Adobe-backed SlimLM delivers cloud-free mobile AI capabilities

Thanks to a new small language model (SLM) that has been developed by a team of AI researchers at Adobe Inc. alongside colleagues from Auburn University and Georgia Tech, new AI apps for smartphones will be able to process documents locally without cloud access. This SLM is called SlimLM, and all details concerning it have been published on arXiv.

"Through extensive experiments on a Samsung Galaxy S24, we identify the optimal trade-offs between model size (ranging from 125M to 7B parameters), context length, and inference time for efficient on-device processing. SlimLM is pre-trained on SlimPajama-627B and fine-tuned on DocAssist, our constructed dataset for summarization, question answering, and suggestion tasks. Our smallest model demonstrates efficient performance on S24, while larger variants offer enhanced capabilities within mobile constraints,"

revealed the research team.

SlimLM is not available to the masses yet, but this could happen soon enough. Thanks to this piece of code, a smartphone can use the power of AI locally to process documents without even going online. While cloud-based AI solutions often use far less local computing power than apps running locally, the most important attribute they offer is privacy. Until now, the big names in the field, such as Google, Apple, and Meta, have already developed such apps. Sadly, their status remains rather experimental, and none of them found their way to the masses. If the plans of the aforementioned researchers come to fruition, SlimLM could become the first such solution.

Those interested in SlimLM and similar solutions should grab Jesse Sprinter's Demystifying AI: Your Guide to Mastering Small Language Models, which is currently available for $7.11 in Kindle format while the paperback version goes for $18.99.