Researchers have been looking at how to develop a computer processor that’s inspired by the human brain. Right now, most of the computers people use, even the most powerful ones, have a limitation. They experience something called the Von Neumann bottleneck. Imagine someone is playing with building blocks.
But, they have to constantly run back and forth to get the blocks needed. In computers, data has to be frequently moved between the memory, where data is stored, and the processor, which does the calculations. This whole back-and-forth is said to slow things down and make computers less efficient. Could a processor be more energy-efficient? IBM, which stands for International Business Machines, is an American tech company. In October 2023, it was said that they had come up with something called the NorthPole processor.
This processor is said to be designed like the human brain and could do AI tasks faster. It’s as if one’s thinking and memory were combined into a single super-efficient process. NorthPole is said to be more than 20 times faster and up to 25 times more energy-efficient than most of the chips available today for AI tasks. This means it can perform AI jobs much faster while using way less power.
It’s like a car could go 20 times faster while using less fuel. So, all of this sounds great theoretically. But, what does it mean for the real world? The NorthPole chip could be used for understanding images, videos and speech. It could be beneficial to self-driving cars, robots, digital assistants and more. Envision a self-driving car that can quickly identify objects in its path or a robot that can navigate complex environments efficiently. One of NorthPole’s advantages is said to be not needing fancy and energy-consuming cooling systems.
Simple fans and heat sinks could keep it running. So, this could address the growing demand for energy-efficient AI. AI is said to consume a lot of energy, which could be bad for the environment. It’s said that in some ways, generative AI could be unsustainable.