OPEN LANGUAGE MODEL

The Allen Institute, a nonprofit organization, is on a mission to make artificial intelligence (A.I.) more transparent and accessible. They have released the largest open dataset, which has been downloaded over 500,000 times. This dataset will be used to train a new A.I. program, Open Language Model (OLMo), set to be released soon.

Many commercial A.I. models are often considered “black boxes,” meaning their inner workings are hidden. The Allen Institute wants to change that. They aim to create a “glass box” A.I., where the technology is transparent, and people can better understand how it operates.

There are only a few open-source A.I. models of this size, including Meta’s LLaMA and Falcon, backed by the Abu Dhabi government. The Allen Institute is well-suited for this endeavour as it’s well-funded yet adheres to academic values, focusing on open science and A.I. technology advancement.

They are not alone in this effort. The nonprofit Mozilla Foundation invested $30 million in Mozilla.ai, a startup dedicated to open-source software for A.I. engines, such as the Allen Institute’s. Their aim is to make these A.I. technologies more user-friendly and transparent.

This initiative aligns with Mozilla’s mission to keep the internet open and accessible to all, ensuring that technology and economic power don’t become too concentrated. The Allen Institute’s commitment to transparency and collaboration is a positive step towards democratizing A.I. and advancing open science.