Open-source AI models, which allow anyone to view and manipulate the code, are growing in popularity.
This is due to variables like cheaper costs for developing high-quality AI models, enhanced openness and auditability of open-source systems, and the potential for open-source models to encourage economic innovation. However, there are issues regarding open-source AI models’ security and abuse. Some researchers are concerned that these models may be exploited to develop synthetic biological weapons or to disseminate disinformation.
With over 37 open-source language models available, some of which perform almost as well as the biggest models, the open-source AI movement is developing. Two prominent competitors in the AI sector, Meta and Mosaic, have made available open-source models. The LLaMA model from Meta includes 65 billion parameters, and the MPT-30B model from Mosaic is reported to outperform the original GPT-3.
These releases make it easier for smaller businesses and researchers to access high-quality AI models, and they help to increase the transparency and auditability of AI systems.
The sources for this piece include an article in Axios.