Mozilla outlines the current landscape of open source AI, highlighting diverse global initiatives and adoption rates among developers. While open models see significant usage, they lag behind closed models in reaching production due to operational challenges.
Various initiatives worldwide are leveraging open source AI models. In New Zealand, a MΔori broadcaster is developing speech models for the te reo language. In Switzerland, a consortium trained a national model on public supercomputers, providing open access to the weights and code. These projects emphasize community ownership and data sovereignty.
In finance, PwC customized an open model, allowing it to serve hundreds of clients without incurring costs per token. Medical research in Lausanne collaborated with the Red Cross to create an open medical model, preparing for clinical trials. Additionally, farmers in East Africa utilize offline models for diagnosing cassava diseases, illustrating practical applications of AI in underserved areas.
According to Mozilla's survey, 79% of developers implementing AI functionality are using open models, slightly ahead of the 71% using closed models. The report indicates that both types of models often complement each other, with half of developers utilizing both.
Despite high adoption rates, a significant gap exists in production readiness. Only 51% of open model teams reach production, compared to 63% for closed models. This gap is attributed to operational tooling and trust issues rather than a lack of capability in the models themselves.
β¨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors β check the original sources. How BrevFeed works β
Mozilla outlines the current landscape of open source AI, highlighting diverse global initiatives and adoption rates among developers. While open models see significant usage, they lag behind closed models in reaching production due to operational challenges.