[ad_1]
Microsoft has introduced a new, smaller AI model for businesses looking to benefit from AI’s efficiency and productivity enhancements on a smaller scale.
The company’s latest model, Phi-3-mini, offers all the abilities of other generative AI models such as content creation and summarization for small businesses, but should require fewer resources, giving making it both lighter and cheaper.
Those improvements also help the worldwide effort of democratizing AI by making it more accessible to smaller companies without the resources of multinational enterprises.
Microsoft launches Phi-3-mini AI model
A statement released by Microsoft (via Reuters) notes Phi-3-mini boasts superior performance to other small models and even some that are twice the size, surpassing benchmarks across language, coding and mathematical capabilities.
The 3.8 billion parameter language model is trained on 3.3 trillion tokens and is likened to GPT-3.5, the model that underpins the publicly available free version of ChatGPT. Microsoft Azure AI Platform Corporate VP Eric Boyd confirmed this in a statement to The Verge: “just in a smaller form factor.”
It’s small enough to be deployed on a smartphone and sits beneath 7B and 14B-parameter versions called Phi-3-small and Phi-3-medium.
The tech giant’s commitment to developing lightweight AI solution reflects a strategic shift towards catering to more diverse markets following on from the early successes that Microsoft has seen with its bigger models and its multibillion-dollar investment in OpenAI, the company behind the GPT models that underpin many of the generative AI tools available across the Microsoft suite.
Microsoft isn’t the only company exploring smaller-scale AI models – Google has already lifted the wraps off its Gemma 2B and 7B models and Meta has its own Llama 3 8B model.
Although ChatGPT and its related GPT models have stolen the show for many months, looking ahead, it’s clear that a growing diversity of models are set to offer companies endless options when it comes to building and deploying generative AI applications, which can only be good news for competition.
More from TechRadar Pro
[ad_2]
Source link
