Tech
Why a Computer Scientist Thinks Superintelligence Won’t Happen
Publication: IEEE Spectrum
Author: Cal Newport
Source published: 19/11/2026, 12:00:03 UTC
Briefing created: 19/11/2026, 18:38:48 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 19/11/2026, 15:13:25 UTC.
Briefings
A computer scientist argues that while AI companies like Anthropic and OpenAI are debating the risk of superintelligence, the technical feasibility of recursive self-improvement remains highly speculative due to data limitations. The author notes that AI has accelerated internally but cites diminishing returns on model scaling and the lack of easy-to-generate training data as key obstacles preventing uncontrollable intelligence explosions.
Key points
- Anthropic researcher Jacob Coxon resigned in 2025 warning that AI companies are 'racing straight to self-improving superintelligence and gambling with our lives'
- AI scaling from 117 million parameters in GPT-1 to over a trillion in GPT-4 has shown diminishing returns, limiting the data needed for recursive self-improvement
Context
The article discusses existential risks from recursive self-improvement (RSI) and artificial superintelligence (ASI), noting that while AI companies like Anthropic and OpenAI are actively debating this, the technical feasibility remains uncertain due to data constraints.
Limitations
Evidence is limited to a single computer science doctoral seminar at Georgetown University; specific dates for the seminar are not provided, and the author's personal conclusions are not verified by external sources.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 19/11/2026, 18:38:48 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeiawf2tijqfz5gzw65etboze7wb2zsnjmo4v22kkxjeof5shd5yocy