4 Ways To Make AI Greener And More Sustainable

Artificial Intelligence (AI) has long been a technology of speculation and anticipation, owing to its potential to transform industry and the relationship between people and technology. AI is growing faster than ever, surging through the inundation of data, evolutions in computing power, and a feverous need for business to generate new insights and make better decisions. In fact, AI has infiltrated many aspects of modern society: from voice assistant like Siri to self-driving cars, and facial recognition on mobiles to software that could challenge professional board game players. While AI is transformative and will play a key role in society’s future, the process of creating AI is a resource intensive practice that can have negative environmental impacts. The implication? Developing AI must be pursued in a sustainable way if we hope to reap the benefits without creating new problems. In this article, we explore three ways that AI could be made more environmentally friendly: production of hardware, design of algorithms and the use of greener electricity.

Algorithmic innovation:

When it comes to making AI greener, algorithmic innovations should be considered a priority. Algorithms tell software how to make sense of text, visual, and audio data so that they can, in turn, draw inferences from it. Algorithms are like sports tactics, and software and hard wares are like the players and their equipment. The more efficient the algorithm, the less work the software has to do, and less computing power should be needed.

Some people might doubt the necessity of optimizing the algorithm and think the whole point of having a computer is to let it do as much of the heavy lifting as possible to save human’s time and effort. Thus, it has become common practice to come up with a simple and straightforward algorithm without regarding the efficiency (or implications for scalability) and let computers do the rest. If the problem is easy and only requires limited computer resource, then this approach might not be problematic. But AI related projects often involve huge amount of data and very complicated algorithms. Inefficient algorithms would result in a waste of time, resources and computing power (energy). However, if the goal is to pursue AI sustainably – and outsmart the competition - it is important to recognize this trade off and think of the optimal strategy overall, rather than outsourcing critical and creative thinking to computers.

Efficient Hardware:

AI projects involve a huge amount of calculation, and especially during the model training stage. One should always consider getting the appropriate hardware like power supply and CPU/GPU. Choosing a better GPU not only boost performance, but also lowers energy consumption, which is what all companies dealing with AI should think about.

More importantly, utilizing all the available resources is another great way to make AI more efficient. Nowadays, most AI projects can and are being built on cloud service provided by the big three (AWS from Amazon, GCP from Google and Azure from Microsoft). One could always choose to carry out an AI plan on such platforms. They are fast, scalable, cost-effective and definitely more efficient in every aspect compared to PCs. If it is confusing why these platforms are so much better, think of you iPhone’s Siri. Siri is a type of AI voice assistant that “talks” to you and executes commands. Have you noticed that Siri only works with Internet connection? That’s because every time Siri receives a command, it sends that command to Apple’s cloud server, works out the solution and sends it back to your phone. This process is so quick that it appears Siri “answers” you instantly. If Apple tried to put such AI model and algorithm in one’s phone and have Siri work offline, then a single commend like “call Tom” may exhaust the phone battery before the action was taken.

Green Energy:

As for greener energy, it is intuitive how it could make AI more environmentally friendly by releasing less carbon. There are five main types of renewal energy technologies:

• Onshore and offshore wind power

• Water (hydro power, wave power and tidal energy)
• Solar energy
• Geothermal energy
• Biofuels

Companies should consider switching for greener energy. However, one problem is that people aren’t always aware of the options. Some utilities don’t always do a great job of advertising their eco-friendly options. One simple option for companies dealing with AI is to ask their energy supplier for renewable energy plan options, which nowadays most suppliers have.

Invest in Environment:

Although scientific advances could make a huge difference, sustainability should not become dependent on the idea that technology will be our only savior. Everyday citizens are not helpless simply because they do not understand how this technology works, rather it is concerned, proactive and vocal citizens that are able to drive change. Such initiatives are already being made possible by organisations like EARTHLY. Both companies and individuals can join them to invest in the protection of environment. For example, companies could ask them to calculate their carbon impacts using latest industry standards. EARTHLY would then come up with a tailored proposal for your carbon mitigation or reduction strategy. Companies would be able to see and track their impact on environment on EARTHLY’s platform. Also, individuals can help by investing in forest preserving and expansion in Kasigau Corridor.

Final Word:

AI’s wide-range applicability and its great potential makes it even more important to consider a sustainable route of development. AI will no doubt play a huge part in human being’s future, and companies regardless of whether they’re just starting their AI journey or in the cutting edge should consider how they can pursue AI in a way that makes AI green and sustainable. Although it’s easy to think sustainability is all about the technology and hardware, human mindset and behaviour is the defining factor that will make sustainable AI a reality.

Written by: Clayton Black and Alex Shen