There is no free benchmark: An institutional view of legal AI benchmarking | PNAS
First, an academic study found that a Thomson Reuters AI -based legal research tool—touted as “avoiding” hallucinations — hallucinated in 33% of ...
PNAS 11
Covers: Chatbots & LLMsLegal & CourtsAI Safety & Existential RiskBias & FairnessAI Agents & AutomationGovernment & Public SectorMarkets & InvestmentGovernance & Risk ManagementRegulation & PolicyPrivacy & Surveillance
Assessing the feasibility of collective licensing of in- copyright works as training data for ... - PNAS
Copyright owners have sued several developers of large-scale generative AI systems for copyright infringement because of their uses of massive ...
Legal infrastructure for transformative AI governance - PNAS
... AI risks . Will generative AI models made available either with open ... AI poses existential risks to humanity that governments should be acting ...
Governance & Risk ManagementLegal & CourtsAI Safety & Existential RiskGovernment & Public Sector
Reply to Wang: Policy failure can be assessed from observed outcomes - PNAS
We ask whether the current journal AI policy regime has produced restraint in AI -associated writing signals or public transparency. The evidence shows ...
With qualitative research, the risks of data sharing can outweigh the rewards - PNAS
Today, confidentiality risks are also amplified by social media, surveillance, and AI tools, which make ample data about people publicly available ...
Evolvable AI : Threats of a new major transition in evolution - PNAS
Evolvable AI (eAI), i.e., AI systems whose components, learning rules, and deployment conditions can themselves undergo Darwinian evolution, ...
How public involvement can improve the science of AI | PNAS
Researchers have also reduced bias by purposively sampling juries involved in participatory evaluation to match the demographics of stakeholders in a ...
Reply to Yu et al.: Datasets, human judges, and future directions for evaluating AI– AI bias
W. Laurito et al., AI- AI bias : Large language models favor communications generated by large language models. Proc. Natl. Acad. Sci.
Could humans and AI become a new evolutionary individual? - PNAS
Y. Bengio et al., Managing extreme AI risks amid rapid progress. Science 384, 842–845 (2024). Go to reference.
The benefits and dangers of anthropomorphic conversational agents - PNAS
While the AI industry directs much effort into improving the reasoning abilities of LLMs—with mixed results—the progress in communicative abilities ...
Existential risk narratives about AI do not distract from its immediate harms - PNAS
We address this “distraction hypothesis” by examining whether a focus on existential threats diverts attention from the immediate risks AI poses today ...