Training AI is carbon-intensive. MIT researchers are changing that

The different ways artificial intelligence (AI) can be used to improve our lives is vast, but one thing we often overlook is the environmental cost that comes with AI. After all, running a training model to improve AI requires a whole lot of energy. For that reason, researchers at Massachusetts Institute of Technology (MIT) have developed a solution that not only lowers costs but, more importantly, reduces the AI model training’s carbon footprint.

Back in June 2019, the University of Massachusetts at Amherst revealed that the amount of energy utilized in AI model training equaled 626,000 pounds of carbon dioxide. How so?

Contemporary AI isn’t just run on a personal laptop or simple server. Rather, deep neural networks are deployed on diverse arrays of specialized hardware platforms. The level of energy consumption required to power such AI technologies is approximately five times the lifetime carbon emissions from an average American car, including its manufacturing. That means that if we want to continue with AI, we must create more sustainable ways for AI to engage in machine learning.

That’s where MIT’s research team comes in. What they have created is a groundbreaking automated AI system, termed a once-for-all (OFA) network, described in their paper here. This AI system — the OFA network — minimizes energy consumption by “decoupling training and search, to reduce the cost.” Essentially, the OFA network functions as a ‘mother’ network to numerous subnetworks. As the ‘mother’ network, it feeds its knowledge and past experiences to all the subnetworks, training them to operate independently without the need for further retraining.

This is unlike previous AI technology that had to “repeat the network design process and retrain the designed network from scratch for each case”, which led to excessive energy consumption. In other words, with the OFA network in use, there is little need for additional retraining of subnetworks. This efficiency decreases costs, curtails carbon emissions and improves sustainability.

For all of us, this story serves as a reminder that with any great technology comes an environmental cost that we don’t always perceive. Fortunately, in AI’s case, researchers are already working to make it more sustainable.

Solution News Source

Training AI is carbon-intensive. MIT researchers are changing that

The different ways artificial intelligence (AI) can be used to improve our lives is vast, but one thing we often overlook is the environmental cost that comes with AI. After all, running a training model to improve AI requires a whole lot of energy. For that reason, researchers at Massachusetts Institute of Technology (MIT) have developed a solution that not only lowers costs but, more importantly, reduces the AI model training’s carbon footprint.

Back in June 2019, the University of Massachusetts at Amherst revealed that the amount of energy utilized in AI model training equaled 626,000 pounds of carbon dioxide. How so?

Contemporary AI isn’t just run on a personal laptop or simple server. Rather, deep neural networks are deployed on diverse arrays of specialized hardware platforms. The level of energy consumption required to power such AI technologies is approximately five times the lifetime carbon emissions from an average American car, including its manufacturing. That means that if we want to continue with AI, we must create more sustainable ways for AI to engage in machine learning.

That’s where MIT’s research team comes in. What they have created is a groundbreaking automated AI system, termed a once-for-all (OFA) network, described in their paper here. This AI system — the OFA network — minimizes energy consumption by “decoupling training and search, to reduce the cost.” Essentially, the OFA network functions as a ‘mother’ network to numerous subnetworks. As the ‘mother’ network, it feeds its knowledge and past experiences to all the subnetworks, training them to operate independently without the need for further retraining.

This is unlike previous AI technology that had to “repeat the network design process and retrain the designed network from scratch for each case”, which led to excessive energy consumption. In other words, with the OFA network in use, there is little need for additional retraining of subnetworks. This efficiency decreases costs, curtails carbon emissions and improves sustainability.

For all of us, this story serves as a reminder that with any great technology comes an environmental cost that we don’t always perceive. Fortunately, in AI’s case, researchers are already working to make it more sustainable.

Solution News Source

SIGN UP

TO GET A Free DAILY DOSE OF OPTIMISM


We respect your privacy and take protecting it seriously. Privacy Policy