Forecasters pull forward their timeline for superhuman AI systems to 2027

Forecasting group AI Futures Project updated its timeline, predicting that artificial intelligence systems capable of replacing human software researchers could arrive as early as 2027. The shift follows research evaluations suggesting internal AI tools are accelerating software development inside major laboratories far faster than expected. According to the group's analysis, AI tools are responsible for nearly doubling the speed of recent software research at frontier labs like Anthropic.

For regular people, this shift suggests that fully automated office work and self-improving artificial intelligence might be just around the corner rather than decades away. The researchers estimated a significant chance that systems capable of outperforming top human experts across all white collar professions could emerge before 2028. If software systems gain the ability to conduct their own engineering research and development, the pace of technological change could compound at a rate that traditional institutions and regulators are unequipped to handle.

What remains uncertain is whether these internal corporate productivity metrics translate to real-world capability or if they simply reflect narrow gains in routine coding tasks. The authors themselves acknowledge that their modeling relies on early corporate self-reporting, which might prove overly optimistic. It also remains to be seen whether tech companies would ever voluntarily pause development to establish safety protocols, or if intense geopolitical competition will force them to deploy highly autonomous systems before they can guarantee safety.