Iris Coleman
Dec 17, 2025 06:09
Dan Fu from collectively.ai argues that synthetic basic intelligence (AGI) is achievable by optimizing software-hardware co-design, enhancing present chip utilization, and overcoming perceived {hardware} limitations.
The controversy surrounding the potential for attaining synthetic basic intelligence (AGI) is intensifying, with Dan Fu, Vice President of Kernels at collectively.ai, offering an optimistic outlook. In accordance with collectively.ai, Fu challenges the notion that developments in AI are being stymied by {hardware} limitations. As a substitute, he posits that present chips are considerably underutilized and {that a} strategic method to software-hardware co-design might unlock substantial efficiency enhancements.
Present Limitations and Future Potential
Because the AI panorama evolves, issues about reaching the bounds of digital computation have gotten extra prevalent. Some specialists recommend that {hardware} constraints, notably in GPUs, may impede progress in the direction of growing typically helpful AI. In distinction, Fu presents a extra hopeful perspective in his publication, “Sure, AGI Can Occur – A Computational Perspective,” which argues that the ceiling has not but been reached for AI capabilities.
Underutilization of Present {Hardware}
Fu highlights that state-of-the-art AI coaching runs, comparable to DeepSeek-V3 or Llama-4, usually obtain solely about 20% Imply FLOP Utilization (MFU), with inference utilization generally within the single digits. These figures recommend a big alternative to boost effectivity by means of higher integration of software program and {hardware}, in addition to improvements like FP4 coaching.
Developments in Computational Fashions
Present AI fashions are based mostly on older {hardware}, and the potential of newer computational assets has not been totally realized. Fu emphasizes that large clusters of the most recent era GPUs, numbering over 100,000, have but to be totally built-in into AI growth processes, indicating a promising horizon for future developments.
Current-Day Utility and Future Implications
Regardless of the perceived limitations, present AI fashions are already revolutionizing complicated workflows, comparable to writing high-performance GPU kernels with human help. This transformation underscores the speedy utility of AI applied sciences and hints on the huge potential for future purposes.
For these within the intersection of methods engineering, {hardware} effectivity, and AI scaling, Fu’s evaluation supplies helpful insights. The complete evaluation might be accessed on the collectively.ai web site.
Picture supply: Shutterstock
