AI solutions look to create a sustainable future

AI solutions look to create a sustainable future

When we talk about unleashing the positive potential of artificial intelligence, sustainability is one of the first things that comes to mind.

Whether it’s streamlining industrial operations and unlocking network efficiency to reduce energy costs or preserving environmental resources to create a more sustainable society, AI offers hope at a time of ecological and climate crisis.

For this to come to fruition, though, we must think sustainability first. That requires changing our mindset and innovating AI technologies to specifically reduce the energy cost and environmental footprint of the technology from the start and not just in retrospect. This is particularly relevant in a telecom industry that consumes a lot of energy.

The information and communication technologies sector already accounts for 2%-3% of global energy consumption and 4%-6% of global green-house gas emissions. These figures are only expected to grow as communications traffic volume increases in an ever more digitized world. Therefore, telecom operators are committed to eventually becoming carbon neutral.

As a company, Nokia is deeply committed to sustainability. In addition to limiting our own footprint, we are devoted to maximizing our handprint in digitalization and connectivity. That means creating technology solutions that make industries more efficient, help minimize waste and enable greater reuse of precious resources and materials.

This commitment is clearly demonstrated in our Nokia Bell Labs approach to AI, as sustainability is one of our six fundamental pillars of responsible AI.

A Nokia Bell Labs original solution

We have struck a fundamental trade-off between network performance and energy consumption by targeting one of the most energy-hungry parts of the radio network: the base station.

We have developed a cutting-edge solution that figures out at any given time the minimal number of frequencies that a base station requires to maintain the mobile users’ quality of connection. In other words, our AI algorithm targets which layers can be switched off without adversely affecting performance.

This method doesn’t require any expensive changes to the electronic hardware that makes up base stations. Instead, it can be deployed through a straightforward software update on existing infrastructure.

A base station consists of different sets of antennas, each transmitting and receiving electromagnetic waves that oscillate at different frequencies. These effectively act as the “entrance door” for mobile phones to find the right frequency to connect to the network.

Think of frequencies as being akin to highway lanes, and mobile data being the vehicles that ride them. Just as more lanes make the traffic move more fluidly, more frequencies improve the mobile user experience. So, the simple solution would be the keep open as many frequencies as possible.

The problem is that more frequencies involve more antennas and therefore more electronic circuitry that translates into higher energy consumption. Finding the right balance between network performance and energy consumption in no easy task and calculating the optimal number of frequencies to be employed has thus far required a lot of trial and error.

Using machine learning techniques, our AI base station solution solves a hard math problem that humans simply could not achieve on their own. We tested this solution on live Nokia networks, with real user traffic on 20 sites, and the result has been a proven reduction in energy consumption by 10-15% without any noticeable degradation in the quality of service for connected users.

If deployed in a typical European nation, this could enable Gigawatt-Hour scale energy savings. That’s roughly the equivalent of the energy it takes to power a million low-consumption light bulbs for a year, at a cost of millions of Euros.

Deploying such networks would have an even greater impact in the developing world, where poorer nations often struggle to keep up with the pace of technological innovation. In sub-Saharan African countries, for instance, 4G is expected to make up most of the connections until at least 2028 since they can’t afford to replace their current hardware with a more energy-efficient one. Our solution is a great fit since it doesn’t require a hardware update and can sit atop any legacy infrastructure. Therefore, developing improved network sustainability software solutions is key to making green solutions accessible globally and for reaching carbon neutrality globally in the telecom sector.

There is obviously also a financial component to this push toward sustainability as operators are desperately seeking new ways to reduce electricity bills that are eating into their earnings due to skyrocketing energy prices, especially in Europe. Indeed, energy consumption already accounted for between 15% to 40% of the operating costs of telecom operators in early 2022, with projections that it would increase dramatically due to the fallout of the Russia-Ukraine war and other geo-political events.

Consequently, we are witnessing a paradigm shift in telecom away from “maximum performance,” where connectivity speed was prioritized more than anything else, to “sustainable performance,” where a balance between speed and energy consumption is most sought.

AI for good

The recent innovation burst in generative AI chatbots has sparked controversy and doomsday projections about how these powerful systems could harm us. But in sustainability, we see one of AI’s greatest upsides without some of the pitfalls that have dominated the news lately.

At Bell Labs, we are committed to using AI to meet the needs of individuals, societies and our planet. That’s why we focus on how it can build communities and serve humanity to make a better world.

We can do so by pursuing principles such as fairness and privacy and solving conflicts between morality and technology – and preserving these principles through the AI design process.

These will all be key elements in reaching our sustainability goals in the 6G era.

Naturally, there are also AI concerns from a sustainability standpoint. Training AI systems consumes enormous amounts of energy, so there is a need to create and train these systems while minimizing their environmental impact. In the pursuit of AI, we must make sure we are not endangering the sustainability of our planet in the process.

Our base station solution though, doesn’t include a separate high-energy training phase as part of its deployment. It’s a powerful example of how AI can be an agent for good and spark our hopes, rather than our fears.