Several options can be considered to reduce the consumption of the required environmental resources. First, renewable energy sources, such as solar and wind energy, can be used to reduce the carbon footprint of AI systems.

Likewise, resource optimization techniques, such as model compression, data precision reduction, and the development of “Green Algorithms,” can be used to reduce energy and memory consumption.

Model compression is a set of techniques that aim to reduce the memory footprint and computational requirements of AI models. These techniques can be broadly classified into two categories: model pruning and model quantization.

• Model pruning involves removing redundant or unimportant neural connections in a neural network, reducing the size of the model and improving its efficiency.

• Quantization, on the other hand, involves reducing the precision of the data used in a model, which reduces the amount of memory and power needed to store and process the data.

So by reducing the memory footprint and computational requirements of AI models, you can reduce the amount of energy and resources needed to train and run the models. Furthermore, model compression can also improve the efficiency and performance of AI models, which can have a positive impact on the accessibility and usefulness of AI technology.

Data precision reduction is a technique used to reduce the number of bits required to represent the data used in an AI model. This technique is often used in AI systems that require a large amount of memory and power to store and process data, such as image and video processing AI systems.

This technique reduces the amount of memory and power needed to store and process data. For example, if you use 16-bit precision instead of 32-bit precision to represent a number, you can reduce the amount of memory required to store the number by half.

Reducing data precision is another possible alternative to reducing the consumption of environmental resources required by AI systems. By reducing the amount of memory and energy required to store and process the data used in AI models, you can reduce the amount of energy and resources required to train and run the models.

For their part, “Green Algorithms” constitute another technique that can be used to reduce the amount of energy and resources necessary for the development of Artificial Intelligence (AI). These algorithms are designed with the aim of using AI that is more inclusive and respectful of the planet. The design of these algorithms, in addition to an environmental necessity, is already considered a requirement for compliance with the Sustainable Development Goals set by the 2030 Agenda.

“Green Algorithms” seek to achieve less polluting AI, capable of reducing the carbon footprint with algorithms that consume less energy and are more efficient. To achieve green algorithms, an environmental commitment and greater economic resources are needed. This is something that not all laboratories or technology centers can assume. To have efficient models, many experiments are needed, and this entails an energy and economic cost that not everyone can afford.

In Spain, the National Green Algorithms Plan establishes a set of voluntary good practices for the certification of sustainable hardware and software services companies. The plan also establishes standards and tools to measure the energy consumption of algorithms, in order to raise awareness among AI developers about the environmental impact of their decisions.

Voluntary good practices for the certification of sustainable hardware and software services companies include, but are not limited to, the following:

• Energy efficiency: Companies can improve the energy efficiency of their AI systems by using more efficient hardware and software, as well as by implementing energy optimization techniques, such as model compression and reducing the precision of data.

• Use of renewable energy: Companies can reduce their carbon footprint by using renewable energy sources, such as solar and wind energy, to power their AI systems.

• Recycling and reuse: Companies can reduce their environmental impact by recycling and reusing hardware and software components.

• Sustainable design: Companies can design their AI systems to be more sustainable, for example by using recycled materials and reducing the size and weight of devices.

• Transparency and accountability: Companies can improve the transparency and accountability of their AI systems by implementing data governance practices and adopting ethical and privacy standards.