Microsoft CTO Kevin Scott on how AI and art will coexist in the future

Computational Demands

The computational demands of AI are heavily reliant on hardware like Nvidia's H100 GPUs, which are integral to Microsoft's AI models. Kevin Scott acknowledges this dependency but highlights the flexibility Microsoft has in potentially switching to other vendors like AMD without disrupting customer experience 1. He emphasizes the importance of using a portfolio of models to optimize costs and performance, allowing for a mix of large and small models to handle different tasks efficiently 2. This approach not only addresses cost and latency concerns but also ensures sensitive data can be processed locally when necessary 3.

You want both, and you want both to be getting cheaper and faster and more performant and higher quality over time.

--- Kevin Scott

   

Cost Strategies

Managing the costs associated with AI development is a critical focus for Microsoft. Kevin Scott discusses the early stages of pricing generative AI, emphasizing the need to find a balance that maximizes user value without overpricing 4. He notes that OpenAI's recent cost reductions for developers, achieved through performance optimizations, demonstrate the potential for lowering AI costs 2. This strategy involves leveraging both large and small models to enhance efficiency and reduce expenses, ensuring that AI remains accessible and valuable to a broad audience.

The market sets the price. If the market tells us that the price for these things is too high, then the price goes down.

--- Kevin Scott