Meta and IBM launch new AI alliance, but where's Google and Microsoft?

IBM and Meta are set to lead a new AI alliance focused on the responsible, open development of artificial intelligence (AI), but notable industry competitors will not participate.

The coalition’s stated goal is to support a transparent and collaborative approach to AI development that the duo said will help produce evaluation standards and tools for building AI systems, as well as educational content to inform public discourse on the technology.

The two technology giants are joined by over 50 founding members and collaborators, including hardware and infrastructure producers, model builders, research institutions, and open source foundations.

Other industry stakeholders pledging support for the alliance include AMD, Intel, Hugging Face, Dell Technologies, Cerebras, Sony, and Oracle.

A number of research institutions have also joined the coalition including notable institutions  such as Harvard University, Yale University, Imperial College London, the University of Tokyo, UCLA Berkeley, and Notre Dame.

IBM AI Alliance: What will the coalition do?

The focus of the alliance appears to be on ensuring the future of AI development is open and retains considerations around safety and trust. 

In a blog post announcing the partnership IBM said an open, collaborative approach to AI innovation will be critical to advancing the technology. 

“Open and transparent innovation is essential to empower a broad spectrum of AI researchers, builders, and adopters with the information and tools needed to harness these advancements in ways that prioritize safety, diversity, economic opportunity and benefits to all.”

IBM and Meta hope the coalition will bring together influential members in the sector to pool resources and knowledge to ensure safety concerns are adequately addressed during the ongoing wave of AI development.

This will see the alliance launch or enhance AI projects that satisfy a list of core objectives centered around fostering a healthy environment for responsible AI innovation.

One of these objectives aims to develop and implement evaluation standards for AI systems, produce a list of vetted safety and trust tools, and advocate for the widespread adoption of these standards and tools.

These projects should also facilitate wider AI skills development on a global scale and produce resources to inform public discourse on AI, the duo said, which long-term will enable more precise and productive regulations on the technology.

Does this alliance represent a new or false hope?

Notable omissions on the list of alliance members include Google, Amazon, Microsoft, Anthropic, and OpenAI, which make up the most influential AI model developers in the sector.

Microsoft, Google, OpenAI, and Anthropic all joined the Frontier Model Forum in July 2023, perhaps explaining their disinclination to join this new alliance.

The stated goals of the partnership run along the same lines as the coalition announced by IBM and Meta, but retain a particular focus on safeguarding the development of frontier models – defined as large language models (LLMs) that surpass the most advanced models available today.

Some experts expressed pessimism about how realistic the partnership was in achieving its stated goals of cooperation with the Forum’s members being major competitors.

Is open AI collaboration a pipe dream?

There’s no doubt that new methods for improved model training - or frameworks through which models could circumvent unwanted hallucinations or harmful output entirely - are welcomed. Collaboration between the public and private sector will be a necessary part of this.

Rory Bathgate

Rory Bathgate is Features and Multimedia Editor at ITPro, overseeing all in-depth content and case studies. He can also be found co-hosting the ITPro Podcast with Jane McCallion, swapping a keyboard for a microphone to discuss the latest learnings with thought leaders from across the tech sector.

Ultimately, however, these kinds of partnerships come second to the guardrails of the market as defined by legislation and competition. AI developers will continue to collaborate on best practices and enter into dialog with world leaders over the potential risk of AI implementation, but the bar for transparency will always lie with regulators.

The ‘black box’ approach to AI development, in which training weights and the nature of data used to train AI models are kept largely secret by each company, is likely to be a sticking point for open AI development. If you don’t know the precise decisions and information that has gone into making a model, evaluation systems can only go so far.

With its focus on open source AI, Meta is arguably in a better position to engage in the kind of meaningful collaboration with academia that will be necessary to lay the groundwork for ethical AI. 

But even Meta has set limitations according to market competition. The firm’s open LLM, Llama 2, has been advertised as an open source model but can only be used by firms with more than 700 million monthly active users with permission from Meta.