A competitive landscape is emerging in AI chip development among major cloud providers. Google recently launched Trillium, a custom chip for AI, while Microsoft’s Maia chip is also anticipated soon. Amazon Web Services (AWS), a leader in cloud technology, is also expanding its AI chip offerings, with notable products like Trainium, Inferentia, and Graviton. To spotlight Trainium, AWS has introduced a grant initiative called "Build on Trainium," offering $110 million in credits for AI research, with up to $11 million per institution and grants up to $500,000 for individual researchers. Additionally, AWS plans to establish a research cluster of 40,000 Trainium chips for scholars to access through managed reservations.
Gadi Hutt from AWS’ Annapurna Labs explained that this program aims to provide researchers with the necessary hardware to advance AI studies, especially for those constrained by limited resources. Hutt emphasized the growing gap between the tech industry and academia, citing that many university researchers lack the infrastructure to develop large-scale AI models, unlike corporations like Meta, which use over 100,000 AI chips for their projects.
Despite AWS’s efforts, some are skeptical about the intentions behind Build on Trainium. Os Keyes, a PhD student at the University of Washington, expressed concerns about corporate influence on academic research. AWS retains control over project selection, with a panel of AI experts reviewing submissions. There’s evidence that corporate-funded AI research often leans towards commercial applications rather than ethical studies, raising concerns about the diversity and breadth of research supported.
Though AWS claims researchers won't be tied to its ecosystem, grant recipients are required to publish their findings and make them open source. The issue, however, underscores a larger funding imbalance: in 2021, U.S. academic funding for AI was $1.5 billion, while private industry spent over $340 billion on AI globally. With nearly 70% of AI PhD holders transitioning to private industry, often attracted by resources and competitive salaries, most groundbreaking AI models today originate from industry.
Policymakers are working to reduce the academic-industry divide. For instance, the National Science Foundation allocated $140 million to establish National AI Research Institutes, and a $2.6 billion initiative is underway to give AI researchers access to computational tools. However, these efforts remain modest compared to the scale of corporate funding, suggesting that the dominance of industry-backed AI research is unlikely to change soon.
Post a Comment