The working paper proposes a dual-layer transparency regime requiring AI-gatekeepers to disclose aggregate testing methodologies and standards at the collective level while providing compact certifications or scores at the individual level.
Sources
- AI Gatekeepers (blogs.law.ox.ac.uk)
- https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6819158 (papers.ssrn.com)
- https://www.ecgi.global/publications/working-papers/ai-gatekeepers (ecgi.global)
More in Technology
- Researchers created a benchmark called VPBench, a visually prompted benchmark containing 16 visual marker variants.
- A study comparing SBLD configurations (using minimum, median, and maximum available data) to textual search used Wilcoxon signed-rank tests and the Vargha–Delaney measure of stochastic superiority.
- SBLD and textual searches for failure-related patterns were roughly equally effective at reducing diagnostic effort, but textual search achieved slightly better recall.
- ADRON-RM is part of the science payload aboard the European Space Agency's Rosalind Franklin rover.
- Agreement between AI models and physician majority vote on NEJM Image Challenge cases was low, with Cohen's kappa of 0.08 for both GPT-4o and Claude 3.7, and 0.24 for Doubao.
- Adding only a small fraction of ranking data to the training blend enabled instruction-tuned LLMs to outperform existing expert ranking models.