THE COMPUTER CHRONICLES: AI BEFORE THE CHAT WINDOW
These notes cover three AI episodes of The Computer Chronicles:
Artificial Intelligence from 1984, another program under that title
cataloged in 1986, and Neural Networks from 1991.
[1, 4, 5]

1984: PUTTING KNOWLEDGE INTO A PROGRAM

The 1984 episode brings together Edward Feigenbaum, Nils Nilsson, John
McCarthy, Tom Kehler, and guest co-host Herb Lechner. Its archive
description identifies knowledge engineering and Lisp among the
subjects. [1]

The opening uses computer chess to raise a larger question about
intelligence. Lechner distinguishes reproducing useful results from
reproducing human thought. An oil-drilling advisory system then makes
that distinction concrete. It questions the operator and offers
advice; when asked why a question matters, it explains the reasoning
behind it. [2]

Another demonstration is especially approachable in plain text.
Jeffrey Perrone shows a system built from examples of decisions. An
airway-diagnosis example is followed by a deliberately simple
Sunday-afternoon planner. Weather and family company become inputs;
the suggested outing becomes the result. The presenter enters examples
rather than explicitly writing every rule. [2]

These demonstrations give the term "knowledge engineering" some
substance. Somebody must decide which distinctions matter, supply the
examples, and determine what the result means. The small Sunday
example exposes those choices clearly.

Kehler's own account identifies his contribution as a demonstration of
KEE, the Knowledge Engineering Environment, applied to a nuclear
reactor diagnostic. He places it around fourteen minutes into the
program. That firsthand description supplies a more specific name and
purpose than the archive's short catalog entry. [3]

1991: LEARNING FROM EXAMPLES ON THE DESKTOP

Neural Networks is a useful next stop. Its catalog lists Braincel 1.1,
BrainMaker Professional 2.0, MacBrain 3.0, NeuroSMARTS, and
ExploreNet, alongside visits to NASA and Intel. [5]

The Braincel demonstration starts inside a spreadsheet. The example
concerns estimating new construction from economic data, for a
business selling heating and air-conditioning equipment. The presenter
describes a situation in which the user has observations but lacks a
convenient equation. Training a network supplies a relationship the
spreadsheet can use. [6]

The demonstration sets aside two years of data for testing, so the
network is checked on examples it wasn't trained on. [6]

The NASA report describes research on recognizing aircraft even
when their positions differ from the training examples. These are
particular recognition tasks, with an intended application and a
testable result. 
[6]

ANOTHER EPISODE TO LOOK UP

The later Artificial Intelligence program, cataloged in 1986, offers a
companion to these two. Its listed products include Texas Instruments'
Personal Consultant and Symantec's Q&A; the subjects are expert
systems, decision support, and natural-language communication. [4]

FOLLOW THE KNOWLEDGE

While watching, ask what information goes into each system and how the
presenters check its answers.

SOURCE NOTE

These summaries draw on archive descriptions, automatic captions for
the 1984 and 1991 episodes, and Kehler's account of his demonstration.
The captions can contain errors. The summaries describe the
demonstrations, not independent tests of the products. The 1986
description comes from the archive catalog, which dates the episode
January 7, 1986. Links below require an HTTPS-capable client.

REFERENCES

[1] VIDEO / WEB: Artificial Intelligence (1984), Internet Archive.
https://archive.org/details/CC1024_artificial_intelligence

[2] WEB / TEXT: Machine-generated captions for the 1984 episode.
https://archive.org/download/CC1024_artificial_intelligence/CC1024_artificial_intelligence.asr.srt

[3] WEB: Thomas Kehler, account of his KEE demonstration.
https://thomaskehler.com/

[4] VIDEO / WEB: Artificial Intelligence, cataloged January 7, 1986.
https://archive.org/details/Artifici1986

[5] VIDEO / WEB: Neural Networks (1991), Internet Archive.
https://archive.org/details/neuralnetwor

[6] WEB / TEXT: Machine-generated captions for Neural Networks.
https://archive.org/download/neuralnetwor/neuralnetwor.asr.srt