The AI Alliance, a new collaboration among 50 companies and institutions led by IBM, Meta and AMD, is expected to bring the rivalry between open source and proprietary technology to new frontiers in generative AI, challenging current big players including AWS, Microsoft, and Nvidia.
“Meta and IBM are both embracing open source and contributing to it hugely in contrast to Google, Amazon, and Microsoft, who do not contribute to open source near as much. This alliance alongside AMD and others is a strong foot forward for the open source community from software to hardware,” said Dylan Patel, chief analyst at semiconductor research and consulting firm Semianalysis.
“AMD is a big beneficiary due to the open software ecosystem benefiting them more than Nvidia, which has a big portion of their stack closed,” Patel added.
What is the AI Alliance and what will it do?
The AI Alliance, announced on Tuesday, comprises tech companies as well as scientific and academic institutions including Oracle, Red Hat, CERN, Cleveland Clinic, Cornell University, Dartmouth College, Dell Technologies, Hugging Face, Intel, Linux Foundation, MLCommons, and NASA.
While IBM and Meta have played an important catalyst role, the AI Alliance is a collaborative initiative of its member organizations, said Anthony Annunziata, director of accelerated discovery incubation at IBM Research.
“There is no official headquarters, and our governance is intended to be representative of all our members,” Annunziata added.
The AI Alliance, according to IBM, will enhance and start open source software projects working with open source foundations like the Apache Foundation and Linux Foundation, but its mission and scope is broader than these foundations.
“The scope of the Alliance includes developing safety and trust standards and benchmarks, AI foundation models and applications, as well as initiatives in the field of research and education, and policy advocacy,” Annunziata said, adding that the alliance will begin its work with the formation of member-driven working groups across all its focus areas.
The Apache Foundation, for example, is focused on the governance of open-source software projects.
The alliance will also establish a governing board and technical oversight committee dedicated to advancing projects in specific focus areas while partnering with existing initiatives from governments, nonprofit and private-sector organizations, according to a statement from IBM.
Explaining further, Annunziata said that the AI Alliance members will utilize pre-existing collaborations as well as create new ones with government agencies globally, including in the US and EU, to identify opportunities that develop open AI resources that meet the needs of business and society equally and responsibly.
“These efforts will also include advocating for smart policymaking that is focused on regulating the application of AI,” the director added.
AI Alliance to counterbalance proprietary software?
The AI Alliance, according to IBM’s Annunziata, is seeking to “restore the balance in the ecosystem between open and proprietary work by leveraging a critical mass of resources to democratize the creation, deployment, and management of AI technology, especially open foundation models.”
In fact, Arun Chandrasekaran, market research and advisory services firm Gartner’s vice president of technology innovation, said that the alliance is “a counterbalance to the closed-sourced ecosystem of large technology companies.”
“Meta is one of the leading research labs in AI and by partnering with enterprise-focused vendors such as IBM, both IBM and Meta, can enable open source software models to be spearheaded by the alliance partners,” Chandrasekaran added.
However, he pointed out that the success of this alliance is predicated on the quality of the models they will release, the permissibility of their licensing, the rigor of their model audit process and the transparency they can bring to the overall model training and release process.
This alliance, according to Constellation Research’s principal analyst Andy Thurai, seems like IBM’s “B team” to compete with the likes of Microsoft, Google, and OpenAI. “IBM might benefit a lot if the consortium start to use IBM’s AI chips, cloud, software and others for their internal use and/or integrate with their products,” Thurai said, adding that he was skeptical about the value add to other companies in the alliance.
However, Ritu Jyoti, group vice president of research at IDC, believes that the alliance is IBM and Meta’s strategy to provide compelling and viable alternatives to offerings from their rivals.
AI Alliance to help AMD, Intel take on Nvidia?
The AI Alliance, which has members such as AMD, Intel and Cerebras, will help these chip makers and hardware accelerator software providers take on Nvidia, as it claims that it will foster an AI hardware accelerator ecosystem by boosting contributions and adoption of essential enabling software technology, analysts said.
Nvidia is a dominant player in the chipmaker space, including hardware that’s required in developing large language models and inferencing them.
Nvidia’s moat is its software ecosystem to run its chips and the creation of a developer community around it while partnering with other technology providers, such as AWS, Google, Microsoft, Oracle, Dell, HPE and Lenovo.
“AMD, Cerebras, IBM, and Intel, all of whom compete with Nvidia, have had a tough time because of Nvidia’s software prowess. So this alliance proposes to develop an open source software ecosystem that is as rich as the proprietary CUDA ecosystem and that runs on these accelerator types,” said Peter Rutten, research vice president of performance intensive computing at IDC.
“This would allow for more choice in the market, help these four vendors gain more share, diversify the ecosystem, possibly bring prices down on accelerators, and therefore make AI development more accessible,” Rutten added.
In October, AMD announced plans to acquire open source machine learning and AI software provider Nod.ai as the chipmaker looks to expand its AI capabilities in a bid to shore up competition against Nvidia.
Will the Al Alliance be able to set standards for models and AI safety?
While some analysts believe that the AI Alliance will have a positive impact on model standards and AI safety, others believe that the consortium has not put out enough information to gauge its potential.
“The Alliance also has an opportunity to work with regulators and bring clarity to the process of regulating open source models and approaches that constitute safe release of models,” Gartner’s Chandrasekaran said.
However, Constellation Research pointed out that it wasn’t clear what benchmarks and standards the Alliance will offer.
Depending on what is being measured — including models and generative AI assistants along with their biases — there are several benchmarks and standards already available. These include GAIA, MLPerf, Fairlearn, AI fairness 360, Accenture Fairness Tool, DAWNBench, GLUE, and SQuAD among others.
Additionally, Amalgam Insights’ principal analyst, Hyoun Park, pointed out that technical standards for AI are still very much a work in progress.
“The United States National Institute of Standards in Technology (NIST) doesn’t even have a clear dictionary of terms to define AI or metrics to define AI performance and safety,” Hyoun said, adding that the Alliance has been structured in a manner to understand and collaborate with agencies trying to develop AI safety standards for most countries, including the US and EU.
Copyright © 2023 IDG Communications, Inc.