A groundbreaking survey of 475 artificial intelligence researchers has revealed widespread skepticism about the current approach to achieving artificial general intelligence (AGI). The findings suggest that the tech industry's massive investments in computing power and data centers may be heading down an unproductive path.
According to the survey conducted by the Association for the Advancement of Artificial Intelligence (AAAI), 76% of researchers believe that simply scaling up existing AI models with more computing power is "unlikely" or "very unlikely" to lead to AGI - systems that match or exceed human-level cognition.
The tech industry has bet heavily on the "bigger is better" strategy. Last year alone saw $56 billion in venture capital funding flow into generative AI, while the semiconductor industry reached $626 billion in 2024. Tech giants like Microsoft, Google, and Amazon have even begun securing nuclear power deals to fuel their expanding data centers.
However, the returns on these massive investments appear to be diminishing. Recent AI model releases have shown only modest improvements despite exponential increases in computing resources. UC Berkeley computer scientist Stuart Russell noted that the heavy focus on scaling without deeper understanding of AI systems has been "misplaced."
The survey also revealed shifting priorities in the AI research community. While 77% of researchers emphasize developing AI systems with acceptable risk-benefit profiles, only 23% focus directly on pursuing AGI. A strong majority (82%) believe that if AGI is achieved, it should be publicly owned rather than controlled by private companies.
Some companies are exploring alternative approaches. OpenAI has experimented with "test-time compute," allowing AI models more processing time before generating responses. While this has shown promise, Princeton's Arvind Narayanan cautions it's unlikely to be a complete solution.
Despite mounting evidence against the scaling strategy, some industry leaders remain committed to it. Google CEO Sundar Pichai continues to advocate for scaling up AI systems, though he acknowledges that easy gains may be over.
The survey results suggest a growing divide between AI researchers' understanding of the field's challenges and the tech industry's investment strategies. As billions continue flowing into computing infrastructure, the question remains whether this approach will yield the breakthroughs companies are banking on.