Microsoft announced new updates to Azure AI, focusing on customization of generative AI models and support for new models. Key highlights include the general availability of fine-tuning for GPT-4o and GPT-4o mini in Azure OpenAI Service, enabling users to tailor these models for specific needs like enhancing translation accuracy or brand voice consistency. The update also expands the Azure AI model catalog with additions like Phi-3.5-MoE, Phi-3.5-vision, Meta's Llama 3.2, SDAIA's ALLaM-2-7B, and updated Command R and Command R+ from Cohere, offering capabilities ranging from multi-lingual support to image processing and Arabic language understanding.
Microsoft emphasizes its commitment to enterprise-grade security and privacy with enhancements like Azure OpenAI Data Zones for enhanced data isolation and support for full network isolation and private endpoints in Azure AI Studio. Responsible AI features are bolstered with the addition of Correction in Azure AI Content Safety, new evaluations for assessing output quality and security, and Protected Material Detection for Code.
These updates signify a substantial step towards making generative AI more accessible and potent for businesses. By providing more powerful customization options and a diverse range of models, Microsoft empowers companies to leverage this advanced technology to address their unique challenges and achieve their goals. An intriguing example of the benefits is Lionbridge, a leader in translation automation, leveraging fine-tuned GPT models to enhance translation accuracy. Their data indicates that fine-tuned GPT models outperform both baseline GPT and Neural Machine Translation engines in languages like Spanish, German, and Japanese.
Overall, these updates demonstrate Microsoft's dedication to delivering comprehensive and robust AI solutions for businesses. By focusing on customization, security, and innovation, Microsoft paves the way for an exciting future of generative AI.
