Kamprad was known for driving an old Volvo, recycling tea bags and taking home little packets of salt and pepper from restaurant visits. He was known as “Uncle Scrooge” and “The Miser” in the Swiss village of Epalinges, near Lausanne, where he moved in the 1970s before returning to Sweden a few years ago. He also avoided wearing suits and ties and traveled coach when flying.
Ikea’s corporate culture mirrors Kamprad’s celebration of frugality. Executives of the company travel on low-cost airlines and lodge in budget hotels. Its employees follow a basic pamphlet written by Kamprad in 1976, “The Testament of a Furniture Dealer,” which states that “wasting resources is a mortal sin,” and stipulates Ikea’s “duty to expand.”
The name Ikea is made up of the founder’s initials and the first letters of the Elmtaryd farm and Agunnaryd village where he was raised. His flat-pack furniture was invented by Ikea employee Gillis Lundgren in 1956 when he tried to fit a table into the back of a car. Realizing the table was too bulky, Lundgren removed the legs. Storing and selling Billy book shelves or entire kitchens in pieces has let Ikea cut storage space and fill its trucks with more goods. The concept of having customers pick up most of their own furniture in adjacent warehouses and transport it home for self-assembly also helped drive down costs.
But Amazon has a huge set of data that Facebook and Google can’t access—namely, its own. Already, more than half of all online searches for products start on Amazon, and of those a majority end there, according to various surveys. That figure has grown every year that pollsters have tracked it.
The Amazon Advertising Platform lets advertisers manage ad buys across multiple advertising exchanges, and it has quietly become as familiar to marketers as its equivalent from Google-owned DoubleClick.
Amazon also needs to expand the number of places it can sell advertisements, which is one reason the company bought videogame-streaming behemoth Twitch and is investing so heavily in its own streaming-video offerings.
…by designing its own chips, Apple cuts component costs, gets an early jump on future features because it controls research and development and keeps secrets away from frenemies such as Samsung…Those ultimately failed or stumbled because chip-making is the sport of kings: It’s brutally expensive and requires massive scale. Apple has wisely focused on designing its silicon (for its system on a chips, Apple uses reference designs from Arm Holdings Plc). Manufacturing is left to others, including Taiwan Semiconductor Manufacturing Co.
Over 40% of Standard & Poor’s 500 revenues now comes from abroad.
No other country is shrinking its equity base to the extent we are. We’re now in our ninth year of share buybacks equal to 3% of the market value of all S&P 500 stocks, based on Laszlo Birinyi’s work.
For 20 years, the average price/earnings ratio has been 19.3. If you go back 50 years, it’s 15.6 times. In periods where inflation grew 3% or less—which is 22 of the past 50 years—the P/E of the market was 19.7.
Using self-play to recursively improve an agent’s ability to play a game isn’t new. Why hasn’t this method yielded a champion chess or Go engine until 2017? Historically, systems that improve via self-play have been very unstable. Previous attempts often ended up in cycles, forgetting and relearning strategies over and over rather than improving to superhuman levels. Or sometimes the agent would get stuck, failing to improve after achieving moderate success.
AlphaZero’s main contribution was solving these problems. After lots of experiments, DeepMind developed a series of new tricks and discovered a value function and tree search that reliably learned through self-play alone. They then leveraged their engineering talent and infrastructure resources to demonstrate that the system could work on the massive scale required to master complicated games such as chess and Go (the version that played Stockfish employed 5,000 custom machine learning chips).
One of the (many) reasons I stopped heeding the macro forecasts of others and quit trying to come up with my own is that even if you knew what the future data would be, you’d still not be able to predict how people would react to it. You could certainly try, but markets are set up to confound us, not confirm our hypotheses.