Curated Insights 2017.08.27

Inside Waymo’s secret world for training self-driving cars

Collectively, they now drive 8 million miles per day in the virtual world. In 2016, they logged 2.5 billion virtual miles versus a little over 3 million miles by Google’s IRL self-driving cars that run on public roads. And crucially, the virtual miles focus on what Waymo people invariably call “interesting” miles in which they might learn something new. These are not boring highway commuter miles.

And in both kinds of real-world testing, their cars capture enough data to create full digital recreations at any point in the future. In that virtual space, they can unhitch from the limits of real life and create thousands of variations of any single scenario, and then run a digital car through all of them. As the driving software improves, it’s downloaded back into the physical cars, which can drive more and harder miles, and the loop begins again.

Not surprisingly, the hardest thing to simulate is the behavior of the other people. It’s like the old parental saw: “I’m not worried about you driving. I’m worried about the other people on the road.”

“Right now, you can almost measure the sophistication of an autonomy team—a drone team, a car team—by how seriously they take simulation. And Waymo is at the very top, the most sophisticated.”

And in reality, those 20,000 scenarios only represent a fraction of the total scenarios that Waymo has tested. They’re just what’s been created from structured tests. They have even more scenarios than that derived from public driving and imagination. “They are doing really well,” Peng said. “They are far ahead of everyone else in terms of Level Four,” using the jargon shorthand for full autonomy in a car.


Halliburton and Microsoft do not compute for OPEC

Contrary to conventional wisdom, the oil business is a high-tech one. You don’t map out complex rock formations thousands of feet beneath the ground in three or four dimensions and then drill into them without advanced tools. For example, Total SA, the French oil major, boasts the 19th-most powerful supercomputer in the world, Pangea, which has clocked a speed of more than 5 quadrillion calculations a second (technical note: pretty fast) according to The Top 500 List.

…the average well has less than 10 gigabytes of data associated with it, equivalent to a couple of high-definition movies.

…the company’s teams now have access to more than 80 real-time data streams and sensors embedded in wells, giving them a constantly updated picture of what’s happening beneath the ground. The killer app here, in every sense of the word, is providing crews and their managers with an integrated platform; a suite of sensors and software communicating seamlessly, updating constantly and available to all involved.


Costco is playing a dangerous game with the web

Costco’s reluctance to embrace the web is understandable. Its warehouse club business model is based on selling a limited assortment of bulk-size food and household items at low prices, alongside an ever-changing selection of general merchandise—everything from margarita machines to kayaks. This creates an in-store treasure hunt experience. Both elements are costly and difficult to replicate online.

Potentially more worrisome: Half of Costco’s shoppers are Amazon Prime members, Kantar Retail says, up from 14 percent five years ago. Sharing too many of the same subscribers could be risky, since Planet Retail RNG analyst Graham Hotchkiss says Amazon now offers many bulk-size goods at prices that rival Costco’s. And Amazon’s pending $13.7 billion deal to buy Whole Foods Market Inc. will give it a firm foothold in groceries—the primary reason people shop at Costco, according to Barclays’s Short.


What is Amazon, really?

At last count, Amazon’s delivery infrastructure included more than 180 warehouses, 28 sorting centers, 59 local package delivery stations, and 65 hubs for its two-hour Prime Now deliveries. Investment bank Piper Jaffray estimates that 44% of the US population lives within 20 miles of an Amazon warehouse or delivery station. Amazon’s proposed $13.7 billion acquisition of Whole Foods could add another 431 distribution nodes in bougie neighborhoods to that network.

“Our goal with Amazon Prime, make no mistake, is to make sure that if you are not a Prime member, you are being irresponsible,” Bezos told shareholders in May. The plan is working: 63% of US Amazon users subscribe to Prime, and estimated to reach more than half of American households by the end of the year. Prime doesn’t just lift $99 off of regular Amazon users each year—it’s proven to be a powerful customer loyalty program. The average Prime user spends $1,300 each year on the site, with 78% of Prime users still citing free 2-day shipping as the main reason for coughing up the fee.

The service first reached customers by 2005, and was officially launched in the summer of 2006. Tom Szkutak, Amazon’s CFO at the time, said the business was “exposing the guts of Amazon,” using the knowledge gained from 11 years of building Amazon.com. Today AWS is on a tear. It’s the world’s dominant cloud computing provider, and the nearest competitors aren’t even within shouting distance: Amazon’s servers deliver 34% of the world’s public cloud services, reports Synergy Research Group, while Microsoft, IBM and Google provide 24% combined.


Amazon vs Maersk: The clash of titans shaking the container industry

Manufacturing is new step for Amazon and they won a patent earlier this year to develop a system to rapidly create clothing and other products after a customer order is placed. This forms a cheap and simple method for Chinese exporters as Amazon have effectively wiped out the middle man, acting as a shipbroker for itself and on behalf of smaller companies.

Freight forwarders may find it hard to compete with companies as powerful as Amazon and Maersk, who can afford to develop disruptive technology and prioritize increasing market share over higher profits.

Small independent ship owners will be left behind unless they adapt their business model to seek different shipping routes, for example choosing container lanes that do not feed into deep sea ports where the ultra large container vessels operated by Maersk can only dock.


Great Wall Motor’s better path leads to emerging markets

The Proton purchase not only gives Geely inroads to the Malaysian consumer market but also access to production plants in the region that could be used to manufacture other car brands. Being closer to the end customer would lower production costs. While Proton’s Tanjung Malim plant has the capacity to churn out one million cars annually, it made only 72,000 last year, according to the Malaysian government.

A lean company, analysts estimate Great Wall makes 60 percent of its parts in-house. It spends little on marketing and is the fourth-most-profitable automaker globally by net margin and return on equity. Its return on invested capital ranks number one among 40 manufacturers tracked by Bloomberg Intelligence.


A handful of companies control almost everything we buy — and beer is the latest victim

A whopping 182 beauty brands fall under the massive umbrellas of seven huge manufacturers: Estée Lauder Companies, L’Oréal, Unilever, Procter & Gamble, Shiseido, Johnson and Johnson, and Coty.

A 2015 Morgan Stanley report found that 10 companies controlled 41% of the clothing market. No other retailer had more than 2% of market share. The retailers dominating the market were Walmart, T.J. Maxx, Macy’s, Gap, Kohl’s, Target, Ross Stores, Amazon, Nordstrom, and J.C. Penney.

According to a Bank of America Merrill Lynch chart, in 2014 AB InBev and SABMiller alone controlled about 58% of the beer industry’s $33 billion in global profits.

Winner-takes all effects in autonomous cars

… it seems pretty clear that the hardware and sensors for autonomy will be commodities. There is plenty of science and engineering in these (and a lot more work to do), just as there is in, say, LCD screens, but there is no reason why you have to use one rather than another just because everyone else is. There are strong manufacturing scale effects, but no network effect. So, LIDAR, for example, will go from a ‘spinning KFC bucket’ that costs $50k to a small solid-state widget at a few hundred dollars or less, and there will be winners within that segment, but there’s no network effect, while winning LIDAR doesn’t give leverage at other layers of the stack (unless you get a monopoly), anymore than than making the best image sensors (and selling them to Apple) helps Sony’s smartphone business. In the same way, it’s likely that batteries (and motors and battery/motor control) will be as much of a commodity as RAM is today – again, scale, lots of science and perhaps some winners within each category, but no broader leverage.

Maps have network effects. When any autonomous car drives down a pre-mapped road, it is both comparing the road to the map and updating the map: every AV can also be a survey car. If you have sold 500,000 AVs and someone else has only sold 10,000, your maps will be updated more often and be more accurate, and so your cars will have less chance of encountering something totally new and unexpected and getting confused. The more cars you sell the better all of your cars are – the definition of a network effect.

The more real world driving data that you have, the more accurate you can make your simulation and therefore the better you can make your software. There are also clear scale advantages to simulation, in how much computing resource you can afford to devote to this, how many people you have working on it, and how much institutional expertise you have in large computing projects. Being part of Google clearly gives Waymo an advantage: it reports driving 25,000 ‘real’ autonomous miles each week, but also one billion simulated miles in 2016 (an average of 19 million miles a week).

So, the network effects – the winner-takes-all effects – are in data: in driving data and in maps. This prompts two questions: who gets that data, and how much do you need?

This leads me to the final question: how much data do you really need? Does the system get better more or less indefinitely as you add more data, or is there an S-Curve – is there a point at which adding more data has diminishing returns? That is – how strong is the network effect?


The stereo speaker company giving sight to self-driving cars

Although it will soon face plenty of competition, Velodyne has become the industry’s go-to lidar supplier and is cranking up production to match. Last year, Ford Motor Co. and Chinese Internet giant Baidu pumped $150 million into Velodyne, money the company used to open its “mega-factory” on San Jose’s southern edge.

Lidar works by firing laser beams — thousands per second — at nearby objects and measuring how quickly they bounce back. With the notable exception of Tesla, most companies pursuing autonomous vehicles rely on lidar, along with radar and cameras.

“The prevailing view is that in the near term — at least a decade — you’re not going to be able to execute this safely without lidar,” said Mike Ramsey, research director at Gartner.

“One major automaker told me they had vetted 50 lidar companies,” Ramsey said. “So more than 50 companies exist, but only Velodyne is producing a lidar they can use.”

Now, the race is to cut lidar’s cost. Velodyne’s most popular lidar, about the size of two stacked hockey pucks, sells for $8,000. As it ramps up production, the company hopes to bring prices down to “a few hundred dollars,” Hall said. “We’re in the inventing business, so we’re going to keep working on this thing until we crack that nut.”


The internal combustion engine is not dead yet

Mazda, which now markets no hybrid vehicles, calls the engine Skyactiv-X and says it is scheduled for a 2019 introduction. In simplest terms, the big difference with the new engine is that under certain running conditions, the gasoline is ignited without the use of spark plugs. Instead, combustion is set off by the extreme heat in the cylinder that results from the piston inside the engine traveling upward and compressing air trapped inside, the same method diesel engines use. The efficiency gains come with the ability to operate using a very lean mixture — very little gas for the amount of air — that a typical spark-ignition engine cannot burn cleanly.

…addresses the challenge of gasoline’s future from a somewhat different direction: the practical limitations of battery electric cars. “Holding a gas nozzle, you can transfer 10 megawatts of energy in five minutes,” he said, explaining today’s refueling reality. To recharge a Tesla electric at that rate today, he said, would require “a cable you couldn’t hold.”

By 2050, Dr. Heywood’s studies project, today’s fuel economy could be doubled. “A quarter to a third of that improvement would come from improvements to the vehicle,” he said, in areas like aerodynamics and weight reduction. Other promising areas include variable compression ratios — a technology Nissan plans to introduce next year — and making better use of available fuels.


Wind power is all grown up now

People tend to think of renewable energy companies as the new kids on the block but Vestas Wind Systems A/S, the world’s biggest wind turbine manufacturer, is no pimply teenager. The Danish group entered the turbine business almost 40 years ago and went public in 1998…

The wind industry is consolidating — Siemens merged its wind business with Gamesa in April — and competition is intensifying. This puts pressure on margins and makes it more difficult to lift revenue.

A bigger concern is that more countries are adopting auction-based contract awards. These promote projects that deliver the cheapest electricity as opposed to feed-in tariffs, which guaranteed a fixed electricity price. So life’s getting a little tougher for Vestas.

There are other ways to make money. High-margin maintenance contracts are an increasing share of business. There are opportunities too to upgrade the installed base with those newer, better turbines.


The very symbolic collision of Sotheby’s-Christie’s and Poly-Guardian in China art

If Sotheby’s and Christie’s are purely commercial Giants, then Poly Culture and Guardian are something else. They are certainly Giants, dominating the domestic art auction industry. But Poly in particular is also a direct extensions of the State. Because it turns out, what happens to historic Chinese art is a significant concern to the Chinese government. Part of this sensitivity is about repatriating works that were stolen and misappropriated over the centuries. Many of the works that have been returned can be seen on display at Poly’s headquarters in Beijing.

Poly Auction is now not just one of the top two auction houses in China. It is also the number three art auction house in the world (after Christie’s and Sotheby’s). Their 2013 turnover was over a billion dollars (about one-fourth of Sotheby’s). They sell approximately 10,000 objects each week, with as many as 40 different catalogs per show.

Because at the same time, Poly and Guardian have been expanding internationally. And they are now on Sotheby’s and Christie’s home turf for the first time. Both have moved into Hong Kong. And Poly is now moving aggressively into New York City, where Sotheby’s is headquartered. Thus far, they have focused mostly on finding consignments in the US for sale in China, particularly Chinese collectibles. But Poly’s openly stated ambition is to become the world’s top art auction house. According to CEO Jiang Yingchun, “We are very big in the art auction market in Mainland China but still have a long way to go to become the biggest auction house worldwide”.


It’s hard to keep up with all that lithium demand

Australia is the biggest lithium producer, though Chile and Argentina account for 67 percent of global reserves, according to the U.S. Geological Survey.

Extracting lithium from the salt flats that dot the arid northern regions of the South American countries is a lot easier and cheaper than digging underground for metals like copper. Producers just pump the brine solution into evaporation ponds, harvesting the mineral once the moisture is gone.

With demand expected to keep rising as electric cars gain a bigger share of the global auto fleet, Argentina and Chile are attracting interest from mining companies because it costs about $2,000 to $3,800 a ton to extract lithium from brine, compared with $4,000 to $6,000 a ton in Australia, where lithium is mined from rock.

Of the 39 lithium ventures tracked by CRU, only four have firm commitments, and all of those are in China, adding about 24,000 tons of annual supply. Another 10 projects representing 400,000 tons are rated “probable” — in Canada, Chile, China, Mexico, Argentina and Australia — but probably only about 30 percent will make it into production, CRU said.

“But we have a window of only 25 years to develop these projects because prices can fall again as soon as a replacement to lithium appears.”


Hunt for next electric-car commodity quickens as prices soar

As one of the key components in the new breed of rechargeable batteries and with supply dominated by the Democratic Republic of Congo, prices have surged at four times the pace of major metals in the past year.

The cobalt market is in a 5,500-ton deficit, according to CRU, with global supply contracting 3.9 percent in 2016.

“The mix of iron and cobalt is tricky. Cobalt is already mined as a byproduct of copper and nickel, but iron has the most negative impact on cobalt, which means processing would be more difficult and more expensive.”

Aging Japan wants automation, not immigration

In the absence of large-scale immigration, the only viable solution for many domestic industries is to plow money into robots and information technology more generally.

With unemployment down to 2.8 percent, companies are increasingly realizing they need to pay up to attract and keep qualified personnel. The other option — increased immigration — is politically difficult.

Bank of America Merrill Lynch forecast IT investment in Japan to rise as much as 9 percent annually in coming years, with the difference in software investment per worker versus the U.S. falling to 5 to 1 by 2020 from about 10 to 1 now.


Who really owns American farmland?

Farmland, the Economist announced in 2014, had outperformed most asset classes for the previous 20 years, delivering average U.S. returns of 12 percent a year with low volatility.

Today, the USDA estimates that at least 30 percent of American farmland is owned by non-operators who lease it out to farmers. And with a median age for the American farmer of about 55, it is anticipated that in the next five years, some 92,000,000 acres will change hands, with much of it passing to investors rather than traditional farmers.

It’s a tenuous predicament, growing low-cost food, feed, and fuel (corn-based ethanol) on ever-more-expensive land, and it raises a host of questions. Is this a sustainable situation? What happens to small farmers? And are we looking at a bubble that will burst?

In practice, our best hope of true stewardship of the land will come from enlightened, committed owner-farmers. But the trend toward treating farmland as a financial investment, and the high prices that have come with it, make it harder and harder for new young farmers to enter the field.

By buying land in other countries and farming it, foreign buyers are able to support their domestic food supply and other markets that depend on agriculture without having to compete for essential products on the global market.

The government of China now controls more than 400 American farms consisting of a hundred thousand acres of farmland, with at least 50,000 in Missouri alone, plus CAFOs (concentrated animal feeding operations), 33 processing plants, the distribution system—and one out of every four American hogs.


Yesterday’s “plastics” are today’s crypto tokens

Our ability to profit from our investments relies on two things: having the resources needed to purchase the asset and then having a way to sell it — a concept known as liquidity. Tokenizing real-world assets will allow buyers to access assets never before within their reach, and sellers to move assets that were previously difficult to unload. The secret lies in the possibility of fractionalization.

Imagine unlocking cash from the equity in your home without having to borrow or pay interest. Tokenize your home and sell fractions to the public. Buy the tokens back, or pay the investors their value at the time the property is sold.

In the future, you’ll be able to tokenize the value of unused bedrooms and backyards in your home. You’ll be able to tokenize use of your vehicle for Uber driving while you’re away on travel. You’ll even be able to tokenize access to your phone so marketers have to pay you tokens in order to gain access to your attention. Yes, this will happen.

At 5,4000 pounds, the Tesla is no lightweight, and the Aventador is a good 1,200 pounds lighter. But the Ludicrous + enabled vehicle not only beats the Lambo, it sets a world record for the quickest SUV with a quarter mile time clocking in at 11.418 seconds at nearly 118 miles-per-hour.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.