The AI Alliance – values, nationalities and absentees

This week 57 organizations announced their creation of the AI Alliance – a consortium designed to advance the altruistic use of open source code and methods for the creation of generative AI models (see my summary of their 6 initial ‘values’ in my Figure above). The IT and IT component suppliers involved are IBM, Dell, Sony, Meta (Facebook), Intel and AMD. Notably absent for now at least are the largest public cloud suppliers Amazon, Google and Microsoft, although confusingly each of these is a member of the ‘Partnership on AI’, which has itself become a member off the AI Alliance. For me the most notable absentees are Nvidia, which is the stand-out leader in the AI hardware market for both its GPUs and Infiniband networking and OpenAI, whose ChatGPT (developed through massive investments from Microsoft) is currently the outstanding leader in Large Language Models (LLMs). Also absent for now are other network vendors (including the market leader Cisco). I may be reading too much into the announcement, but the phrase ‘boosting contributions’ when applied to the IT hardware components involved in the alliance implies Intel’s and AMD’s GPU products.

US suppliers and universities dominate the initial make-up of the alliance (see my Figure above for the overall geographic spread of members). Missing in general are Chinese organizations, apart for Singapore’s Agency for Science Technology and Research (ASTAR). There are currently no South American, African or Australian organizations either, although many of the members have multi-national employees and/or students working in non-HQ countries.
The AI Alliance is focused on those developing, training and deploying AI models, but overlaps to a degree with other pre-existing groups such as the EU’s European AI Alliance, with which it could easily be confused. The EU has also begun the process of introducing the AI Act to be adopted in all EU countries over time. The EU as ever is looking after its citizens by banning companies using unsafe AI and promoting its ethical creation and use.

There are 6 types of organization in the AI Alliance, which I have captured in 4 boxes in my Figure above). The largest of which are universities (25), followed by AI specialist suppliers/developers (15), IT suppliers (8), other consortia (5) and hardware component suppliers (Intel and AMD only at the moment). there are only 2 commercial organizations involved for now, one in healthcare; the other in automotive insurance. Interestingly 2 people – Travis Oliphant (founder of 3 member organizations and Matt Harward (2) – have a significant interest in the success of the AI Alliance.
I fully support the AI Alliance’s encouragement of a fuller understanding of AI modelling, training and deployment; this is an IT industry sub-market full of moral, social and economic peril if not addressed by legislation and policed successfully. However I am less sure that the alliance will be successful in reducing the success of Nvidia, anymore than the promotion of Linux’s hypervisor KVM adversely affected Vmware’s dramatic rise in server virtualization. Certainly the AI Alliance needs to avoid the fate of the Ace consortium, created to challenge the dominance of Microsoft Windows and Intel x86 chips in the workstation market, but quickly sunk by being overrun by suppliers, many unaligned with the consortium’s aims.

The AI Alliance’s initial membership – December 2023