Mark I Perceptron (1958)

Mark I Perceptron은 1958년 Frank Rosenblatt가 개발한 최초의 인공지능 기계로, 3계층(감각, 연관, 반응 유닛) 구조를 갖춘 감독 학습 이미지 분류 시스템입니다. 단일 층 500~1000 뉴런으로 구성된 이 시스템은 종이 위의 기호를 위치, 회전 변형을 포함해 높은 정확도로 구분했으며, 10,000개 이상의 이미지로 학습했습니다. 당시 3초~15초의 처리 시간으로 99.8%까지 정확도를 달성했고, 이는 현대 AI 기술의 기초가 된 퍼셉트론 아키텍처의 초기 실험적 성공 사례입니다.

https://en.wikipedia.org/wiki/Mark_I_Perceptron

#perceptron #supervisedlearning #imageclassification #historicalai #frankrosenblatt

Mark I Perceptron - Wikipedia

A computer science student trained an AI exclusively on Victorian texts from 1800-1875. When prompted about 1834, the model accurately referenced real historical protests and Lord Palmerston—events the developer didn't know were factual. The "quarantine strategy" suggests what you exclude from training data may matter as much as what you include for authentic outputs.

#AITraining #HistoricalAI #MachineLearning

https://www.implicator.ai/the-quarantine-strategy-to-make-an-ai-remember-the-past-you-have-to-protect-it-from-the-future/

The Quarantine Strategy: To make an AI remember the past, you have to protect it from the future

A college student trained an AI on 7,000 Victorian texts. When he prompted it about 1834, it described protests he didn't know were real.

Implicator.ai
College student’s “time travel” AI experiment accidentally outputs real 1834 history

Hobbyist training AI on Victorian texts gets an unexpected history lesson from his own creation.

Ars Technica