AI
Microsoft unveils the smallest AI Phi-3.
Microsoft introduces the Phi-3 Mini: Launch of a series of compact AI models, three are planned in total.
Microsoft has released the next version of its lightweight AI model, Phi-3 Mini. This model, the first in a series of smaller models, includes 3.8 billion parameters and is specifically optimized for use on personal devices such as smartphones and laptops. Compared to larger language models like GPT-4, Phi-3 Mini is trained on a smaller dataset and is now available on platforms such as Azure, Hugging Face, and Ollama.
Eric Boyd, Corporate Vice President of Microsoft Azure AI Platform, explains that Phi-3 Mini is as capable as larger language models like GPT-3.5, but in a more compact format. The smaller size leads to lower operating costs and improved performance on end devices.
Microsoft also plans to introduce additional models in this series, including Phi-3 Small with 7 billion parameters and Phi-3 Medium with 14 billion parameters. The development of these models is part of a corporate strategy that focuses on more efficient, resource-conserving AI models designed specifically for less complex tasks such as document summarization or programming assistance.
The training methods for Phi-3 are inspired by pedagogical approaches. Developers used a curriculum based on simplified texts, similar to the stories that are read to children. This method helps the model to grasp basic concepts and structures more quickly. According to Boyd, the training was supported by "children's books" created by another LLM to systematically train Phi-3.
Despite the ability to improve complex tasks such as programming and logical reasoning, the Phi-3 Mini cannot match the breadth of knowledge of larger models like GPT-4. However, for many companies using smaller, specific datasets, it provides a cost-effective and powerful alternative.
With this latest development, Microsoft continues to focus on innovations in the field of artificial intelligence, with an emphasis on practical applicability and efficiency. This underscores Microsoft's ambition to remain a leader in the development of AI solutions that are both powerful and accessible.
