{"id":119576,"date":"2024-08-22T21:43:15","date_gmt":"2024-08-22T19:43:16","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/the-linux-foundation-zapuskaet-inicziativu-omi-dlya-prodvizheniya-otkrytyh-ii-modelej"},"modified":"2024-08-22T21:43:15","modified_gmt":"2024-08-22T19:43:16","slug":"the-linux-foundation-zapuskaet-inicziativu-omi-dlya-prodvizheniya-otkrytyh-ii-modelej","status":"publish","type":"post","link":"https:\/\/prohoster.info\/en\/blog\/news\/the-linux-foundation-zapuskaet-inicziativu-omi-dlya-prodvizheniya-otkrytyh-ii-modelej","title":{"rendered":"The Linux Foundation launches the OMI initiative to promote open AI models","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"
<\/p>\n
The Linux Foundation has announced the launch of a new initiative aimed at promoting open artificial intelligence models<\/p>\n
Last week, the Open Model Initiative (OMI) was officially presented, now operating under the auspices of The Linux Foundation. Together, they aim to stimulate innovation in the development of open AI models.<\/p>\n
OMI was created by three organizations specializing in AI: Invoke, Comfy Org, and Civitai. The initiative focuses on developing and distributing AI models for generating images, videos, and audio, which will be available under open licenses.<\/p>\n
Project management will be conducted by a committee consisting of community representatives. All models released under OMI will have licenses that remain unchanged, will not include terms for content removal, and will not require mandatory fees for access.<\/p>\n
The main objectives of this program are:<\/p>\n
\n
to create a governance structure and working groups to support collaborative efforts within the community;<\/li>\n
to conduct surveys among community members to understand their needs and expectations regarding future research and model training;<\/li>\n
to develop competitive and ethical AI models with open licenses;<\/li>\n
to establish standards to ensure model compatibility and a unified approach to metadata;<\/li>\n
to work with transparent datasets for training and to create automatic data tagging systems;<\/li>\n
to launch an alpha version of the model for targeted safety testing (Red Teaming);<\/li>\n
to publish the alpha version of the model and its configuration scenarios by the end of 2024.<\/li>\n<\/ul>\n
![\"The](\"\/wp-content\/uploads\/2024\/08\/2fd9a96755359a0be02a9b78aa84f1b4.jpg\")
<\/p>\n