Microsoft Develops New AI Model To Rival Google, Anthropic, And OpenAI

According to Foresight News, Microsoft has internally developed a new artificial intelligence model, which is said to be capable of competing with models from Google, Anthropic, and OpenAI. The new model, known internally as MAI-1, is led by former Google AI leader Mustafa Suleyman. Suleyman recently served as the CEO of AI startup Inflection, most of whose employees were hired by Microsoft in March this year. Microsoft also paid $650 million for the intellectual property rights of the startup. Sources reveal that the new model is separate from the Pi model previously released by Inflection and may be based on the startup's training data and other technologies. The MAI-1 model is expected to be much larger than the small open-source models Microsoft has trained in the past. It will require stronger computing power and training data, thus increasing the cost. The MAI-1 model will have about 500 trillion parameters, or adjustable settings, to determine the content learned during the training process. In comparison, OpenAI's GPT-4 has over 1,000 trillion parameters, while small open-source models released by companies like Meta Platforms and Mistral have 70 billion parameters.