AI's Progress: Can Neuromorphic Computing Help?

Saving Energy While Making AI Smarter: The Neuromorphic Computing Idea

Artificial Intelligence (AI), which means using computers to think like humans, has gotten much better recently. For instance, programs like ChatGPT can understand and use language, and computers can “see” things better too. This helps different businesses and society as a whole. But, these improvements also use up a lot of energy, which can be a problem.

A study says that if we keep using more and more data, the energy AI needs might be more than the world can make by 2040. This shows that we have to think about how AI affects the environment as it gets even better.

One problem with AI is how computers are built, like the Von Neumann architecture. This way of building computers separates their “thinking” part from their “memory” part. This makes them slow to talk to each other and stops them from working together really well. This is why they use so much energy and make pollution.

To fix this, there’s a new idea called Neuromorphic Computing (NC). This idea looks at how the human brain works, which is very energy-efficient. The brain only needs a little bit of power, but AI computers need much more. NC tries to copy how the brain works by making the “memory” and “thinking” parts closer together. They use a special kind of technology that’s like the brain’s connections, called memristors. These help with memory by changing how they work.

Big tech companies like IBM, Intel, and BrainChip are making new ways of doing this, like TrueNorth and Loihi. These ways use less energy than the usual ones.

But, there are problems with using NC. We need to find better computer rules that work with this new way of building computers. Also, we have to test these new ways with big and complicated sets of data. We need to do more research and work on this idea to make it really good for everyday things.

To sum up, as AI gets better, it needs more and more energy. Neuromorphic Computing is a way to fix this by copying how the brain works, which uses very little energy. This could change how AI works, make it use less energy, and help the environment.