A brief terminology adjustment

“Rationalists” are now automatists. Please update your equipment accordingly.

I like the word “automatist” because it reinforces the connection between the idea that reason can be reduced or transcended, and the discipline of constructing artificial reasoning devices, also known as AI. I don’t think it’s too hard to see how the desire to build AI could generate a bias in favor of formulaic epistemologies.

One book on AI I can recommend is HP Newquist’s The Brain Makers. Newquist’s book came out in 1994, and he was writing about the 1980s generation of AI, which focused its efforts on symbolic logic, very different from this new Bayesian revival. The collapse of ’80s AI, sometimes known as the AI winter, after a decade-long avalanche of pure sickening hype, should have taught us a permanent lesson. Sadly, humans are not good at permanent lessons.

My own involvement with ’80s AI was quite tangential—my first job ever, at age 15, involved translating an expert-system engine from Basic to C, on the Wang VS. Since I didn’t know C at the time, and since I was using the first C compiler ever released for the VS, there was a great deal of merriment. But I also wondered: who the hell would use this product? And for what? The answer was no one, of course, though since my employer specialized in government work, this may not have even mattered.

But much later I had a coworker who was a veteran of Intellicorp—(Andi? Are you out there, Andi?)—and, after borrowing Newquist’s book, he confirmed its general accuracy. I trust him, he trusts Newquist, and you trust me, and since you can get The Brain Makers for only a quarter plus shipping and handling, if you give a rat’s ass at all about AI you should order it.

Of course the contents are industry dirt, not actual technical information. But as Boltzmann proved, information is fragile whereas dirt lasts forever. Why read today’s dreamers, when you can read about yesterday’s failures? It’s the difference between grape juice and wine.