Smart vs. Seeing

— In 1975 a Kodak engineer brought his bosses the first digital camera. They told him: "that's cute — but don't tell anyone about it." Six years later Kodak commissioned a big internal study — and it predicted the death of film absolutely correctly. And on the basis of that correct forecast they made a flawlessly logical decision: use digital to improve film. Smart people, sound analysis, accurate numbers. Thirty years later — bankruptcy. That isn't a different mark on the scale. That's a different operation.

— They knew and they did the math. Their head of market intelligence put two curves together: how fast sensors get cheaper and how much the silver in film costs. The conclusion came out double. Bad news — digital will eat film. Good news — we have about 10 years. The forecast turned out to be nearly flawless; they were off by a couple of years.

— Yes. And it isn't stupidity — try coming up with something better right now. You have a profitable business, a billion dollars of margin a year, a hundred-odd thousand employees, and a new technology that so far gives you a black-and-white frame of one hundredth of a megapixel and takes 23 seconds to write it. What do you do? You reinforce what works at the expense of what doesn't work yet. It's the best move. Just on a different board.

— That's how it looked to me too, for about 15 years. Then I looked at how it ended, and the picture flipped. In 2005 Kodak was number one in unit sales of digital cameras in America.

— Number one. Twenty-something percent of the market, the digital line grew maybe 40 percent in a year. And according to the Harvard school's analysis they were losing about 60 dollars on every camera they sold. Seven years later — Chapter 11, bankruptcy. Do you see the trick? They didn't lose the game they were playing. They won it.

— Much nastier. And here's a second one just like it. In the seventies Xerox invented not one thing but the entire modern computer at once: windows, the mouse, editing that looks like paper, Ethernet, the laser printer, the workstation. Out of that whole list Xerox took exactly one item.

— The printer. And made billions on it. That's not stupidity, that's a direct hit. The laser printer translated into their language: we're a copier company, a printer is a copier in reverse. Windows and the mouse didn't translate into that language at all. It isn't even a "bad idea." It isn't an idea — they had no category to put it in.

— They tried. It came out in 1981 at 16,500 dollars, they sold maybe 25,000 units. A failure. And management read that failure absolutely correctly inside their own frame: so personal computers aren't our business. The data confirmed the premise. And now the detail that still unsettles me. Jobs didn't get into their lab for free.

— He gave Xerox the right to buy 100,000 pre-IPO Apple shares at 10 dollars. A year later Apple went public at 22. Which means that inside the frame "venture investment" Xerox made a good deal and turned a profit. It only lost on the scale it didn't have.

— There are, and this one is the limit case, because there the intelligence was certified. The hedge fund LTCM, with two Nobel laureates in economics on the board.

— For exactly the work the fund went broke on. The prize was awarded in October 1997. Minus 4.5 billion — in August and September of 1998. Eleven months. And the mathematics in their models was correct; people checked it line by line afterwards. It's just that the model contained an assumption that the market would stay a market, and the history it was trained on was about five years long. What actually happened simply didn't exist inside that window.

— The error was made before the calculation started. And that's the most important thing I want to say in the first ten minutes. From inside the calculation it cannot be detected, in principle. Picture it: you're driving on your navigation app, it works perfectly, routes around a jam, saves you seven minutes, warns you about a speed camera. You arrive exactly on time at the right house in the wrong city, because when you typed the address you tapped the first line of the dropdown. The app isn't at fault and cannot be at fault. For it the address isn't a hypothesis, it's an input.

— Exactly. And now look at where all this is held in place. An exam, an IQ test, an olympiad, a job interview — what do they have in common?

— The given. In all of them the problem statement is handed down from above. Your whole life you're graded on the quality of your move given a fixed problem, and never once on the choice of problem. We built an entire civilization for measuring moves and never built a single instrument for measuring the choice of board. The only attempt I know of is an old study at an art institute: students were given 27 objects and an hour for a still life. Some started drawing right away, others spent a long time rearranging and reshuffling. Seven years later, and then 11 years after that, the second group turned out to be more successful as artists. That's it. That's the only measurement I know of, and it's about drawing.