceres
👽 Seven temperate exoplanets discovered
What you wrote
Two essays by EV readers this week are worth reading.
Nick Russell asks is it possible to modify individual human brains via social media?
Husayn Kasai makes the case for Universal Basic Income
End note
I watched I, Daniel Blake, a great film about a carpenter dealing with unemployment, navigating Britain’s social safety net and trying to maintain his dignity. It’s a wonderful film on its own merits. One element that stood out as particularly interesting. The protagonist needed to steer a byzantine bureaucratic process with no real explicability of the how it worked and with a limited right of redress: it was essentially a giant algorithm. And an inhuman one at that. Recommended viewing!
Have a super week
Azeem
P.S. Take time to find us on Twitter if you haven’t already.