Majority of AI Researchers Say Tech Industry Is Pouring Billions Into a Dead End

A recent survey of 475 AI researchers, conducted by the Association for the Advancement of Artificial Intelligence, has cast doubt on the tech industry’s current approach to achieving artificial general intelligence (AGI). The overwhelming majority—76%—believe that merely scaling up existing AI models is “unlikely” or “very unlikely” to result in AGI, which would match or surpass human cognition.

For years, major tech companies have poured billions into generative AI, relying on ever-larger models and vast data centers to fuel progress. In 2024 alone, venture capital investments in generative AI exceeded $56 billion. Microsoft, for example, plans to spend $80 billion on AI infrastructure in 2025. This has driven massive energy consumption, with companies even securing nuclear power deals to sustain their operations.

Yet, AI experts suggest that scaling is reaching its limits. Stuart Russell, a UC Berkeley computer scientist involved in the report, argues that the industry’s obsession with brute-force scaling has resulted in diminishing returns. Signs of this plateau became evident in late 2023, when OpenAI’s newest GPT model showed little improvement over previous iterations. Even Google’s CEO Sundar Pichai acknowledged that “easy AI gains” were over, though he still advocated for continued scaling.

Meanwhile, alternative approaches are gaining traction. OpenAI has experimented with test-time compute, allowing models to process information more efficiently instead of relying solely on scaling. Chinese startup DeepSeek has introduced a “mixture of experts” model, which uses specialized neural networks rather than a single, all-encompassing model.

Despite these innovations, industry giants appear committed to the scaling model, while smaller startups are leading the charge in finding more efficient ways to advance AI. Whether brute-force scaling can still yield breakthroughs remains an open question—but many researchers believe the future lies elsewhere.