Brain dumping

INTEL’S ANDY GROVE says Moore’s Law has hit the wall. Good. We don’t need computers that execute an empty loop twice as fast as last year’s models. We need software that takes our fast computers beyond their roles as glamorized typewriters and ledgers.

In academia and in vendors’ research labs, computers are being taught to pull the meaning out of text, handwriting, and speech. Software that comprehends, rather than merely records, will make this decade’s headlines. Businesses must think about what projects the smart software of the future will work on.

Metaresearch digs knowledge that the original authors never considered out of assembled documents. Such efforts have changed medical treatment approaches and altered the application of law. Today, technology only assists the metaresearch process; the real analysis is done by humans and takes years to complete. But the raw material is already in digital form and much of it is encoded according to standards. Imagine what we’ll learn from these vast databases when computers can understand them.

If you allow intelligent software to mull over every bit of data you poured into your company’s computers, what new insights would emerge? Perhaps not many. Most of us are downright stingy when it comes to sharing knowledge with our computers. Typed information potentially lives forever, but essential supporting data exists only in transient forms: phone calls, office discussions, scribbled notes, and thoughts never shared.

Two things keep me from pouring all my thoughts and experiences into my computer. First, I’m concerned about how that information might be used. U.S. Sen. Trent Lott said something about segregation at a recent event feting Strom Thurmond; reporters’ Nexis searches quickly unearthed a similar statement Lott made some 20 years ago. That turned what might have been a lapse in judgment into an apparent expression of long-held (and backward) principles.

If Lott’s private thoughts were also part of the record, maybe they’d exonerate him. Maybe they’d dig him a deeper hole. It would probably depend on the person guiding the analysis. I don’t want to build the database that could be used to harm me later on, yet not taking that risk denies me the benefit of my own knowledge and experience.

I encode only what my computers can interpret right now. I’d think it a poor use of my scarce time to record and store my phone calls on disk or to scan my handwritten notes. As a result, a fraction — maybe 5 percent — of what I know, think, and experience ends up encoded as Word documents. Even those are scattered and easily lost.

Then again, maybe the phone call I was too lazy to record or the digital photo I deleted to save 100K of disk space would complete some puzzle down the road. Disk space is cheap. Since I can’t know now what will be important in five, 10, or 20 years, I’m inclined to find ways to capture all the knowledge I can. Maybe there are enlightenments hiding in the data we’re throwing away.