How the world can tackle the power demands of artificial intelligence

The world must reshape its technology infrastructure to ensure artificial intelligence makes good on its potential as a transformative moment in digital innovation.

New technologies, such as generative AI, need huge amounts of processing power that will put electricity grids under tremendous stress and raise sustainability questions.

But pioneering technologists are working on a potential game changer that goes some way to address these issues: photonics.

John Gallant, CIO.com’s Enterprise Consulting Director and Vito Mabrucco, NTT Corp. Chief Marketing Officer, recently engaged in an extensive discussion on exactly how photonics technology could help meet the power demands of AI.

Mabrucco first explained that AI will put exponentially higher demands on networks to move large data sets.

He said: “We know the current infrastructure that we have can’t possibly support all of the new innovations that are going to result from the very wide and broad and deep implementation of AI.”

The demands of AI call for an entirely new approach – a paradigm shift that replaces electronics-based computing and networking with photonics-enabled computing and networking.

Photonics addresses a variety of issues and concerns: first, because photonics uses less energy than electronics, it can reduce the amount of power needed to do the same amount of work. This is vitally important because there are legitimate concerns that AI will outpace the capacity of the power grid.

Photonics technology also delivers exponentially higher bandwidth rates with lower latency. This enables use cases such as near real-time disaster recovery over photonics-based links in industries like banking and finance, vehicle-to-vehicle communication in an autonomous vehicle scenario, and real-time edge-to-data center connections for robotics applications in factories, or at remote sites in mining or oil and gas industries.

How does it work?

Mabrucco explained that NTT is working to take fiber-optic technology, which has been used for decades to transmit data over long distances (think undersea cables) and shrink it down for deployment inside computers and networking gear, even down to the chip level.

He also says that photonics can “change the paradigm of computing” by enabling a disaggregation of the traditional computing stack.

With photonics-based interconnects, organizations will be able to create efficient pools of processing units for specific use cases, such as large language model (LLM) data processing in one location, data storage in another location, and a high-speed link between the two.  

NTT takes leadership role

NTT is taking a leadership role on a variety of fronts, Mabrucco said. NTT created, alongside Sony and Intel, the IOWN Global Forum. Over 150 leading organizations are involved in it, with the aim of achieving early and successful uses cases that can then be scaled.

Mabrucco also encouraged CIOs to get involved in the IOWN Global Forum.

NTT has its own internal set of principles that guide its approach to AI. NTT believes that AI should respect human rights and diversity; that it be fair, unbiased and transparent; that it protects personal data; that it be secure; and that it will not only create new business opportunities, but also benefit people and the planet.

Speaking of NTT’s AI Charter, Mabrucco said NTT was looking to take a leadership role in AI governance and ethics. It’s clear that the technology firms must understand how this technology will impact if it is to deliver on the promise of a secure and trustworthy AI.

View the entire discussion here: