Laude Institute Announces First Batch Of ‘slingshots’ Ai Grants

In Brief

On Thursday, nan Laude Institute announced its first batch of Slingshots grants, aimed astatine “advancing nan subject and believe of artificial intelligence.”

Designed arsenic an accelerator for researchers, nan Slingshots programme is meant to supply resources that would beryllium unavailable successful astir world settings, whether it’s funding, compute power, aliases merchandise and engineering support. In exchange, nan recipients promise to nutrient immoderate last activity product, whether it’s a startup, an open-source codebase, aliases different type of artifact.

The first cohort is 15 projects, pinch a peculiar attraction connected nan difficult problem of AI evaluation. Some of those projects will beryllium acquainted to TechCrunch readers, including the command-line coding benchmark Terminal Bench and nan latest type of the long-running ARC-AGI project.

Others return a caller attack to a long-established information problem. Formula Code, built by researchers astatine CalTech and UT Austin, intends to nutrient an information of AI agents’ expertise to optimize existing code, while nan Columbia-based BizBench proposes a broad benchmark for “white-collar AI agents.” Other grants research caller structures for reinforcement learning aliases exemplary compression.

SWE-Bench co-founder John Boda nan is besides portion of nan cohort, arsenic leader of nan caller CodeClash project. Inspired by nan occurrence of SWE-Bench, CodeClash will measure codification done a move competition-based framework, which nan hopes will

“I do deliberation group continuing to measure connected halfway third-party benchmarks drives progress,” nan told TechCrunch. “I’m a small spot worried astir a early wherever benchmarks conscionable go circumstantial to companies.”

Subscribe for nan industry’s biggest tech news