OpenAI CEO Sam Altman announced the indefinite postponement of the AI open model’s launch, initially slated for this week, due to the need for extensive safety tests. Altman emphasized the importance of careful preparation, stating that once the model’s parameters are released, they cannot be reversed. He highlighted that the potential for misuse of open-source AI necessitates thorough internal risk assessments. This open model has garnered significant attention this summer alongside GPT-5, which is a cloud-based commercial offering. While OpenAI aims for a high standard in open-source models, the competition is heating up, evidenced by Chinese startup Moonshot AI’s introduction of ‘Kimi K2,’ a public model boasting 1 trillion parameters that has outperformed GPT-4.1 on specific benchmarks. OpenAI is committed to launching its model only when fully confident in its safety and effectiveness, reflecting the broader industry’s focus on responsible AI development.
Source link
Share
Read more