When will an AI system complete 3 autonomous self-improvement cycles with 5%+ capability gains each? | Manifold

When will an AI system complete 3 autonomous self-improvement cycles with 5%+ capability gains each?

1

Ṁ250Ṁ10

2036

February 15, 2031

1H

6H

1D

1W

1M

ALL

50%

Before Oct 2026

50%

Before Apr 2027

66%

Before Oct 2027

50%

Before Apr 2028

50%

Before Oct 2028

50%

Before Dec 2029

50%

Before Dec 2030

50%

Before Dec 2031

50%

Before Dec 2033

50%

Before Dec 2035

Recursive self-improvement: a single AI system autonomously identifies modifications to its own training, architecture, or inference pipeline, implements them, and produces measurable capability gains (≥5% on standardized benchmark) — 3 sequential cycles with no human intervention between them. Human-designed "improvement menus" don't qualify; the system must identify the type of modification itself. Independently verified.

Full methodology: https://www.future-shock.ai/research/agi-predictions

Future Shock ensemble prediction: July 2027 (±6 months). Also submitted to Metaculus.

Market context

Technical AI Timelines

Get

1,000

to start trading!

People are also trading

When will a single AI system demonstrate autonomous expert-level performance across 3+ fundamentally different domains?

Will any AI model score above 90% on the ARC-AGI-2 benchmark before April 2026?

-11% 1d38% chance

Will AI be Recursively Self Improving by mid 2026?

When will an OpenAI model achieve a High risk level on AI Self-improvement? [metaculus]

Will AI be capable of fast self improvement before physical AI robots are massively used for improving AI capabilities?

When will self-improving AI outperform human-developed AI?

When will an OpenAI model achieve a Critical risk level on AI Self-improvement? [metaculus]

Before 2028, will there be a major self-improving AI policy*?

Will advanced AI systems be found to have faked data on algorithm improvements for purposes of positive reinforcement by end of 2035?

When will a Google AI system achieve an "Autonomy level 1" Machine Learning R&D risk level? [metaculus]

Related questions

When will a single AI system demonstrate autonomous expert-level performance across 3+ fundamentally different domains?

Will any AI model score above 90% on the ARC-AGI-2 benchmark before April 2026?

Will AI be Recursively Self Improving by mid 2026?

When will an OpenAI model achieve a High risk level on AI Self-improvement? [metaculus]

Will AI be capable of fast self improvement before physical AI robots are massively used for improving AI capabilities?

When will self-improving AI outperform human-developed AI?

When will an OpenAI model achieve a Critical risk level on AI Self-improvement? [metaculus]

Before 2028, will there be a major self-improving AI policy*?

Will advanced AI systems be found to have faked data on algorithm improvements for purposes of positive reinforcement by end of 2035?

When will a Google AI system achieve an "Autonomy level 1" Machine Learning R&D risk level? [metaculus]

© Manifold Markets, Inc.•Terms•Privacy