What happened
The Laude Institute introduced its inaugural Slingshots AI grants program, providing financial aid, computing power, and engineering support to 15 projects. This initiative aims to convert AI research into startups or open-source projects, focusing on AI evaluation. Specific projects include Formula Code, measuring AI agent code improvement, and Columbia's BizBench, benchmarking 'white-collar AI agents'. Additional grants support reinforcement learning and model compression, with CodeClash evaluating code through a competition-driven model.
Why it matters
The introduction of new AI evaluation benchmarks, including for 'white-collar AI agents' and code improvement, creates an oversight burden for organisations. This initiative increases exposure to evolving performance metrics and validation methodologies for AI systems, potentially leading to a policy mismatch with existing procurement and compliance frameworks. Platform operators and IT security teams will face higher due diligence requirements to assess and integrate AI solutions against these emerging, potentially more complex, evaluation criteria.
