Jürgen Schmidhuber
Deep-learning researcher at KAUST and IDSIA; co-invented the LSTM, and keeps a long public record of who published which idea first.
Jürgen Schmidhuber did not write this page. ppll compiled it from their own words, each linked to the source. Is this you? · Report a mistake
Latest x.com
What they believe
-
Recurrent backpropagation learns slowly over long intervals mainly because error backflow decays.
Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow.
doi.orgSaid 1 Nov 1997This on korrents
The bold line is korrents’ wording; the quote is theirs. These one: what they wrote before what they said aloud, one per source, newest first.
Beliefs
Korrents What they believe 2 beliefs — each backed by an exact quote.
Each is a — compiled by korrents.com, not by them: the one-line wordings are korrents', the quotes are theirs.
Recent
Recurrent backpropagation learns slowly over long intervals mainly because error backflow decays.
Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow.
Long Short-Term Memory Said 1 Nov 1997
LSTM can solve complex artificial long-time-lag tasks that prior recurrent network algorithms never solved.
LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.
Long Short-Term Memory Said 1 Nov 1997
Beliefs others hold too
Recurrent backpropagation learns slowly over long intervals mainly because error backflow decays. 2 hold this
Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow.
Long Short-Term Memory Said 1 Nov 1997
LSTM can solve complex artificial long-time-lag tasks that prior recurrent network algorithms never solved. 2 hold this
LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.
Long Short-Term Memory Said 1 Nov 1997
What is a korrent?
A korrent is a belief a person has stated in their own words: one sentence stating the claim, backed by a quote and a source, kept at korrents.com.
Under a name here, the quoted block is what they actually said. The korrent beneath it is the claim those words support, in korrents' wording — tap it to see the record, its source, and who else holds it.
Nobody here wrote their own korrents. They are compiled from public statements, and a person can change their mind, which is recorded too.
Recommends rcmnd
rcmnd What they recommend 1 thing — each backed by an exact quote.
Compiled by rcmnd.app, not by them: the quotes are theirs, and each links to the post it was copied from.
Recent
Disliked A New Kind of Science
In fact Wolfram's book adds nothing substantial to Zuse's ideas, besides the pretty illustrations. Nor does it offer any predictions that go beyond Zuse's.
Where they publish
Site people.idsia.ch/~juergen His page: papers, the AI history posts and the priority disputes.
Link verified 20 Sept 2026.
x.com @SchmidhuberAI Their X account.
Recent
- This is an old hat. That's exactly what the neural fast weight programmers did in the 1990s (including the 1991 unnormalised linear Transformer). Overview with links to the original papers: 3 Nov 2026 people.idsia.ch
- Our embodied AI draws on memory. It recalls lived experiences, which contain depths of meaning that enable the AI to model the world, to plan, to minimize expected pain & maximize expected reward, and connect more deeply with the goodness of these 10 Oct 2026 people.idsia.ch
- My 1991 paper [1] introduced 3 techniques that have become essential for modern large language models: ★ Pre-training for deep neural nets - the P in ChatGPT ★ Positional encoding (Sec. 5.1) ★ Distilling the knowledge in a neural net (Sec. 3.2.2 & 4) [1] Neural sequence chunkers. TR FKI-148-91, TUM, April 1991 https://t.co/rg2nQnw5Ck Journal publication: Learning complex, extended sequences using the principle of history compression. Neural Computation, 4(2):234-242, 1992. See also the overview 6 Oct 2026 people.idsia.ch
Link verified 23 Sept 2026. Recent items update automatically from the channel.
About the English under a post
Some people here publish in a language other than English. Where they do, this site shows a machine translation beneath the post, in this typeface — the site's own, not theirs.
The post itself is never changed, moved or hidden: what is set in the serif above is exactly what the person published, and it is what to quote them on. A translation can be wrong in ways that matter, especially about tone.
Only the post's own words are translated. A quoted post, a linked article and a belief on korrents.com are left in their original language.