Microsoft's compact AI models for smartphones rival GPT-3.5 in performance
In a significant shift from the tech industry's usual emphasis on larger, high-powered AI systems, Microsoft has ventured into the domain of compact and cost-effective AI technologies. On Tuesday, the tech giant revealed three compact AI models as part of a new technology suite named Phi-3. Remarkably, even the smallest of these models showcased performance levels nearly equivalent to GPT-3.5, which powered OpenAI's ChatGPT chatbot at its launch in late 2022.
Phi-3 is designed to operate on smartphones without internet connection
The most compact model in the Phi-3 suite is designed to operate on a smartphone without requiring an internet connection. It's also engineered to function on the type of chips found in everyday computers, eliminating the need for expensive processors produced by NVIDIA. This shift toward smaller models is driven by their lower processing demands, allowing tech providers to offer these services at a reduced cost to customers.
Microsoft's move toward compact and affordable AI technologies
While these smaller systems may be less powerful and potentially less precise or smooth-sounding, the tech industry believes that customers will accept some performance trade-offs for cost-effectiveness. The ultimate goal is to make AI more accessible for applications where larger, more sophisticated models have previously been financially out of reach. Although Microsoft has not disclosed specific pricing information, the company's move toward compact and affordable AI technologies signals a significant shift in the industry's approach to accessibility and affordability.