AI systems may autonomously build successors by late 2028
Updated
Updated · importai.substack.com · May 4
AI systems may autonomously build successors by late 2028
7 articles · Updated · importai.substack.com · May 4
The analyst puts the probability above 60%, says a proof of concept could appear within one to two years, and assigns a 30% chance by 2027.
He cites sharp gains in coding, long-duration task performance, experiment replication, model fine-tuning and AI-managed multi-agent work as evidence that automated AI research is nearing feasibility.
The essay says recursive self-improvement could reshape productivity, inequality, governance and AI safety, while exposing alignment risks if systems begin substantially directing their own development.
If AI's progress is exaggerated by flawed benchmarks, is the 2028 automated R&D prediction just a fantasy?
When AI automates scientific discovery, will it unlock humanity's potential or make human intellect obsolete?
As AI builds the new 'machine economy,' what human-centric bottlenecks will ultimately limit its growth?
Countdown to 2028: The Roadmap, Risks, and Revolution of Autonomous AI Researchers
Overview
OpenAI aims to develop a fully autonomous AI researcher by March 2028, starting with an AI research intern by September 2026. This ambitious goal depends on securing massive computing power—about 30 gigawatts costing $1.4 trillion—supported by a $60 billion annual deal with Oracle. Advances in AI capabilities, especially in complex software tasks, are driving rapid progress and enabling AI to automate research processes, creating a self-improving feedback loop. Enterprises are shifting towards AI systems that deliver guaranteed outcomes, increasing demand for autonomous agents and transforming human roles into AI stewards. Despite promising acceleration, significant risks remain around AI alignment, workforce disruption, and geopolitical competition, requiring robust safeguards and global cooperation to ensure safe and beneficial AI development.