This is Juergen Schmidhuber’s keynote “Modern Artificial Intelligence 1980s-2021 and Beyond,” posted on his own official YouTube channel. The talk first premiered on 3 December 2020 at the AIJ conference in Moscow and was subsequently presented at NVIDIA’s GTC-21 and several other venues, with the description noting it reached millions of viewers across platforms.
Schmidhuber uses the talk to lay out his account of the history of modern AI, tracing a line from his lab’s work in the late 1980s and early 1990s through Long Short-Term Memory, developed with Sepp Hochreiter, and on to the deep-learning systems that power today’s applications. He highlights early ideas his group pursued - including artificial curiosity, self-supervised learning, and what he frames as precursors to later methods - and he is candid about crediting contributions he believes are often overlooked.
This is a practitioner-level talk that doubles as a firsthand, occasionally contrarian, history lesson. Its value as a primary source is precisely that it is one of the field’s pioneers giving his own version of where the building blocks of modern AI came from.