🎉🔮 The Google Brain; redefining mobility; the new factory; Facebook; VR in 2017; fairness & democracy in AI; Euler, opioids, metformin++ #92
On the Google Brain and how it will reshape the firm. On the humans who train AI. Disrupting the mobility industry. Reinventing the factory. Algorithmic fairness. Progress in solar power. VR in 2017. Euler’s identity. The opioid crisis. Monopoly.
Last issue of the year - have great conversations!
❤️ Love this? Share it via Twitter, on LinkedIn, by Facebook.
🚀 Forwarded this from a friend? Sign-up to Exponential View here.
Dept of the near future
⭐️ The Google Brain: on the great AI awakening in Google and how it will change the firm forever. Grab a coffee for a long, sumptuous read.
🏭 Reinventing the factory: Great interview with EV subscriber, Nick Pinkston on how is thinking about manufacturing’s future.
🚲 The curious world of Mechanical Turks. Half-a-million humans train AI systems every day, pennies a task, on Amazon’s crowd-working platform. Excellent read. (Also; Izabella Kaminska: Inside the gig economy. Izabella becomes a Deliveroo driver.)
🚘 Business model disruption in the mobility industry. Olaf Sakkers provides a solid analysis of the different attacks on the traditional auto industry and the underlying shift from “selling [consumers] miles rather than cars.”
📺 Are screens ruining intimate, human time? “[The] big news … is that parents are doing a lot more parenting. In 1965, parents spent 41 minutes engaged in “primary care” for their little ones. That number had more than doubled, to 88 minutes, in 2012.”
Dept of artificial intelligence
Interesting profile of NumerAi, a fund which makes investment decisions based on an ensemble of models submitted by data scientists. Think Kaggle meets Mechanical Turk meets Renaissance Technologies. Contributors are anonymous and paid in Bitcoin.
Crowdsourcing has been a favourite tool for the data scientist, as it is one of acquiring ground-truth data. In NumerAi’s case, investment strategies are crowdsourced from scientists contributing their ingenuity. Several years ago, PeerIndex used Amazon Mechanical Turk to build some of our ground truth datasets.
🤖 The boom in AI is increasing demand for gold-standard data and with that in Turking. More than 500,000 people work as Mechanical Turks, being paid pennies per training instance. (Great essay which outlines the growth in the industry and profiles Turkers work and feel.)
Last week EV touched on the cost of how the cost of training AI systems was getting out of reach for many researchers and, indeed, small firms. This week, Andrej Karpathy from OpenAI took sharing one stage further by releasing an image classification network - model & calculated weights into the public domain. The particular model took more than five days of training on an 8 GPU rig, unobtainable to most. It’s an interesting case for how we can democratise access to this technology. (h/t @guillaumeallain). Another example is RASA NLP, an opensource API released this week.
EV reader, Auren Hoffman, makes a similar argument for widening access to training data: “If we want to massively accelerate artificial intelligence and improve human lives, we need to democratize access to data.”
⚖ On algorithmic fairness, great profile of Cynthia Dwork on the importance of algorithmic fairness:
[it is] pretty clear that algorithms were going to be used in a way that could affect individuals’ options in life
One of Dwork’s concerns is the risk of poor design choices in hiring algorithms. In HBR, Cathy O'Neil and Gideon Mann argue we are already seeing these risks play out.
As an aside, Dwork also invented differential privacy (now used by Apple to protect personal information).
In a similar vein, interesting IBM Research paper on “Learning to Trust AI”
Every company should have in place guidelines that govern the ethical management of its operations … [a]nd those same standards of ethical business conduct should guide the development of AI systems.
Elsewhere:
David Kelnar’s highly accessible introduction to machine learning and AI. Recommended for the non-technical.
My talk on AI in finance at the Anthemis Hacking Finance forum is also a decent non-technical introduction to the space.
EV subscriber, Riva Tez, on the hyperbole of RocketAI: “Turns out anyone can make a multi-million dollar company in 30 minutes.”
Brief discussion on a challenge for reinforcement learning. How to find reasonable reward functions in real, rather than toy, worlds?
Good summary of the NVidia DGX-1, an AI supercomputer in a box.
Dept of other tech (clean, VR, etc)
Cleantech
World energy hits a turning point. Solar now cheaper than wind with new installs running at less than a third of the capex of 2010.
Old LiOn battery packs from PHEVs and EVs could add 10GW of storage to the energy grid. (Nice business opportunity.)
Cities will become increasingly important in combatting emissions. (Especially relevant with populist petrochemical-friendly regimes seemingly gaining ground.)
VR
HTC Vive on the next two years in VR. Good read
Facebook starts to address the fake news problem it claimed it didn’t have. (I’m interested in connecting with initiatives tackling the fake news problem, especially any technical platforms. Ping me.)
EV reader Jonathan Wolff on how Facebook influences you: “I’ve spent a hundred million dollars on Facebook advertising - and it works.”
Crypto
Signal is Edward Snowden’s recommended instant messenger app. Why won’t it catch on?
Short morsels to appear smart at dinner parties
𝞹 e^(πi)+1=0 A nice introduction to Euler’s identity, one of my favourites.
Also, clever proof that the surface area of a sphere is 4πr^2
How to build a Death Star
The technosphere weighs 30 gigatons.
📈 12Tb hard drives? 4.6Gbps Wifi? #exponential
💰 Kenya’s M-Pesa lifted nearly 200k families out of poverty
The world’s best navigators are from the Nordics. Why?
Melancholy profile of the opioid crisis in the US & its impact on children
Humans check smartphones 9bn times a day
How many miles have you driven by Uber? One rider on this leaderboard has clocked 29k miles which I estimate to cost more than $75k. Anyone better that?
Miracle metformin: looks like it can suppress certain cancers
Shutting down an illegal bitcoin farm in China.
A robotic hand with a sense of touch.
Monopoly strategies & Markov Chains (My view: the only way to enjoy Monopoly is not to play.)
🐹 Politics got weird because neoliberalism failed.
End note
This is the last Exponential View of 2016. Next Sunday is Christmas Day so I’ll see you next year - a year which I hope will be one of active participation in the opportunities and challenges ahead.
Exponential View has grown from 3,500 readers to nearly 13,000. I was particular excited by the increasing diversity of subscribers. The newsletter started with an email to 30 of my mates. Today, more than 700 company founders subscribe, joined by nearly 300 venture capitalists, 17 University professors, and readers from West Point & the National Security Agency!
This year we held about 10 events with more than 500 attendees in three countries. And we started a podcast channel.
I would love to start 2017 with 14,000 subscribers. This is a massive ask. So if you have enjoyed the newsletter this year, please take a moment to share this via email to your friends/employees/portfolio CEOs/book club with a personal commendation.
Happy New Year and thanks for reading.