:max_bytes(150000):strip_icc()/GettyImages-2136341890-620b2db01414456f8d6549272f9ac84f.jpg "Why Microsoft Is Building Smaller AI Systems That Can Fit in Phones")
:max_bytes(150000):strip_icc():format(jpeg)/GettyImages-2136341890-620b2db01414456f8d6549272f9ac84f.jpg)
Key Takeaways
- Microsoft announced its smallest artificial intelligence (AI) model, the Phi-3 Mini, on Tuesday.
- The small language model (SLM) can fit in a smartphone and could better position Microsoft as companies look for more cost-effective AI solutions and tech capable of running AI locally on personal devices.
- Deutsche Bank analysts said Microsoft could gain amid the “rise of the SLM” and noted that the firm anticipates a future scenario where smaller models “prevail for the majority of tasks.”
Microsoft (MSFT) announced its smallest artificial intelligence (AI) model, the Phi-3 Mini, as the early AI leader expands its small language model (SLM) offerings.
The “tiny but mighty” AI system is one that Microsoft reports rivals OpenAI’s GPT-3.5 and Mistal AI’s Mixtral 8x7B in performance, even though it is “small enough to be deployed on a phone,” according to the company’s technical report.
Microsoft also announced initial results for its Phi-3-Small and Phi-3-Medium, larger SLMs that are more capable.
The company said that the new Phi-3 models are built on a “scaled-up version” of the training dataset it used to train previous versions of its SLMs, including Phi-2 announced in December of 2023.
Large language models (LLMs) like OpenAI’s GPT-4, Amazon-backed (AMZN) Anthropic’s Claude 3, and Meta’s (META) Llama 3 “have created exciting new opportunities to be more productive and creative using AI,” but “their size means they can require significant computing resources to operate,” Microsoft said in its announcement.
The company says its SLMs “offer many of the same capabilities found in LLMs but are smaller in size and are trained on smaller amounts of data.”
Deutsche Bank analysts said that Microsoft could be well positioned to gain amid the “rise of the SLM,” noting they can “see a scenario in the future whereby a smaller model (or the combined power of multiple smaller models) prevail for the majority of tasks, and the biggest LLMs are used more selectively for situations that require a massive knowledge base.”
Microsoft has established itself as an early AI leader through its ongoing partnership with ChatGPT maker OpenAI. The company has found some success in monetizing its AI offerings to its enterprise customers, though analysts have suggested Microsoft could see greater contributions from AI going forward.
The Phi-3 SLMs could give Microsoft a competitive edge as companies look for solutions to run AI locally on computers or phones. Analysts have indicated that tech capable of running AI locally could catalyze an accelerated phone upgrade cycle.
Microsoft shares were up roughly 1.6% at $407.55 around 3:45 p.m. ET on Tuesday, contributing to the stock’s 8% gain so far this year. The company is set to report earnings on Thursday after the closing bell and could provide updates about its AI initiatives.