AI Is Becoming More Powerful—but Also More Secretive

A recent Stanford study reveals that leading AI systems, including commercial giants and so-called open-source models, suffer from severe transparency deficits. No evaluated system achieved more than fifty-four percent on a comprehensive openness index, highlighting a pervasive industry-wide lack of disclosure regarding training data, hardware usage, and environmental impact. This opacity stems from corporate fears regarding competitive advantage and potential misuse, resulting in a significant information blackout surrounding the creation of powerful algorithms. This trend signals a dangerous shift in AI research, moving away from collaborative scientific inquiry toward proprietary secrecy. By withholding critical details about model development and data sources, companies reduce accountability and reliability, making it difficult for the broader community to assess safety risks or verify claims. Consequently, the field is becoming less capable of producing verifiable scientific advances, as external researchers cannot audit or reproduce results without full access to underlying methodologies and resources. For the open data community, this article is critically relevant as it exposes the gap between "open source" labels and actual accessibility. True open data requires transparent documentation of origins, curation, and licensing, yet even models marketed as open fail to disclose essential training data details. This lack of genuine transparency hinders the development of trustworthy AI systems and undermines efforts to ensure ethical standards, demonstrating that without rigorous disclosure requirements, the promise of open data in AI remains largely unfulfilled.

Source: wired.com
Published on 2023-10-20

Actividad de restaurantes cae un 0,73%: ¿Cuáles fueron los negocios más afectados? »

Open Data

Related news