Recent research has revealed that top-tier experts actively working in the field of AI have significantly underestimated the speed of technological progress. Compared to the timelines anticipated by many experts, the enhancement of AI capabilities and its deployment in society are progressing at a much faster pace.
This study was conducted by comparing predictions gathered from leading AI researchers and specialists with the actual dates when technological milestones were achieved. The results confirmed that many crucial tasks and performance improvements are being realized years ahead of expert projections.
This forecasting gap stems from the non-linear improvements in the learning capacity of AI models and the unexpectedly rapid deployment of computational resources. The acceleration of performance gains based on scaling laws is progressing at a rate that surpasses even the expectations of experts.
These findings urge a revision of predictive models for future AI development. As technological advancement accelerates, discussions surrounding social implementation and safety will increasingly demand more dynamic and flexible approaches.