Industry experts accelerate predictions for the arrival of human-level AI

Leading AI experts and researchers predict human-level artificial intelligence could emerge within decades, with estimates ranging from as early as 2027 to later this century.

Current consensus and timeline estimates: Major surveys and aggregate forecasts from the AI research community point to potentially transformative AI developments in the coming decades.

• A 2022 survey of 738 machine learning researchers projected a 50% likelihood of human-level AI by 2059
• Metaculus, a prominent forecasting platform, predicts “the first general AI system” by 2031 and “weakly general AI” by 2027
• Samotsvety forecasters estimate a 50% probability of AGI by 2041, with a 9-year standard deviation

Expert perspectives: Leading figures in artificial intelligence research have offered varying predictions about the timeline for achieving human-level AI capabilities.

• Turing Award winner Yoshua Bengio and renowned AI scientist Geoffrey Hinton both estimate a 5-20 year timeframe, though with different confidence levels
• DeepMind co-founder Shane Legg projects an 80% probability by 2037
• Meta’s Chief AI Scientist Yann LeCun suggests it could take “several years if not a decade,” while acknowledging potential longer timelines
• Paul Christiano, head of the US AI Safety Institute, estimates a 30% chance of transformative AI by 2033

Modeling and research analysis: Technical modeling approaches provide additional context for timeline predictions.

• Ajeya Cotra’s “biological anchors” analysis for Open Philanthropy updated its timeline from 2050 to 2040 based on recent AI developments
• Matthew Barnett’s “direct approach” model, which extrapolates from training loss data, estimates transformative AI around 2053
• A 2023 survey of Conjecture employees unanimously expected AGI before 2035

Analytical considerations: Timeline predictions require careful interpretation given their inherent limitations and assumptions.

• Forecasts vary in their definitions of human-level AI versus transformative AI
• Selection bias may influence both survey participation and which predictions receive attention
• Timeline estimates have generally shortened in recent years as AI capabilities have advanced more rapidly than expected

Future implications: While the range of predictions spans several decades, there appears to be growing consensus among experts that transformative AI capabilities could emerge within many people’s lifetimes, potentially reshaping technology, society, and human civilization in profound ways.