AI goes from great to disappointing
| Source: HN | Original article
AI once hailed for its breakthroughs now faces growing criticism and setbacks.
A new study released this week underscores both the promise and the limits of artificial‑intelligence tools in fundamental research. The paper, highlighted in an article by Nia Robles on 6 October 2026, shows that large language models can construct rigorous proofs in the highly abstract field of celestial holography. The authors demonstrate that, when confined to a well‑defined logical framework, the AI produces “impressive and not straightforward” arguments that open several fresh research directions.
The findings matter because they expose a growing reliance on AI for scientific reasoning while reminding the community that physics is more than formal deduction. The same work notes that AI struggles with the intuitive leaps, experimental judgment and creative synthesis that often drive breakthroughs. As AI‑generated content becomes ubiquitous—from draft papers to grant proposals—understanding where the technology excels and where it falters is essential for maintaining research integrity and for setting realistic expectations among funders and policymakers.
Looking ahead, the paper points to a need for hybrid workflows that combine the speed and consistency of AI with human insight. Researchers will likely explore methods to embed physical intuition into models, while journals may tighten verification procedures for AI‑assisted manuscripts. Observers should watch for follow‑up studies that test AI performance across other subfields of physics, as well as any emerging guidelines from funding bodies or scientific societies on the responsible use of generative AI in discovery.
Sources
Back to AIPULSEN