AI Futures Model Predicts 3-Year Delay for Full Coding Automation Amid R&D Rethink

The latest 'AI Futures Model: Dec 2025 Update' introduces significant revisions to previous forecasts regarding AI timelines and takeoff speeds, particularly for key milestones such as Automated Coder (AC) and superintelligence (ASI). This updated model now predicts approximately three years longer timelines to full coding automation compared to its predecessor, the 'AI 2027' model published in April 2025. The primary driver for this adjustment is a less optimistic view on the speedups of AI research and development (R&D) occurring before AI achieves full automation capabilities. This shift underscores the dynamic and uncertain nature of AI development, where even incremental improvements in modeling can lead to substantial changes in long-term projections.

The developers of the model emphasize the inherent difficulty in predicting the future of AI, acknowledging that while their model incorporates critical dynamics, it cannot account for all factors. Crucially, many parameter values within the model rely on intuitive guesses rather than solely empirical data, a transparency that invites users to interact with the model via an interactive website to adjust parameters based on their own judgments. This approach highlights the lack of universal consensus and the speculative elements inherent in AGI forecasting. The improvement in modeling AI R&D automation, rather than new empirical evidence, was the main catalyst for the 2-4 year shift in the median for full coding automation. This suggests that refinement in theoretical understanding and modeling techniques can sometimes outweigh new observational data in shaping our expectations.

The uncertainty surrounding AI timelines is further evidenced by the wide divergence in expert opinions. A 2023 survey of AI academics, for instance, yielded a median AGI prediction of either 2047 or 2116, depending on the definition used for AGI. In contrast, aggregations from platforms like Metaculus and Manifold markets suggest a 50% probability of AGI by 2030. Such vast discrepancies emphasize the nascent state of AGI forecasting and the need for more robust, transparent methodologies. The 'AI Futures Model' aims to contribute to this by offering a structured, explicit framework for weighing various considerations, factors, and arguments, moving away from implicit, 'in-the-head' assessments.

Ultimately, the goal of such modeling is not to provide an infallible crystal ball but to enhance preparedness and understanding. By attempting to predict future trends and build models of underlying dynamics, stakeholders can gain a better sense of probabilities, allowing for more informed decision-making and better integration of future empirical data into existing forecasts. This updated model serves as a critical recalibration for anyone involved in AI strategy, investment, or policy, signaling that while progress is undeniable, the path to advanced AI capabilities may be more protracted than previously thought, offering potentially more time for societal adaptation and responsible development.