How Should We Define ‘Open’ AI?
From The New Stack: 2024-03-31 09:09:29
The National Information Standards Organization (NISO) is working on defining what constitutes open artificial intelligence (AI). Thomas Padilla, a deputy director at Archive.org, discussed the topic at the NISO Plus conference, emphasizing AI’s impact on achieving collective goals. Meta’s Llama 2 claims to be open source, but restrictions on large-scale commercial use have raised questions about its authenticity. Padilla points to OLMo as a model with more open-source spirit. Transparency and accountability are key factors for open AI, according to Padilla, who highlighted the need for documentation and community collaboration. Distributed AI Research Institute, focusing on reducing research impact on marginalized communities, offers a roadmap for responsible AI development. Finally, Padilla emphasized the importance of adopting AI sustainably by considering interdependencies and evaluating the full lifecycle of technology. Incorporating multiple perspectives into AI development will be crucial for the field’s future.
Read more at The New Stack: How Should We Define ‘Open’ AI?