A recent survey conducted by the Association for the Advancement of Artificial Intelligence (AAAI) found that most researchers consider the current approach to developing artificial intelligence to be a dead end, Futurism reports.
Of the 475 respondents surveyed, 76% said that scaling up existing technologies is unlikely to lead to the creation of artificial general intelligence (AGI) capable of matching or surpassing human intelligence.
Experts like Stuart Russell of the University of California, Berkeley, point out that investing in scaling without a deeper understanding of the processes is ineffective. We are already seeing a decline in the efficiency of new models like GPT compared to previous versions.
Despite this, companies continue to invest billions of dollars in developing infrastructure for generative models. For example, Microsoft plans to spend $80 billion on AI infrastructure by 2025.
However, cheaper and more efficient approaches are currently being explored. For example, OpenAI is exploring alternative methods such as “compute-while-testing,” which allows models to make more informed decisions. This adds performance gains that would otherwise require massive scale to replicate.
While these methods show some progress, experts warn that they are not a one-size-fits-all solution. The industry needs new approaches to avoid stagnation.