Microsoft has announced Phi-3 Mini, a small language model trained on children's books written by other language models. The understanding of the AI model should be in line with approximately GPT 3.5 of OpenAI.
Phi-3 Mini does not appear to be online yet at the time of writing, but Microsoft has put the paper about it online. This is a model with 3.8 billion parameters. That is relatively little and the advantage is that it is light enough to run on laptops and smartphones. According to Microsoft in an article by The Verge, the training data came partly from the web and partly from 'children's books' created by AI chatbots with only the 3000 most common words.
The intention is to demonstrate that it By using the right training data it is possible to build an equally powerful model with fewer parameters and according to Microsoft, Phi-3 Mini can compete with GPT 3.5 and Mixtral. The latter has 45 billion parameters.
The Mini version is the first in a series of three Phi-3 models that Microsoft wants to release. In addition, Phi-3 Small has seven billion parameters and Phi-3 Medium has fourteen billion parameters. It is unknown when they will be released. Microsoft puts its Phi models online on HuggingFace, but they can also be used via Microsoft's own cloud platform Azure and desktop app Ollama.
It is unknown what Microsoft's plans are for Phi. It generally uses OpenAI's GPT models for its CoPilot products, which it trains itself for the services in which Microsoft integrates the model.
/i/2006659970.png?f=imagemedium)
/i/2006659972.png?f=imagemedium)