The activity of the world data centers consume up to 1.5% of world total electricity, a consumption that could reach triple in 2028, according to estimates of the United States Department of Energy, a country in which, according to forecasts … of Lawrence Berkeley National Laboratory, Its ‘IA’ servers could exceed 10% of electricity consumption Total of the US in the next three years.
In this context, in which AI exponentially increases energy needs, programming faces the challenge of ensuring a more sustainable present and future. And thus it stands out in a recent publication of the IA of Ametic Observatory (‘Green Code when developing code’), conducted by three students of the technical masters of Tajamar Tech (cloud infrastructure and development, in addition to cybersecurity) to reflect on how Apply the ‘Green Code’ philosophy in software development.
«Software development (comment) plays a key role and could be responsible for up to 14% of total CO2 emissions by 2040. Each line of code and active server contributes to the global problem of climate change. However, if only 3% of developers and technicians adopt ecological software practices, a notable reduction in the carbon footprint of the technological sector would be achieved. And they point out how Chatgpt’s current models can evaporate up to 700,000 cubic liters of water and generate 502 tons of CO2 (globally, the demand for AI could represent up to 6,600 million cubic meters of water extraction by 2027).
Although the technological ‘majors’ have already gone to work in this inescapable need for sustainable development, the Green Code supposes the opportunity to progressively improve in more efficient codes, optimized algorithms and an intelligent use of hardware and software resources. «The objective of this initiative with Ametic (comments José Antonio Ureta, director of the artificial intelligence master of Tajamar Tech) is to contribute to the awareness, by the new generations, of sustainable development, to produce clean and consume less, a whole responsibility in the mass consumption of the new ones ».
Applied research
In Imdea software, Research Institute belonging to the IMDEA initiative of the Community of Madrid, they know well the performance of facing this challenge, for example with projects such as Tazos with Nomadic Labs; the N-Greens program funded by the Community of Madrid; the traces and procode projects of the National Plan; Or the project enters the European Union, all in collaboration with numerous European universities and companies. «Energy consumption (Manuel Hermegildo and Pedro López, IMDEA SOFTWARE, Polytechnic University of Madrid and CSIC, have recently been recently exacerbated with the emergence of generative artificial intelligence: the training of one of these systems can generate thousands of tons of CO2 and it is estimated that, if the energy growth of artificial intelligence continues to the current step, it could in a short time consume as much as a medium -sized country ».
Hermenegildo and López add another variable in this environment: “Billion internet devices, including small and everyday devices, from phones to implantable medical devices, which work with batteries and where concern is autonomy.” And put as examples of significant energy savings in software those of “relatively small changes in the ‘kernel’ (software that serves as a bridge between the hardware and the rest of the operating system functions) of Linux, which are resulting in reductions up to 30% in the consumption of large data centers ».
Less carbon
To move towards a more sustainable scenario, researchers such as the consulted develop advanced techniques for analysis and optimization of programs that, combined with precise models of energy consumption in the hardware, allow to build advanced tools for the development of energetically efficient programs.
«These tools (the specialist stands out) identify ‘Resource consumption errors’ in the programs in compilation time (that is, during the program development process) or automatically certify that the program consumptions are within the limits admissible ». Actions that allow simplifying, lowering and shortening the development of ‘green programs’ and, therefore, of devices that make a “certifiable more efficient resources of the available resources: energy, and also execution time, memory, disc, space, etc. ».
In the case of Keep Coding, programming training center, with all kinds of options and resources (and levels), Adriana Botelho, CEO and co -founder, points out how “already in 2019, a study revealed that training a model of language processing model Natural can issue up to 284 tons of CO2, equivalent to life emissions of five average cars in the United States. Therefore, Green Code is not only an emerging trend, but a prevailing need, an ethical and environmental obligation. Writing efficient code, minimizing superfluous processes and designing sustainable architectures at all levels are fundamental actions to reduce the digital carbon footprint ».
Social impact
«In Keepcoding (continues Bottleo), we are aware of this reality and, for years, we have incorporated ethics and Green Code modules in our formations. We believe that future programmers should not only be technically competent, but also to understand the social and environmental impact of their work and how to reduce it. Although the software is intangible, its impact on the physical world is undeniable ». Evidence on which Fernando Rodríguez, Clo and Cofundador introduces a concept to take into account: “Green software requires being approached from two different points of view and in two different layers: the underlying software and hardware.”
«Ultimately (points) it is the hardware that consumes resources. From the point of view of the hardware, huge advances have been made in recent years, due to the almost absolute prevalence of portable systems (‘laptops’, ‘tablets’ and telephones) that require moderate use of the battery to succeed in the market. This has caused a gradual change, almost absolute today, towards more efficient platforms such as ARM processors, which have displaced Intel in almost all markets, from mobile phones to large servers in the cloud. And from the point of view of software engineering, «the objective of low consumption and greater efficiency are achieved with good engineering practices, to use the best algorithms and data structures. It should be the usual practice and, nevertheless, it is often forgotten ».
#Green #software #sustainable #code #digital #avalanche
