The great AI hype correction of 2025 (www.technologyreview.com)

In 2025, the AI/ML community is experiencing a significant shift in expectations, as industry leaders confront the limitations of generative AI and the overpromised advancements that were hyped post-ChatGPT's launch. The enthusiasm for AI technologies, which led to explosive growth and substantial investment, is now overshadowed by reports indicating stalled uptake among businesses. Many AI projects remain in pilot phases, with studies revealing that 95% of companies see little to no value from AI implementations, despite the technology's potential to enhance workflows. This has prompted a reevaluation of the narrative that AI could revolutionize the workforce and drive massive economic change. Key technical challenges are surfacing as the excitement surrounding models like GPT-5 has waned, exemplifying a moment of reckoning for the industry. Experts like Ilya Sutskever are emphasizing the inherent limitations of large language models (LLMs), which may mimic human-like behavior but lack true understanding and generalization capabilities across tasks. As the community grapples with what constitutes realistic advancements in AI, researchers are redirecting focus to the practical applications and integration strategies needed to realize AI's promises, while cautiously considering whether the current landscape reflects a bubble akin to the dot-com era. The ongoing conversation underscores the necessity of tempering expectations while recognizing the substantial innovations still unfolding within the field.