The Real AI Threat?
rsk at gsp.org
Thu Dec 10 08:35:17 UTC 2020
On Thu, Dec 10, 2020 at 12:34:33AM +0000, Mel Beckman wrote:
> So don???t be fooled by Siri and Google voice response. There is no
> intellect there, only pattern matching. Which we???ve been doing with
> machines since the Jacquard Loom.
On this particular point: many years ago, some of us at Purdue discussed
this at great length and eventually coined the term "ad-hockery" --
which found its way into the New Hacker's Dictionary. The gist of
the idea is that it's possible to craft a sufficient number of ad hoc
rules, plug them into a pattern matcher, and present a modestly plausible
appearance of "intelligence"... when in fact nothing resembling actual
intelligence is involved. Such systems are brittle and demonstrate
it when presented with input not covered by those ad hoc rules, which
is why they're often made progressively less so by repeated tweaking.
(Also known as "release 3.0" and accompanied by prose touting it as
an innovative upgrade.)
But to borrow Mel's phrasing, even a very large collection of ad hoc
rules that performs its task tolerably well is no more intelligent than
More information about the NANOG