Microsoft Phi-3, its smallest language, released on Tuesday is the AI model so far. Small AI models are important because they have the potential to run on smartphones. The latest AI model is the successor of Phi-2 released in December 2023, with a higher training database and larger parameters. The added parameters help the AI model understand and respond to more complex problems than its predecessor. It is said that the number of training parameters is 10 times greater than Phi-3.
The preprint paper A detailed introduction to Small Language Models (SLMs) has been published on arXiv. However, since arXiv does not conduct peer review, the validity of this claim has yet to be determined. AI enthusiasts can test AI models via Azure and Ollama. Counterweights for a Hug face index Fee-3-min have also been developed, but the weights have not yet been released.
In terms of performance, the AI model was trained on 3.3 trillion tokens, which are units of data containing words, phrases or subsets of words fed into the system to train the AI model. It has 3.8 billion parameters, highlighting the level of complexity that a chatbot can understand. They are basically neural connections where each point is connected to different points that contain knowledge about a particular subject and information about the main point.
Microsoft claims – based on internal benchmarks – that competing models much larger than the Chabot SML include the Mixtral 8x7B and GPT-3.5. AI adapts to the chat format, meaning it can answer conversational questions. “We also provide some preliminary parameter scaling results for the 7B and 14B models trained on 4.8T tokens, called phi-3-small and phi-3-medium, which are both more efficient than phi-3-mini,” Tech says. monster
Reuters reports AI models designed to perform simple tasks are also hosted on Microsoft Azure and Ollama. The company did not disclose details about Phi-3-mini’s open source license. Notably, Grok AI recently acquired an Apache 2.0 license allowing academic and commercial use.
Affiliate links may be generated automatically – see our ethics statement to learn more.
Source link
