🔮 AI & Moore’s Law; the end of the petrol car; sewbots, cobots, & sexbots; big data’s big advantage; cash in China++ #123

🔮 AI & Moore’s Law; the end of the petrol car; sewbots, cobots, & sexbots; big data’s big advantage; cash in China++ #123

Cash is disappearing. Moore's law meets AI. Putting the combustion engine to rest. Collaborative robots booming, trouble with sex robots. Why business leaders should read sci-fi. Airbnb's 100m+ stays this year.

😃  Have a friend who would enjoy this? Forward it!

🌍  Go ahead, tell the world! Share via *|SHARE:facebook,twitter,linkedin|*


👻 Cash is disappearing in urban China, replaced by electronic payments on proprietary platforms like Alibaba and Tencent.

As the country builds its entire consumer economy around two private smartphone payment platforms, it is slowly locking out people unable to get onto those networks, and locking itself into those companies.  At the simplest level, that makes life difficult for tourists and business travelers who are unlikely to open a bank account in China and so will find it hard to turn their phones into wallets.

🇸🇪 Sweden has made even more progress being cashless through its Swish system. Fascinating details here.

💱 Adam Ludwin has a great introductory tweetstorm about Blockchain and decentralised apps.

🔔 Smartphone notifications are a classic case of ‘tragedy of the commons’ argues Scott Belsky. Here is how we can fix them.

🔌 California invested so much in clean energy it is having to pay other states to take its electricity.

🎚️ Just how large are the computational demands of artificial intelligence? Many, many times larger than the smartphone industry, I argue. (Not paywalled.)


When will electric vehicles finally put the internal combustion engine to rest? Experts seem to disagree. Bloomberg New Energy Finance, a source I generally trust, reckons that by 2040 33% of light-duty vehicles on roads will have electric powertrains. OPEC recently tripled its estimates for EV penetration to 22% by 2040.

🇪🇺 This week Dutch bank ING produced a more stark conclusion that by 2035, only electric passenger vehicles will be sold in Europe. The bank reckons that by 2024 the three major consumer hurdles to buying battery EVs, price, range and charging time, will have been successfully conquered.

An interesting analysis in this must-read report is how the shift to EVs will challenge Europe’s car industry. Compared to a petrol vehicle, labour’s share of value in a battery EV is much lower. A much larger part of the value held in the battery (i.e. a resource-driven component).

Additionally, the emphasis on the ‘product attributes’ of the car (Swedish safety, German reliability, Italian raciness) will decline in favour of a commoditised utility. On both these dimensions, Chinese manufacturers hold the lead.

Like ING, I have long been bullish on the progress that electric vehicles will make. Aside from the product dimensions above, in 2015 I argued that:

A small drop in consumer demand (by a switch to EVs or through more conservative driving) could make independent retail petrol outlets (forecourts) unprofitable and putthem out of business. This in turn hurts retail distribution and makes gasoline more inconvenient, and making EVs less inconvenient.

To add to carmakers woes, their dealer channel survives on really flaky economics dependent on the barely-profitable sales of cars and then significant service and financing revenue. One of the largest dealership/service networks, Penske Automotive, ekes out a 3% or so margin on about $20bn in revenues. This isn’t the kind of margin that gives you room to manoeuvre around a business model change that maintenance-parsimonious EVs are likely to create.

In addition, the US and, increasingly, the UK, many new cars are financed by quirky personal loans. These loans allow an owner to acquire a car for a small deposit and low monthly payments. After 3-4 years, they can trade the car in, buy it out right or just hand the car back. This model of financing is predicated on stable residual values. But right now most of the loans are on internal combustion (ICE) vehicles. These may hold their value if bought in 2017, but it seems possible that residual prices start to decline as the system effects of the EV transition make owning an ICE vehicle cumbersome. This could be a ticking time bomb for an already overheated market.

I’m more aligned with Stanford economist, Tony Seba, who reckons that by 2040 essentially all the miles driven will be done so by electric vehicles.

🛢️ See also: Min Zhu: The oil price is on borrowed time. (FT paywall.)

Apple is working on electrical batteries with a leading Chinese manufacturer.


⏲️ Tim Wu on how we’ll increasingly need tests to prove we aren’t dealing with a robot.

Using robots to fake support, steal tickets or crash democracy really is the kind of evil that science fiction writers were warning about.

Textile robots may impact emerging markets, where millions are employed in the garment industry. Of course, Polanyi’s Paradox very much comes into play. The deft actions of a skilled garment worker may hide many hundreds of nuances that may be hard to detail let alone turn into code.

The market for collaborative robots (or cobots) is hotting up. (Good read on some real-life examples.)

💤 John Naughton: how we can get robots to take on our least interesting jobs.

💓 Laura Bates on the trouble with sex robots.


A recent Google paper demonstrates how new deep learning approaches continue to improve in performance as they are lavished with ever more data. (The full paper from ArXiv is here.)

We’ve long known that improving the amount of data will improve algorithm performance. But pre-deep-learning approaches, like support vector machines, tended to a limit of effectiveness. There was a point at which adding more data didn’t help because the model couldn’t resolve with any greater granularity.

Not so with deep-learning. You can add more and more data, and so long as you make the model deeper (more layers) and chuck more processing at it, at least for machine vision tasks, it will improve linearly.

As Tom Simonite points out in Wired:

that there could be even greater benefits to being a data-rich tech giant like Google, Facebook, or Microsoft than previously realized.

🤐 Apple’s approach to privacy (which occludes personal data so it is non-identifiable) could hurt its AI efforts. Also, see EV#76 on the same issue.

I’ve been bullish about China’s potential to make progress in artificial intelligence. Now the government has announced it wants the AI industry to surpass other nations technologically and worth $150bn by 2030. China’s particular political economy means that such government edicts are likely to transmit in quantity to local governments, businesses and the academic establishment. (My comment last month on China’s potential in AI is here.)

Secretive Apple publishes its first blog on machine learning. This one is about image enhancement.

🔧 Joanna Bryson: The three sources of AI bias. (Wonderful, short.)


📼 We now spend 30 minutes a day watching videos on our mobiles.

Airbnb's exponential growth continues. It will exceed 100m stays this year.

💌 The internet had changed how we find our lovers. (2016 story, but great data)

How the American opiate epidemic is playing out on Reddit.

💡 Rural electrification in 1930s America was a women’s movement.

🔨 The steep decline in Ethereum’s price has crypto miners dumping their GPUs on eBay.

A Chinese shopping mall has introduced “husband storage”: "There's no ventilation or air conditioning, I sat playing for five minutes and was drenched in sweat."

🥂 Chip Hall of Fame: celebrating the best-of-the-best integrated circuits.

🚀 Insulin resistance, not calories or cholesterol, is the key predictor of CVD and Type-2 diabetes.

Business leaders should read more science fiction. Investors should read more novels.

Bees through a macro lens. 🐝

🌤️ When is your best chance to see the total solar eclipse?



I'm glad you all enjoyed last week's guest issue from Saul Klein and his team at LocalGlobe. We've got a couple more guest curators coming up when I take my family vacation.

Exponential View is a whisker away from 20,000 subscribers, so please take a moment to forward this email to a few colleagues with a personal recommendation or hit the tweet button.


P.S. If you want to join the sponsor roster from the fall, drop Marija en email.

This week's issue is brought to you with support from our partner, WorkShape


Helping companies hire software engineers


Sign in or become a Exponential View member to join the conversation.