The Laude Institute has launched its inaugural Slingshots grants program, dedicated to fostering advancements in the field of artificial intelligence (AI). This initiative serves as an accelerator for researchers, offering essential resources that are typically scarce in academic environments, such as funding, computational power, and engineering assistance. Recipients of the grants are required to deliver tangible outcomes, ranging from startups to open-source projects, in return for the support provided.
The initial cohort comprises fifteen projects, with a strong emphasis on tackling the challenge of AI evaluation. Among the featured projects are Terminal Bench, a command-line coding benchmark, and the latest iteration of the renowned ARC-AGI project. Additionally, initiatives like Formula Code from CalTech and UT Austin seek to evaluate AI agents’ performance in optimizing existing code, while BizBench, developed at Columbia University, aims to establish a comprehensive benchmark for ‘white-collar AI agents.’ Furthermore, several grants explore novel approaches to reinforcement learning structures and model compression techniques.
Noteworthy among the recipients is SWE-Bench co-founder John Boda Yang, leading the innovative CodeClash endeavor. Inspired by the success of SWE-Bench, CodeClash will introduce a dynamic competition-based framework to assess code quality. Yang emphasizes the significance of ongoing evaluations on standardized benchmarks to propel advancements in the AI domain and expresses concerns about the potential fragmentation of evaluation metrics in the future.
Source: TechCrunch