OECD audits global capacity to support government AI

By on 28/01/2021 | Updated on 28/01/2021
Counting clouds: the OECD is totting up the world’s government AI processing and storage capacity. Credit: StockSnap/Pixabay

The OECD has embarked on an exercise to benchmark the computational capacity needed by national governments, as they seek to develop state-sponsored artificial intelligence (AI) solutions to transform services or boost economic growth.

The new OECD Task Force will assess the “computational infrastructure” at governments’ disposal: the data centres and cloud computing capacity needed to “train” machine learning or AI models, and the software, datasets and personnel to underpin AI solutions.  

Measuring and comparing computational firepower will be essential to governments as they make investment and policy decisions on AI, the intergovernmental economic group believes.

The task force will also assess the roll-out of “national AI clouds” – state-backed or state-controlled isolated cloud networks, usually run in partnership with private companies such as Google or Microsoft, but confined within the geographic border of specific countries.

According to the OECD, national AI clouds are already a growing trend in Asia, Europe, and the Middle East.

ONE vision

The task force is one of several projects from the OECD Network of Experts on AI, known as ONE AI, which in turn is part of the OECD’s AI Policy Observatory, launched in 2020.

Explaining the project, the ONE AI webpage says: “While data and machine learning algorithms have received significant attention in policy circles at the OECD and beyond, the computational infrastructure that makes AI possible has been comparatively overlooked.

“Since understanding domestic AI compute capacity is increasingly critical to formulating effective AI policies and informing national AI investments, the OECD is focusing efforts on this area in 2021.”

The OECD is currently recruiting members of the new taskforce, which will be co-chaired by Keith Strier, vice president of worldwide AI initiatives at NVIDIA and formerly global AI leader at consulting firm Ernst and Young.

Task force members will be drawn from hardware providers and manufacturers, cloud service providers, academia, data centre operators and consulting firms, the webpage says.

Speaking to US tech news website Venturebeat, California-based Strier said: “There’s a remarkable gap in understanding AI despite it being a publicly identified policy priority for many governments

“If you’re a prime minister or the president of a country, you want to know three things: How much AI infrastructure do I have? How does it compare to other countries? And is it enough? Those sound like simple enough questions, but if you can’t answer them, how could you possibly know you’re making the right investments?”

Count your assets

Venturebeat also named Jack Clark, former policy director at research laboratory OpenAI policy director and co-chair of tracking initiative AI Index, as a member of the OECD Task Force.  

Clark is quoted as saying: “Think of it this way — if no one measured resources like electricity or oil, it’d be difficult to build national and international policy around these things.

“Compute is one of the key inputs to the production of AI, so if we can measure how much compute exists within a country or set of countries, we can quantify one of the factors for the AI capacity of that country.”

Karine Perset, the administrator of the OECD AI Policy Observatory, told the website: “There’s nothing that helps our member countries assess what they need and what they have, and so some of them are making large but not necessarily well-informed investments.”

She also told Venturebeat that the task force intends to develop an initial framework by autumn 2021, then begin gathering a data gathering exercise.

This will benchmark domestic AI computing supply by country and region, and also look at countries’ budgets and investment priorities.

However, the ONE AI webpage acknowledges that the task force must be “mindful that the AI computer landscape is unusually dynamic with technical shifts on a frequent basis”.

One issue is that national governments might enter into partnerships with private sector cloud services, but these arrangements might not prove lasting: the task force will have to decide whether to count these when calculating national computational infrastructure.

Clouds in the desert

An example cited by Venturebeat is the deal signed in December 2020 by state-owned Saudi Arabian oil company Saudi Aramco and Google Cloud. The new infrastructure is expected to accelerate Saudi Arabia’s tech industry.

The OECD was founded in 1961 to stimulate economic progress and world trade. The Paris-based intergovernmental economic group currently has 37 member countries

ONE AI is running a number of projects to guide national governments as they invest in AI solutions. These include writing a “user-friendly framework to classify AI systems” according to their impact on public policy in areas.

The ONE AI webpage names these as economic and social benefits; human rights, privacy and fairness; safety, security, and risk assessment; transparency; accountability; research; data, compute and technologies; labour and skills; and international cooperation. Another working group is to produce guidance and standards to ensure that member states’ AI systems are “trustworthy”, which is defined as being inclusive and of benefit to both people and the planet, while also being unbiased and transparent.

About Elaine Knutt

Leave a Reply

Your email address will not be published. Required fields are marked *