Troyale

Troyale

Share this post

Troyale
Troyale
The Defense Department Created Siri

The Defense Department Created Siri

Troyale's avatar
Troyale
Jun 06, 2025
∙ Paid
Share

Sort of.

By the early 2000s, artificial intelligence research had narrowed to focus on specialized tools with a very limited scope. Voice-activated systems could transcribe dictated speech, and search engines could gather and curate results matching typed requests, but nothing out there could interpret intent, learn from user behavior, or perform tasks with any level of complexity. Within the many cubicles and offices of private research agencies and corporate-run tech labs, the concept of a digital assistant that could understand, reason, and act independently was generally considered to be unrealistic, at least at the time. The American Defense Department felt differently, however.

Through its experimental research department DARPA (Defense Advanced Research Projects Agency) - created directly in response to the Sputnik satellite launch in 1957 - they launched a $150 million inititiative designed to create an AI that could support military decision-making they’d eventually end up calling the ‘Personalized Assistant that Learns,’ or PAL. More specifically, its purpose was to reduce the logistical burden on command staff by letting computers handle tedious administrative and data-intensive tasks. Busy work, essentially.

The Agency ultimately awarded the development contract to SRI, or the Stanford Research Institute, a then-independent nonprofit organization originally founded by a group of trustees from the university in 1946.


SRI engineer Doug Englebart invented the original computer mouse in 1963.


To serve as the metaphorical spine, nervous system, and brain of the PAL program, SRI created the ‘Cognitive Assistant that Learns and Organizes,’ or CALO, a desktop-based system designed to function like a digital secretary, observing its user’s behavior and then adapting to help them complete tasks more efficiently. It could monitor and analyze emails, manage digital calendars, and search the internet like any traditional search engine. Unlike other AI’s of the day, it didn’t have a static training set or database loaded with fixed patterns of pre-programmed behavior. It ran in real time.

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Stephen J. Hunt
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture

Share