The AI rollercoaster of expectations and concerns continues to twist at breakneck speeds as enterprises inch ever closer to understanding the rapidly changing technology and its possible functions within their business. Most recently, advanced artificial intelligence platforms such as generative AI and large language models (LLMs) have fallen under scrutiny for their voracious energy consumption and consequent ecological impact, with some researchers hypothesising that LLMs consume hundreds of litres of fresh water and produce annual emissions equivalent to that of a small country (according to Pengfei Li, Jianyi Yang, Mohammad A. Islam, and Shaolei Ren).
With the recent WEF meeting in Davos, global stakeholders are questioning where the bulk of responsibility should lie for limiting warming to 1.5 degrees by 2025. Climate change remains an issue of critical importance for consumers and companies alike amidst these global efforts to reduce CO2 emissions, boding poorly for the public image of any company that uses consumptive AI tools without keeping its carbon footprint in check. More importantly, rampant unchecked AI use could have disastrous consequences for the environment – research from MIT suggests that training just one AI model produces almost five times the lifetime emissions of an average American car, which has the potential to significantly counteract global progress in combatting climate change.
Despite the apparent ecological apathy of recent legislation like the EU AI Act and President Biden’s executive order, which focus largely on other facets of AI responsibility, some major AI players have begun to proactively self-regulate and work towards sustainable AI use. Here are ways in which the leaders in artificial intelligence are approaching AI with ecological consciousness, while preserving the profound business value of the technology.
Opting for purpose-built AI
Many drawbacks of generative AI and LLMs stem from the massive stores of data that must be navigated to yield value. Not only does this raise risks in the way of ethics, accuracy, and privacy, but it grossly exacerbates the amount of energy required to use the tools.
Instead of highly general AI tools, enterprises have begun to pivot to narrower, purpose-built AI, specialised for specific tasks and goals. For example, ABBYY has adopted this approach by training its machine learning and natural language processing models to specifically read and understand documents that run through enterprise systems just like a human. With pre-trained AI skills to process highly specific document types with 95% accuracy, organisations can save trees by eliminating the use of paper while also reducing the amount of carbon emitted through cumbersome document processes.
The role of developers
AI companies don’t need to shoulder the burden of sustainable AI all on their own – some are proactively putting the proverbial ball in the court of developers.
OpenAI, the artificial intelligence pioneer responsible for the widely popular ChatGPT, recently announced that developers can create their own “GPT” platforms for specialised purposes. This allows developers and organisations to narrow their AI use with a high degree of customisability, trimming away excessive features and data that amplify ecological damage. For example, developers could design GPTs for purposes limited to creative writing advice, cooking information, tech support, or any other niche purpose.
Considering the increased risks for inaccuracy and privacy infringement associated with highly general AI models, developers will likely be motivated to take advantage of these narrower, more specialised GPT platforms not just for ecological responsibility, but for improved business outcomes as well.
Implementing sustainable business practices
Companies should also take a step back from the technology itself and look inside their organisation for more ways to sustainably leverage AI. For example, Microsoft revealed that their AI-supporting hardware runs exclusively on clean energy, absolving them of creating so-called “operational emissions.”
Moreover, companies can use AI as a tool to explore other facets of their business in which sustainability could be prioritised. Forrester highlights the measurement, reporting, and data visualisation capabilities of artificial intelligence to suggest that it could power a climate revolution of its own.
Although objectively important, emissions aren’t the only metric used to encompass ecological impacts – studies have shown that a combination of robotics and AI has reduced herbicide use in some contexts by 90%. As companies continue to grapple with the utility and consequences of AI, they must explore the full breadth of its capability to enhance and contribute to sustainability.
Staying on the right track
So far, early AI legislation has largely failed to reign in the ecological implications of artificial intelligence, focusing instead on privacy and other ethical areas. While these areas are also crucial for responsible AI use, enterprises must keep themselves accountable for how they leverage artificial intelligence to generate business value.
2023 may have been a year of hype, noise, expectations, and misconceptions surrounding artificial intelligence, but the maturity that enterprises have accrued over the past year has given them the means necessary to make informed and responsible decisions regarding AI use. Still, it’s wise to scrutinise, question, and hold large organisations accountable for their carbon footprint and other impacts on the environment – those who prioritise ecological responsibility should have nothing to hide.
Maxime Vermier
Maxime Vermeer is the Senior Director of AI strategy at ABBYY, a global leader in intelligent document processing and digital transformation solutions. With over 10 years of experience in the AI field, Maxime leads the creation and execution of the go-to-market strategy for ABBYY’s cutting-edge AI technology, delivering value to customers and stakeholders across various industries.
