A recent study led by Peter Kirgis and Sayash Kapoor at Princeton University reveals that current AI agents are proficient at solving engineering problems within AI research but fall short in producing original work that meets top conference standards. According to MIT Technology Review, these AI systems lack the necessary judgment and creativity for open-ended AI research.
The findings highlight a significant limitation in AI development, emphasizing that while AI can assist in technical tasks, it cannot yet replace the human element required for innovative and creative scientific contributions. This research underscores the gap between engineering capabilities and creative insight in AI applications.
For Japanese markets, where innovation in AI-driven technologies is rapidly advancing, understanding these limitations is critical for realistic expectations around AI’s role in research and development across FX, crypto, and equities sectors.