Executive Briefing: How Microsoft, Bayer, and Discovery Use AI on the Data You Can't Upload

Large organizations are addressing the challenge of using AI with sensitive data by fine-tuning smaller models on proprietary information, enabling quick, specialized responses. Companies like Bayer and Discovery Bank have successfully implemented this, reducing processing times significantly. For individual use or testing, tools like LM Studio allow sensitive documents to be processed locally on an air-gapped machine, keeping data private.

• Companies are fine-tuning AI models with proprietary data to handle sensitive information securely.
• Bayer uses fine-tuned models for complex crop-protection queries, reducing resolution time from days to seconds.
• Discovery Bank fine-tuned Azure OpenAI models to understand financial language and specific workflows, improving response times.
• Microsoft ensures customer data and fine-tuned models are not used to improve general foundation models without permission.
• LM Studio allows users to process sensitive documents locally on a disconnected laptop, avoiding data sharing with model providers.
• Local testing with LM Studio helps determine if a task is suitable for a laptop or requires an enterprise system.
• Fine-tuning models with user corrections and permissions can lead to vendor lock-in.
  https://bender.layer3.press/articles/8e63a82a-47f8-44ac-a437-8745778680c0