AI Scientists Doubt That Current Models Will Lead to AGI - But Do We Even Need AGI for Disruption?

March 15th 2025

A new survey of AI researchers reveals widespread skepticism about whether today’s AI models will ever achieve artificial general intelligence (AGI), despite tech giants investing billions toward this goal. The study, conducted by the Association for the Advancement of Artificial Intelligence (AAAI), found that 76% of AI researchers believe scaling up existing AI models is unlikely to result in AGI.

This signals a shift from the once-popular belief that “scaling is all you need,” which dominated AI discussions during the rise of transformer-based generative models like GPT. The diminishing returns of larger datasets and computing power are raising questions about whether an entirely new approach is necessary to reach human-level AI.

However, while AGI may still be a distant goal, agentic AI and increasingly capable specialized models are already reshaping industries, automating tasks, and influencing labor markets. Advanced AI does not need to reach human-level intelligence to cause profound economic and social changes - these shifts are already happening.

Despite doubts about AGI, companies like OpenAI, Google DeepMind, and Microsoft continue massive investments in AI infrastructure, with some redefining AGI in economic terms, such as OpenAI’s internal definition of AGI being an AI system that generates $100 billion in profit.

This raises an important question: If the goal of AGI keeps shifting and current AI models are already disrupting industries, do we really need AGI for revolutionary changes, or is the transformation already underway?

Read the full AAAI report here

Source: NewScientist.com