With the rise of AI technologies, there is an increased demand for energy it requires. The new computer chips can enhance energy efficiency by six times. It can also align with computation and data storage. It will also reduce the ecological footprint of artificial technologies.
To reduce the amount of energy consumed by AI, researchers are working towards it. Assistant Prof. Sieun Chae from Oregon State University is researching chips that allow for both computation and data storage. These chips are based on a novel material platform. The system of the chips will mimic the way biological neural networks handle the storage and processing of information.
It is estimated that artificial intelligence will account for around half a percent of the total global energy consumption by 2027. It is equal to the amount that the entire Netherlands uses annually. Researchers at the University of Michigan collaborated with researchers from the University of Oklahoma, Cornell University and Pennsylvania State University.
It is also possible that AI and social media may worsen the climate crisis, say researchers.
Highlights
In the recently published report, she highlights the following:
- These chips feature components called memristors, short for memory resistors.
- Memristors used in this study feature a new material system known as entropy-stabilized oxides (ESOs).
- ESOs comprise more than half a dozen elements that allow fine-tuning of the memory capabilities.
- Memristors do not consume energy to move data to and from the chips.
- By optimizing ESO composition to work best for specific AI jobs, these chips can perform tasks with less energy in comparison to computer CPUs.
- ESO tuning can work on different time scales and would be able to process time-based information, like data for video or audio.
Even though these new computer chips can enhance energy efficiency by six times, this technology will be impactful only when applied on a large scale.
Source: New computer chips show promise for reducing energy footprint of artificial intelligence
