In the summer of 1958, Frank Rosenblatt (a Cornell Aeronautical Laboratory researcher) demonstrated his “perceptron” at the Office of Naval Research (soon to be renamed Defense Advanced Research Projects Agency [DARPA]). In front of a room of reporters, Rosenblatt showed his perceptron a deck of cards that had colored squares on either the left side or the right side. After “seeing” about 50 cards, the perceptron was able to consistently tell left-side cards from right-side cards. Importantly, it could also do this with cards it hadn’t previously seen. For many, this was the birth of neural networks (although Rosenblatt’s perceptron only had a single “neuron.”)

65 years later…

In the fall of 2023, the U.S. Air Force staged its inaugural AI-human dogfight under the Air Combat Evolution (ACE) program. This aerial duel featured a manned F-16 going head-to-head with its technologically advanced sibling, the X-62A VISTA – an experimental F-16 variant equipped with live AI agents, installed less than a year prior.

The development of this AI system, spearheaded by DARPA in December 2022, aimed to autonomously pilot a fighter jet while rigorously adhering to the Air Force’s stringent safety protocols. During the encounter, the two jets engaged in intense “high-aspect nose-to-nose engagements,” demonstrating their aerial prowess as they maneuvered within 2,000 feet of each other at breakneck speeds of 1,200 miles per hour. Notably, DARPA has remained tight-lipped about which aircraft emerged victorious from the engagement. To date, a total of 21 test flights have been conducted, with the program set to continue through 2024.

Just a quick reminder that in 1969, ARPA (the Advanced Research Projects Agency, renamed DARPA in 1972) launched ARPANET, which evolved into the internet.

Thinking about AI in the context of national defense, one has to ask: “Is Stargate (Microsoft and OpenAI’s proposed $100 billion AI supercomputer) being funded by DARPA, or is the Pentagon working on one of its own?” If the U.S. Government is working on its own $100 billion AI supercomputer, where are they getting the engineering talent to build it?

More By This Author:

ChatGPT: Now, Less Chatty… And That’s A Good Thing
Google To “Ground” Its AI Models In Truth
AI May Be Yum Brands’ Secret Sauce