If generative AI demonstrates its deeply transformative nature day after day, it is also proving to be particularly resource and energy intensive. To meet the major challenge of the environmental crisis, it is essential to be selective in our projects and to direct our use of generative AI towards positive impact solutions.
We must rethink our approach to progress by refocusing technological innovation around solutions that serve the general interest and contribute to building a more sustainable and inclusive future. This means making a distinction between complacent AI and impactful AI[MJ1] ; between AI-powered solutions that only serve to entertain or consume, and those that positively impact our world and our daily lives. Today, AIs, whether generative or not, can help a household save energy, choose more environmentally friendly products when shopping, or even adopt a more environmentally friendly diet.
And for an AI to have a positive impact, there is still work to be done. Let us be reminded what too few people are aware of – AIs are not intangible[MJ2] ! Behind a Gemini, you have equipment requiring a lot of precious, rare, and polluting materials to produce. Training an LLM model consumes as much energy as several hundred French households. A ChatGPT query uses about ½ litre of water. Current projections in 2027 are equivalent to Belgium's annual electricity consumption and France's water consumption.
Given these troubling numbers we need to consider not only the financial benefits of an AI, but also how it affects society and the environment.
Towards a new paradigm
The good news is that the ground is ripe for the establishment of this new paradigm. According to an IPSOS study done for Sopra Steria, more than half of the respondents (58%) think that artificial intelligence can make a significant contribution to combating global warming. In addition, more than 6 out of 10 people say they are ready to integrate AI-powered tools into their daily lives to better control their impact on the environment. These are important conclusions which show the need to change the way of designing and especially evaluating digital solutions.
As more and more organizations and individuals incorporate extra-financial criteria into their performance evaluation, and as the European population is [MJ3] clearly ready to welcome AI with a positive environmental and social impact, it is our duty to think of new tools to be able to evaluate the cost/benefit ratio of an AI in terms of impact.
We must complement the purely financial vision of return-on-investment criteria in order to systematically add an impactful ROI. And it is our responsibility; to us, tech companies; to us, users of AI; and to us, citizens to limit our use of AI to projects whose impactful ROI is positive.