We’re pleased to make the Laude Institute’s inaugural Slingshots grantmaking announcement official. These grants are meant to spur breakthroughs in artificial intelligence. The Slingshots program will fund 15 different projects, all tackling the hard challenges of evaluating AI systems in complicated real-world scenarios. Laude offers access to incredibly valuable knowledge and tools that are otherwise difficult to cultivate in academia. This methodology encourages groundbreaking and scientific pursuits within the field of AI.
Some of the resources available through the Slingshots program include funding, compute power, and dedicated product and engineering support. These elements give researchers the opportunity to focus without distraction on their path-breaking work. Second, they liberate themselves from the forms and modes that dominate most academic environments.
One notable project among this first batch is Formula Code. It was developed in partnership with mobility researchers at CalTech and the University of Texas at Austin. This initiative aims to better understand how generative AI agents can be used to make existing software code more efficient. As creative projects such as Formula Code take to the field, they amplify the program’s mission to address central questions about efficacy and equity in AI evaluation.
A second impressive endeavor to mention is BizBench, headquartered at Columbia University. BizBench offers an in-depth, industry-agnostic benchmarking framework that’s purpose-built for evaluating “white-collar AI agents.” This project reflects a growing recognition of the importance of evaluating AI performance across diverse applications, particularly in professional settings.
Martin John Boda Yang, the new project’s leader within the cohort. He placed a value on exceptional evaluation standards for this quickly evolving field of practice.
“I do think people continuing to evaluate on core third-party benchmarks drives progress,” said Yang. “I’m a little bit worried about a future where benchmarks just become specific to companies.”
Yang’s biggest worry is that AI evaluation metrics will become dangerously fragmented. He worries that this duplication and fragmentation could derail the overall advancement of the field. His reflections open up a much larger discussion. There is an urgent need for at least some standardized benchmarks that allow meaningful comparisons between different AI systems.
The Laude Institute’s Slingshots program is an important shot across the bow at three of the biggest challenges facing the AI landscape today. The institute provides researchers with both the tools they need and a collaborative environment. This public first approach promotes a culture of creativity and experimentation, scientific curiosity and discovery.
