Key Findings
A research paper from the University of Cambridge’s AI Science and Policy Program (CASP) thoroughly discusses the potential for an “intelligence explosion,” where AI rapidly develops next-generation AI. The paper reveals a dramatic acceleration in AI’s autonomous R&D capabilities. Notably, at Anthropic, the approval rate for AI-generated code exceeded 80% by May. Furthermore, the proportion of autonomous R&D tasks performed by AI surged from 1% to 26% within a short period from March to August. Researchers project that AI R&D projects that currently take humans months could be fully automated by AI by mid-2028, potentially accelerating AI development by up to 10 times within 1.5 years.
Technical Details
The “intelligence explosion” scenario posits that AI will exponentially enhance its own intelligence through self-improvement, ultimately far surpassing human intellect. The case of Anthropic suggests that early stages of this scenario are already underway. An AI generating code that achieves over 80% approval after human review implies high-quality contributions from AI in the initial stages of software development. Moreover, AI’s 26% involvement in autonomous R&D tasks indicates that AI is beginning to assist, or even replace, humans in more complex cognitive processes such as hypothesis formation, experimental design, data analysis, and paper writing.
Background and Context
The development of large language models (LLMs) has enabled AI to demonstrate high capabilities across various tasks, including text generation, translation, summarization, and programming assistance. AI’s role as a programming assistant, in particular, has been remarkable, with AI-generated code being adopted in many development environments. This situation transforms the possibility of AI not just remaining a tool but becoming an agent capable of improving itself and developing new AIs—a concept previously considered science fiction—into a reality. This research highlights that AI’s evolution is progressing at a pace far beyond predictions, reinforcing the immense impact it will have on society and industry as a whole.
Strategic Significance and Outlook
The acceleration of AI R&D by AI itself presents both positive aspects and significant challenges. Positively, solutions to complex problems facing humanity—such as drug discovery, climate change mitigation, and energy efficiency—could be found more rapidly than ever before. However, simultaneously, challenges related to AI control, safety, and ethical governance will become even more profound. If AI begins to lead autonomous R&D projects by mid-2028, it might become difficult for humans to keep pace with AI’s evolution. This prediction strongly advocates for international coordination and swift action regarding AI governance and regulation. The hyper-accelerated evolution of AI presents us with both new opportunities and unprecedented responsibilities.
Source: https://www.mk.co.kr/en/it/12163345
Get our weekly technology intelligence — free
Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.
Subscribe Free — Weekly Tech Intelligence
By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.
- Your email and selected fields are used only to deliver the newsletter.
- We never share your information with third parties.
- You can unsubscribe anytime via the link in each email.
See our Privacy Policy for details.
Takes about a minute · Unsubscribe anytime
