Provenance record

Artificial intelligence now beats some of the best human forecasters

The Economist: Science and Technology (tier 1, news) 2026-09-16T18:47:08.000Z Original ↗

Source note: Paywalled beyond the standfirst.


Discourse valence
-23
Adverse
confidence 52% · 5 items · range -100 to +15
Adverse readingFavourable reading
Consensus of 5 models from different labs. Spread 115 points, agreement low.
AGI and timelinesEvaluation and measurementScaling and architecture
Excerpt as ingested

Crystal balls give way to LLMs

Every model that read this

ModelProviderStageScoreConf.LatencyPromptWhen
Llama 3.3 70BMetaanalysis +10 60%2489ms v1.0.0 / m1.0.1 2026-09-16 21:17
GPT-4.1 miniOpenAIconsensus +15 60%3803ms v1.0.0 / m1.0.1 2026-09-16 21:17
Claude Sonnet 5Anthropicconsensus +15 30%4529ms v1.0.0 / m1.0.1 2026-09-16 21:17
Gemini 2.5 FlashGoogleconsensus -100 20%1404ms v1.0.0 / m1.0.1 2026-09-16 21:17
Mistral Small 3.1 24BMistral AIconsensus -55 90%5104ms v1.0.0 / m1.0.1 2026-09-16 21:17
Llama 3.3 70B · reading

AI outperforms humans in forecasting.

evidence: speculative horizon: n/a
GPT-4.1 mini · reading

The claim that AI now outperforms some top human forecasters implies improvements in prediction accuracy which could benefit many fields. However, it raises concerns about overreliance on AI forecasts and possible loss of human judgment. The societal effect is likely mixed, improving outcomes in forecasting tasks but with some risk.

evidence: reported horizon: near capability 40 societal 20 existential 0 economic 30
Claude Sonnet 5 · reading

The headline claims AI now beats some top human forecasters, a narrow but notable capability milestone. Without the full article, the societal implication is limited mostly to improved decision-support tools, a mild positive with unclear downstream effects.

evidence: reported horizon: n/a
Gemini 2.5 Flash · reading

The article's title suggests AI has achieved a significant milestone in forecasting. However, the available text is only a title and a very short standfirst, which provides no details or evidence to substantiate this claim or assess its societal implications. Thus, the item's actual impact remains entirely unclear.

evidence: reported horizon: n/a
Mistral Small 3.1 24B · reading

AI outperforming human forecasters suggests a shift in decision-making power. This could lead to over-reliance on AI, reducing human judgment.

evidence: primary horizon: n/a

Evidence extracted

The chain
SOURCE     The Economist: Science and Technology (tier 1)
   ↓
DOCUMENT   5f97b987-7899-4587-a198-ab9ae6817adc
           https://economist.com/science-and-technology/2026/09/16/artificial-intelligence-now-beats-some-of-the-best-human-forecasters
   ↓
EVIDENCE   1 extracted excerpt
   ↓
MODEL RUN  5 runs, methodology 1.0.1
   ↓
SCORE      -23  (Adverse)
   ↓
CONFIDENCE 52%