The impending threat of large-scale computation to the environment often goes unnoticed in our software-orientated world. Shivam Kajale, a PhD student in the Nano-Cybernetic Biotrek research group at the MIT Media Lab, has taken up the mission to bring awareness regarding environmental impact of artificial intelligence (AI)-related computation. In line with this mission, he recently delivered a talk at the TEDx Boston’s Planetary Stewardship program, highlighting the growing threat of large-scale computation to the environment and urged our community to get into a race for developing environmentally sustainable alternatives for computing hardware.

We are all witnessing the meteoric growth of computation across the world fueled by AI and internet of things (IoT). While we are still expanding the application of AI and in awe of what all is possible with it, we are failing to see the growing threat that large scale computation poses to the environment. For example, data centers across the world, which are gigantic colonies of computer we commonly call the “cloud," are already consuming electricity at par with major nations like South Africa and the United Kingdom. Computation is the fastest growing consumer of electricity worldwide and is estimated to account for about 30% of the electricity consumption in the world by 2030. Electricity, 60% of which is still coming from fossil fuels. Moreover, by the year 2040, computers of the world are estimated to require 1e27J of energy to keep up the current rate of growth in computation—energy that is far greater than what humanity may be able to generate by then. Yet, we cannot just abandon computation altogether. What we need are “greener chips," i.e., computer hardware that is exceedingly energy efficient and allows us to reap the benefits of AI and IoT without harming our planet. 

Through his research in Prof. Deblina Sarkar's group at the MIT Media Lab, Shivam is developing “beyond-CMOS” electronic devices using novel materials and physical mechanisms which have been estimated to be over 10,000 times energy efficient as compared to the traditional silicon-transistor based computers.

His approach is to take inspiration from our brain, which is arguably the most energy-efficient computer in nature. After all, it enables human intelligence with just 20W of power. And while we are far from completely decoding our brain, what we do know is that the brain operates using an intricate network of neurons and synapses. Each neuron is a computing entity and is directly linked to thousands of synapses which are the brain’s memory unit. This is unlike traditional computers where processor and memories are physically separated and just the back-and-forth movement of data far exceeds the energy needed for computation alone. And so, to mimic the brain’s strategy for low-power computation, Shivam is making use of two-dimensional magnetic materials to create energy efficient processors and memories for computers.

Shivam’s work is trying to bring energy-efficiency to the most fundamental building blocks of our computers, and hence its applications can be far reaching. These devices can replace the silicon-transistors in all computing devices from PCs and smartphones to data centers and wearables, facilitating massive reductions in energy consumption to allow an environmentally sustainable growth of computing ecosystems.