Paper argues intelligence is learning to turn surprise into understanding
burny_tech · x · 2026-07-23
A paper argues intelligence is the ability to convert surprise into understanding
The preprint "Intelligence from Learnable Novelty" proposes that intelligence is not surprise itself, but the capacity to separate learnable from unlearnable novelty and optimize only the learnable part.
- The authors argue that two common objectives both fail for opposite reasons: novelty search gets trapped by noise, while surprise minimization gets trapped by stasis.
- They introduce a cheap differentiable estimator built on reservoir computing.
- With it, the method:
- recovers decades of complexity-classification results, including Rule 110 among elementary cellular automata,
- grows solitons and traveling/colliding structures,
- clusters MNIST without labels,
- improves reinforcement-learning exploration when used as an intrinsic reward.
- The paper frames complexity generation, abstraction, and exploration as arising from ascent on a single differentiable quantity.
Related event: New Paper Unifies Intelligence Concepts via Learnable Novelty(7 posts)→
More from Research
- RFD4-Proteina runs on an RTX 3060: protein design from $0.001 per design — AllThingsApx · 2026-10-27
- University of Freiburg's Generative Intelligence Lab recruiting funded PhDs via ELLIS, apply by Oct 31 — yuyinzhou_cs · 2026-10-27
- Nature Physics: reprogrammable refractive index enables arbitrary multimode wave control for ML — MarioKrenn6240 · 2026-10-27
- Wakaru open-sources ontologize: labeling datasets from model activations — amplifiedamp · 2026-10-27
- EMNLP Oral paper traces multilingual representations in LLMs with cross-layer transcoders — ZhijingJin · 2026-10-27
- GPT-6 Pro reruns crime research: GTA V's 2013 launch tied to violence dips in four regions — emollick · 2026-10-27