Amazon is investing $110 million into universities to advance research AI using Trainium chips.
- The Top Enterprise AI Trends of 2024: AI Innovation
- Microsoft, Google, and Amazon Sign EU’s AI Safety Pact
- Amazon Launches ‘Project Amelia’ AI-Powered Personal Assistant
The programme, known as Build on Trainium, enables researchers to build new AI architectures, machine learning (ML) libraries, and performance upgrades for large-scale AWS Trainium UltraClusters, which are AI accelerators that work together to process complex computing tasks.
It provides research and training opportunities, which Amazon says is “sparking innovation for the frontier models of the future”.
Todd C. Mowry, Professor of Computer Science at CMU, explains what the impact is of Amazon’s investment: “AWS’s Build on Trainium initiative enables our faculty and students large-scale access to modern accelerators, like AWS Trainium, with an open programming model.
“It allows us to greatly expand our research on tensor program compilation, ML parallelisation, and language model serving and tuning.”
‘Build on Trainium’
AWS Trainium is the ML chip designed by AWS for deep learning and inference.
According to Amazon, the AI advancements made through this research will be open-sourced to ease continued research.
The Trainum research UltraCluster has as many as 40,000 Trainium chips, which have been created for AI workloads and computational structures.
AWS and “leading AI research institutions” have setup dedicated funding for new research and student education.
Amazon will also host its Amazon Research Awards for proposals, which will provide selected proposals with AWS Trainium credits and access to the Trainium UltraClusters for their research.
The Build on Trainum programme may also represent and more cost-efficient approach for universities as developing frontier AI models and apps needs a lot of computing power. Amazon reports that the associated costs have forced a number of universities to cut down on their AI research.
Grants will also be available and those who receive them will have access to AWS’s technical education and enablement programs for Trainium.
Build on Trainum will help to grow the amount of innovative ideas being investigated and support the training of AI experts.
Christopher Fletcher, Associate Professor of Computer Science Research at the University of California at Berkeley, and a participant in Build on Trainium, was full of praise for the programme: “Trainium is beyond programmable—not only can you run a program, you get low-level access to tune features of the hardware itself.
“The knobs of flexibility built into the architecture at every step make it a dream platform from a research perspective.”
Fletcher continued: “AWS is really enabling unexpected innovation.
“I walk across the lab and every project needs compute cluster resources for something different.
“The Build on Trainium resources will be immensely useful—from day-to-day work, to the deep research we do in the lab.”
Researchers of the Build on Trainum program will be able to collaborate with other industry experts to help progress an idea into a working solution.
These researchers will also publish papers on their work and will bring the code into the public domain using open-source machine learning software libraries, which will then become the “foundation for the next round of advancements in AI”.
