Microsoft makes new AI team for small models

Microsoft's latest strategy in the AI game is small models. Meet their new GenAI team that’ll focus on developing smaller, more efficient AI models. At WEF, Satya Nadella (Microsoft’s CEO) mentioned the work on small models as “controlling our own destiny.”

What’s going on here?

Microsoft is focusing on small language models (SLMs) to cut down costs and computing power.

What does this mean?

Microsoft has been big on AI, especially with their heavy investment in OpenAI. But we all know, GPT-4’s too power-hungry. So now, Microsoft is creating this new team to make AI more economical. They've pulled in their top AI brains from research teams into a GenAI team to develop these SLMs, including leading researcher Sébastien Bubeck.

The GenAI team is different from Microsoft’s Turing team which focuses on making larger models. GenAI is led by Corporate Vice President Misha Bilenko and answers to Kevin Scott, the CTO.

Why should I care?

These SLMs are nifty because, with some optimizations and a narrow application range, they can do much of what the larger models do but on a diet of less computing juice. Microsoft’s Phi models, its attempt at SLMs, has gained much attention among companies and fine-tuning hackers. The Phi series of models is already being tested by Goldman Sachs. Nice, coz I’m all in for AI to become more accessible and cheaper to run for us all.