A naukri.com initiative
TechCrunch
2w
233
Image Credit: TechCrunch
We finally have an ‘official’ definition for open source AI
- The Open Source Initiative (OSI) has released version 1.0 of its Open Source AI Definition (OSAID) after several years in the making.
- The OSAID aims to provide a universal standard by which anyone can determine whether an AI is open source, which could help prevent potential confusion that could arise from the misuse of the term.
- To be considered open source per the OSAID definition, an AI model must provide enough detailed information for an interested party to recreate it and disclose details about its training data.
- AI model developers can go ahead and use any open source model for any purpose without requesting anyone's permission.
- The Open Source Initiative (OSI) doesn't have enforcement mechanisms, but it intends to flag models described as “open source” but which fall short of the definition.
- A study last August found that many “open source” models are open source in name only and don't democratize AI but tend to expand centralized power.
- Commenting on the definition, Luca Antiga, CTO of Lightning AI, said that it doesn't go far enough, for example, concerning proprietary training data licensure.
- OSI established a committee that will be responsible for monitoring how the OSAID is applied, proposing amendments for future versions.
- The OSI servers are funded by companies like Amazon, Google, Microsoft, and Salesforce.
- Meta's reluctance to reveal training data may be due to fears of a legal backlash and that companies cite proprietary data set building and tweaking as a competitive advantage.
Read Full Article
14 Likes
For uninterrupted reading, download the app