How Much Can Intelligence Be Compressed?
jacobkimmel · x · 2026-07-11
The author poses a fundamental question of the AI era: Exactly how much can "intelligence" be compressed?
He argues this is crucial because the historical trajectory of model capabilities has consistently demonstrated the potential for compression:
- Early improvements relied primarily on scaling up model size and data;
- Later, distillation emerged to compress large model capabilities into smaller ones;
- Yet, the limits of this compression remain unknown, lacking a mature quantitative framework.
The article explores two potential futures:
- If the compressibility of intelligence is limited, AI development might resemble an S-curve, becoming increasingly reliant on massive compute and energy inputs over time;
- If intelligence is more compressible than we think, we might still be on a very long exponential growth trajectory.
Ultimately, the author asks whether a near-consensus theory currently exists to answer how much intelligence can be compressed and how to quantify it.
Related event: LLMs Approach 10T Parameters Amid Intelligence Compression Trend(4 posts)→
More from AGI Musings
- Better languages emerged from struggle: AI shortcuts may cost the commons — jjvincent · 2026-09-11
- OpenRouter agents now out-consume humans as AI usage arrives in three waves — AccBalanced · 2026-09-11
- Novosad backs Hassabis' AI safety institution-building over kneecapping US labs — paulnovosad · 2026-09-11
- If AI teleports us to solutions, how do underlying fields develop? — jjvincent · 2026-09-11
- Economist argues safe AGI comes from engineers inside big labs, not regulation — paulnovosad · 2026-09-11
- Op-ed: the ">10% extinction" narrative is liability evasion — AI is just software, and the vendor is the defendant — gerardsans · 2026-09-11