What considerations are behind Meta's collaboration with IBM to establish an open-source AI alliance?
TechFlow Selected TechFlow Selected
What considerations are behind Meta's collaboration with IBM to establish an open-source AI alliance?
Neither OpenAI nor NVIDIA is a member of the AI alliance.
Navigating Web3 tides with focused insightsBy Alpha Rabbit
According to a Bloomberg report on December 5, 2023, Meta and IBM are currently working with more than 40 companies and institutions to build an industry organization dedicated to open-source artificial intelligence efforts, aiming to share technology and reduce risks.
The organization, named the AI Alliance, will focus on the responsible development of AI technologies, including tools for AI information security and functional safety. In addition, the AI Alliance will work to increase the number of open-source AI models, develop new hardware, collaborate with academic researchers, and support "open innovation" and "open science" within the AI field.
Supporters of open-source AI believe it is a more effective way to cultivate highly complex systems. Over recent months, Meta has been releasing open-source versions of its large language models (Llama), which form the foundation of AI chatbots.
Nick Clegg, President of Global Affairs at Meta, stated in the announcement: “We believe that developing AI openly is better because more people can benefit from it, build innovative products, and ensure safety.”
The organization will eventually establish a governing council and a technical oversight committee. Participants include Oracle, AMD, Intel, Stability AI, as well as academic research institutions such as the University of Notre Dame and the Mass Open Cloud Alliance.
What are the functions and attributes of the AI Alliance?
Members of the AI Alliance will first form working groups, a governing council, and a technical oversight committee, focusing on advancing areas such as metrics for AI "trust and verification," hardware and infrastructure supporting AI training, and open-source AI models and frameworks.
They will also develop project standards and guidelines, then collaborate with “existing key initiatives” from governments, non-profits, and civil society organizations that are doing valuable and consistent work in the field of AI.
Some reflections
Public information shows that Google, OpenAI, and Microsoft have been major critics of Meta’s open-source AI approach, arguing that this method could be dangerous and may fuel the spread of misinformation. However, Meta continues its planned open-sourcing strategy, releasing text-generating models like Llama. Of course, everything has two sides—while these models may be misused by bad actors, many developers have also built useful applications based on them.
Meta’s Chief AI Scientist Yann LeCun was quoted in an interview with The New York Times saying: “The platform that ultimately wins will be the open one”—he is one of over 70 influential signatories calling for greater openness in AI development. Of course, LeCun’s view holds merit; it is estimated that Stable Diffusion, the open-source AI image generator released by Stability AI in August last year, now accounts for 80% of all AI-generated images.
Why is IBM leading the AI Alliance? What is IBM thinking?
Background: IBM has recently seen some profit gains, mainly driven by enterprise demand for generative AI, but the company remains at a disadvantage amid intense competition from Microsoft and OpenAI (and to some extent Google), who are jointly developing products that directly compete with IBM’s enterprise AI services.
In November this year, IBM announced a collaboration with VMware to bring IBM watsonx into enterprise on-premises environments via VMware Private AI and Red Hat OpenShift. IBM and VMware have long maintained a joint innovation lab where engineers conduct R&D, solving customer problems by launching new products to market.
VMware and IBM plan to develop a validated reference architecture enabling customers of both companies to leverage hybrid cloud and on-premises (regulated) environments for training and fine-tuning ML models—including large language models (LLMs) used in generative AI applications. For example, regulatory restrictions or sheer data volume might prohibit transferring data from a bank or financial institution in one country to a hyperscaler, making cross-location data transfer impossible.
Additionally, IBM Consulting has established a Center of Excellence for Generative AI, currently staffed with over 1,000 consultants specialized in generative AI, collaborating with global clients to boost productivity in IT operations and core business processes such as human resources or marketing.
Of course, neither OpenAI nor NVIDIA are members of the AI Alliance. With IBM and Meta pushing open-source AI and bringing along numerous small and medium-sized companies, the endgame of generative AI remains full of uncertainties—this is precisely the charm of technology.
References:
1. IBM and VMware Help Enterprises Adopt Generative AI with watsonx On-Premises - VMware News and Stories
2. Meta and IBM form an AI Alliance, but to what end? | TechCrunch
3. https://www.bloomberg.com/news/articles/2023-12-05/meta-ibm-create-industrywide-ai-alliance-to-share-technology
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
