Anthropic co-founder Jack Clark has predicted that AI research and development will become fully automated by 2028. Clark's assessment is based on the rapid advancements AI has made in tasks related to AI research, as demonstrated by benchmarks like CORE-Bench and PostTrainBench. These benchmarks evaluate AI's ability to implement research papers and autonomously fine-tune models, respectively. Clark believes that AI is nearing the capability of end-to-end automated development, which could lead to AI systems autonomously building their successors and initiating a cycle of self-improvement. Clark's prediction has sparked debate, with some viewing it as a step toward artificial superintelligence and the singularity, while others, like University of Washington professor Pedro Domingos, question the evidence for increasing returns from AI systems. Clark acknowledges the profound implications of fully automated AI research and expresses uncertainty about society's readiness for such changes. He estimates a 60% probability of achieving AI-driven research without human involvement by 2028, emphasizing the need for careful consideration of the potential consequences.