Schmidhuber accuses AI companies of training models on plagiarized, misattributed work
SchmidhuberAI · x · 2026-09-10
Jürgen Schmidhuber posted that certain companies train their AIs to plagiarize work by unnamed scholars, extending an AI history 'permeated by plagiarism and misattributions':
- 1900s: the method of least squares (Gauss & Legendre, 1795-1805) was renamed 'neural network' without citation.
- 1950s: cybernetics was renamed 'AI' by people who did not want to credit earlier pioneers like Wiener.
- 2000s: 1960s AI techniques (Ivakhnenko & Lapa 1965, Amari 1967) were renamed 'deep learning' without correct attribution.
He says he could cite many more cases, that some plagiarists even won awards, and frames the post as being in the field's and scientific honesty's best interest.
Related event: Schmidhuber blasts AI history as rife with plagiarism(2 posts)→
More from AGI Musings
- Apple A20 Pro's 50% memory bandwidth jump is a big deal for on-device multi-agent AI — Scobleizer · 2026-09-10
- Anthropic Researcher Quits: 'Racing Straight to Self-Improving Superintelligence, Gambling With Our Lives' — Confident_Salt_8108 · 2026-09-10
- 'Culture Is Becoming a Dark Forest': AI Race May Destroy the Knowledge Commons It Built On — erikphoel · 2026-09-10
- Google's Denny Zhou: The Biggest Research Divide Is Access to Top Models and Compute — denny_zhou · 2026-09-10
- Will Automating AI R&D Trigger a Software Intelligence Explosion? Paper Analyzes — nabeelqu · 2026-09-10
- Can Language Models Reproduce Creative Breakthroughs from Historical Data? Scientists Investigate — CatAstro_Piyush · 2026-09-10