The Linux Foundation’s recent integration of the Open Model Initiative (OMI) might enable the development of “more ethical” large language models (LLMs), according to experts.
“A primary goal for OMI becoming a part of the Linux Foundation is to promote an ethical approach to using data (text/images) for training generative AI models,” stated Abhigyan Malik, who is a practice director focusing on data, analytics, and AI at Everest Group.
Malik also highlighted the growing challenges in utilizing ethically sourced data due to increasing data protection awareness and evolving privacy and usage policies of popular data sources.
Notably, several major proprietary LLM companies, including Open AI and Stability AI, are currently embroiled in legal actions alleging copyright violations in their model training processes.
The Open Model Initiative (OMI), established in June by Invoke, Civitai, and Comfy Org, focuses on uniting developers, researchers, and businesses to further the development of open and permissively licensed AI-related model technologies.
As described by the Linux Foundation, permissive licenses facilitate effortless participation and contribution sharing within the community, devoid of secondary obligations.
“This benefits particularly those software areas that necessitate the capability for software creators to distribute proprietary software based on the open-source codebase without disclosing their modifications,” stated the Foundation in its open-source software guide.
The primary goal of OMI is to combine deep knowledge in model training and inferencing to craft models that match or surpass the quality of proprietary models from companies like OpenAI, Google, and AWS, while avoiding restrictive licensing conditions that inhibit the utilization of these models.
The OMI, overseen by a community-led steering committee, will set up governance structures and collaborative groups to enhance community involvement in development.
To understand community needs on future model research and training, OMI will carry out a survey, as per the Linux Foundation’s announcement. The organization plans to also establish universal standards to improve model interoperability and the management of metadata.
Furthermore, the OMI will create a transparent dataset for model training and initiate an alpha test model dedicated to red teaming.
As stated by the Foundation, the primary aim of this initiative is to introduce an alpha version of the model, complete with fine-tuning scripts, to the community before year’s end.
The significance of this move for enterprises lies in the unavailability of source code and the license restrictions from LLM-providers such as Meta, Mistral and Anthropic, who put caveats in the usage policies of their “open source” models.
Meta, for instance, according to Everest Group’s other AI practice leader Suseel Menon, does provide the rights to use Llama models royalty free without any license, but does not provide the source code.
“Meta also adds a clause: ‘If, on the Meta Llama 3, monthly active users of the products or services is greater than 700 million monthly active users, you must request a license from Meta.’ This clause, combined with the unavailability of the source code, raises the question if the term open source should apply to Llama’s family of models,” Menon explained.
In contrast, OMI’s objective, according to analysts, is to create models that don’t present enterprises with caveats and are more freely accessible.
OMI’s vision and objectives elicited a variety of responses from industry experts.
Hyoun Park, the chief analyst at Amalgam Insights, expressed optimism that OMI would foster the creation of more reliable and uniform standards across open source models, enhancing their interoperability. On the other hand, Malik from Everest Group had reservations, doubting OMI’s ability to compete with major players such as Meta and Anthropic.
“The creation of large language models demands immense computational resources and has already cost large tech companies and start-ups billions in terms of capital expenditures to reach their current level with both open-source and proprietary large language models,” commented Malik. He highlighted that this high cost poses a significant hurdle for community-driven large language models.
The AI practice leader also noted that earlier efforts to launch community-based large language models did not achieve wide adoption, mainly because models from larger organizations tend to show superior performance across various metrics.
“Malik cites BLOOM as a notable attempt at utilizing an open LLM to foster a community-driven model, though its adoption has been limited due to its non-chat focused design and other inefficiencies,” he mentioned.
Nevertheless, the AI practice leader believes that OMI could find its footing in specific segments of the content creation industry, such as 2D/3D imagery, adaptation, visual design, and editing, as it continues to evolve its models.
“The targeted niches, like 3D image creation or catalogue image generation and editing for retail, are well-suited for the application of these models, potentially leading to effective performance in these areas,” explained Malik.
According to Malik, this approach is promising, considering that platforms like Invoke and Civitai are already leveraging AI in professional studios and creative hubs respectively.
One of the other use cases for OMI’s community LLMs is to see their use as small language models (SLMs), which can offer specific functionality at high effectiveness or functionality that is restricted to unique applications or use cases, analysts said.
Currently, OMI’s GitHub page has three repositories, all under Apache 2.0 license.
ChicagoVPS is your gateway to unparalleled hosting solutions. Our state-of-the-art datacenters and powerful network ensures lightning-fast speeds and uninterrupted connectivity for your websites and applications. Whether you’re a startup looking for scalable resources or an enterprise in need of enterprise-grade hosting, our range of plans and customizable solutions guarantee a perfect fit. Trust in ChicagoVPS to deliver excellence, combining unmatched reliability and top-tier support.
For Inquiries or to receive a personalized quote, please reach out to us through our contact form here or email us at sales@chicagovps.net.
