This post reflects on the misguided optimism surrounding AI's potential in scientific research. The author, a Princeton PhD graduate in physics, recounts personal experiences where the hype around AI in scientific contexts led to disillusionment. A key point raised is the pervasive issue of negative results being overlooked in scientific publishing, a situation not unique to AI but symptomatic of broader flaws in scientific evaluation methodologies. The post also warns of serious budget cuts proposed for the NSF, driven by claims that AI could replace traditional research methods, suggesting these claims may be overstated. The author argues that such beliefs ignore the complex realities of scientific inquiry, and that true progress requires continued investment in human-led research.