Sunday, April 29, 2007

Book Review: Batchmode by Michel Creton

Batchmode is a complicated techno-financio-thriller about the rise and fall of Thoughtspeed, a Silicon Valley startup leading a renaissance in artificial intelligence. I think Michel Creton is sort of a French Canadian Michael Crichton.

The founder of Thoughtspeed is a very high-powered experienced start-up expert (this is his fourteenth company, although he is not even that old). The founder has a brilliant friend who comes up with a technical plan to achieve artificial intelligence. The brilliant guy becomes the Chief Scientist of Thoughtspeed.

The company gets funded and they hire a few dozen Ph.D.s. The first big challenge comes when they start having a serious HR problem. One of the scientists reads through the business plan and suddenly starts acting like a loose cannon. This gets a little technical, but the guy insists that the metric prefix exa- really means 10 to the 18th power. Not 10 to the 9th power. But the Chief Scientist, who is a brilliant guy, put into the business plan that exa = 10 to the 9th power. So it's obviously mission critical to maintain that and not get distracted. And so people try to help the loose cannon guy get with the program but he just gets more and more shrill and crazy-sounding. Management tries to gently help him out by suggesting that he shouldn't be so bothered by a little bit of extra work, but he tries to shift the blame back on them: "10 to the 9th extra work? Do you understand what that means? You want people to work 40 billion hours a week?" Whatever. We've all known people like that, who let their own egos or laziness stand in the way of new ideas. Anyway, in a typical dumb career move the loose cannon guy quits in a huff.

BUT, in a startling plot twist several months later, the others discover that the loose cannon guy was actually right! So everything is off by 9. (Disclosure: I'm kind of a programming "geek" myself, and even being off by 1 can be a pretty big deal in computer science. So off by 9 is serious, trust me!)

The founder is concerned that the venture capitalists may get confused if he admits the mistake now. "The best thing for everybody is we go ahead with the demo," he explains. And it turns out the VCs are a lot better off not knowing, because they sell most of their stake to small investors in an initial public offering later on. So in the end the loose cannon behavior doesn't hurt anybody.

Fast forward to three years later. The scientists and engineers at Thoughtspeed have created a system that is functionally equivalent to the human brain. With great ceremony the CEO pushes "GO" and mankind literally enters a new era.

But the A.I. program is slower than originally planned. In fact, after two more years it has still only said "Uh...." By that time, money for all those expensive technical people has run out. So pretty much everybody gets laid off. In a really moving scene, an old security guard conveys management's heartfelt thanks for all the long hours. He even breaks down and admits that due to budgetary constraints he is required to escort himself out of the building once everyone else leaves.

And then before long unfortunately the execs have to pull the plug and give back all the rented cubicles and servers and stuff. Some funny moments when the Chief Scientist's necktie gets caught under a crumbling chipboard desk and he almost gets dragged onto Route 880 behind a Ryder van. "Holy shit, not again!" laughs the founder. "You want some advice? No matter how bad you need to take a leak, don't ever whip it out anywhere near a car door!" It sounds like startups must be a lot of fun.

But overall the book was kind of a rip-off. My main reason for buying the book was curiosity about what an artificial intelligence would really be like, personality-wise. But we never really find out. The author was too focussed on his techno-fetishes.

4 stars out of 11. Wait for the real thing.
PW Science Staff


Post a Comment

<< Home