With the rapid development of artificial intelligence (AI) technologies in recent years, many have begun to wonder how these very technologies can help in solving one of the most serious threats that already looms over humanity - global climate change? A new article, written by some of the leading experts in artificial intelligence development and published on the arXiv.org online repository, attempts to answer this question by offering several examples of how machine learning will be able to prevent the decline of our civilization.
The proposed methods range from using AI and satellite technology to more effectively monitor deforestation, to developing new materials that can replace steel and cement (their production accounts for up to 9 percent of greenhouse gas emissions into the atmosphere). Despite this diversity, in their article, specialists repeatedly return to the broader possibilities of using such technologies. Especially against this background, the possibilities of using machine vision technology for environmental monitoring stand out; conducting large data analyzes to determine the inefficiency of industries with a high level of emissions of harmful substances into the atmosphere; and using AI to develop new, more efficient system models, such as our climate models,thanks to which we can better predict and prepare for future changes.
The authors of the article, including the British artificial intelligence researcher, founder and CEO of DeepMind, Demis Hassabi, Turing Prize laureate and one of the "fathers of deep learning" Yoshua Bengio, and co-founder of Google Brain, Google's research project on artificial intelligence Deep Learning - Andrew Ng says that AI can be “invaluable” in minimizing the worst impacts of global climate change, but adds that this technology is not a “silver bullet” - the only remedy for all problems. In their opinion, political forces should take an active part in this issue.
In total, the article discusses several areas at once in which machine learning technologies could find their application, categorized by the time frame of their possible use potential, explained by whether this technology is sufficiently developed. Below you can see this list.
Artificial intelligence will improve the efficiency of power supply systems
If humankind plans to rely on more renewable energy sources in the future, utilities will need ways to more efficiently predict and calculate the amount of energy we will actually need to use. Moreover, these calculations will have to occur in real time and during the entire period of operation of these enterprises.
Promotional video:
Algorithms have already been developed that can predict the demand for energy, but the efficiency of these algorithms can be further improved by introducing into the calculations such factors as the climatic features of certain regions, as well as the specifics of economic activity. Attempts to make the specifics of these algorithms more understandable will also allow utility operators to more accurately interpret the results of their analysis and use them in planning, choosing the most optimal time to launch these renewable energy sources.
Artificial intelligence will help discover new materials
Scientists need to develop new materials for more efficient production, storage and use of energy, however, as a rule, the process of discovering and developing new materials is very slow and not always successful. Machine learning technologies will speed up the process of finding, developing and improving new formulas with the desired properties.
Perhaps this will lead to the development of, for example, a new type of fuel, let us conditionally call it "solar", which will be able to store the energy of sunlight; will allow you to create a new and very efficient absorbent of carbon dioxide or building materials, the production of which will emit less carbon. Such materials may one day replace steel and concrete, the production of which releases almost 10 percent of the world's total greenhouse gas emissions.
Artificial intelligence will help to effectively reorganize the transport system
Delivery of goods around the world is a very complex and very often ineffective logistics process, in which goods of different volumes, weights and sizes interact, and different types of transport are used. At the same time, it is transport that accounts for a quarter of all CO2 emissions into the atmosphere.
Machine learning technologies used in this area will allow to more efficiently combine goods that require delivery to the same destination, which will reduce the number of required shipments. In addition, such a system will be more resilient to unforeseen disruptions in transport systems and will be able to manage huge fleets of unmanned trucks. However, the authors note that the latest technology is not yet ready at this point.
Artificial will lead to rapid adaptation of electric vehicles
Electric vehicles, which are a key element in decarbonizing vehicles, face a number of problems that prevent them from becoming truly mainstream.
Machine learning can help with this issue, the authors of the report say. For example, algorithms could improve the management of battery power consumption to increase the mileage of each charge and reduce the level of concern among potential buyers of such vehicles about limiting the range of travel. In addition, these technologies will optimize charging times.
Artificial intelligence optimizes building infrastructure
Smart control systems based on machine learning can significantly reduce the level of energy consumption of buildings, taking into account weather conditions, the current occupancy of the building and other environmental factors, and then adjust the heating, cooling, ventilation, and lighting in the room accordingly.
Smart buildings will be able to transmit information about the current state of the environment directly to the grid so that energy consumption can be reduced if there is a shortage of low-carbon electricity supply.
AI will be able to more accurately calculate the amount of energy resources used
In many regions of the world, data on the level of local energy consumption and emissions of greenhouse gases into the atmosphere are practically absent, which can be a big problem for the development and implementation of effective offset measures.
Machine vision will allow satellite technology to be used to estimate the built-up spot (area) so that machine learning algorithms can use this data to calculate energy consumption and emissions. Similar methods can be used to identify buildings that require upgrades to improve their efficiency.
Artificial intelligence optimizes supply chains
Using similar capabilities, machine learning technologies will be able to optimize channels and supply chains by minimizing the carbon footprint of transporting various goods.
The possibility of more efficient forecasting of the law of supply and demand will reduce production and transport waste.
Artificial intelligence will make precision farming scalable
Most modern agricultural farms use the principle of growing monocultures. In other words, only one crop is grown over a large area.
This approach makes it easier for farmers to work their fields with agricultural machinery and other basic stand-alone tools, but at the same time depletes the soil, depriving it of nutrients and thus making it less productive. As a result, various fertilizers are often used to increase yields, in particular those based on nitrogen, which can be converted into nitrogen oxides - greenhouse gases 300 times more dangerous than carbon dioxide. Machine learning robots can help agriculture assess the current state of the soil and suggest which crops to plant to restore soil health while reducing the need for fertilizers.
AI will help more effectively monitor deforestation
Deforestation contributes to approximately 10 percent of total greenhouse gas emissions. Tracking and preventing this often illegal activity is usually a very time consuming and routine process that requires personal supervision on site.
In turn, satellite imagery, coupled with machine vision technology, will allow automatic analysis of forest cover loss on a large scale, and special sensors installed on the sites, combined with algorithms that can, for example, detect the sounds of chainsaws, can help law enforcement agencies to more effectively deal with illegal activities.
AI will help change our consumer attitudes
According to the authors of the report, there is a widespread misconception in the world that ordinary people are unable to have a serious impact on climate change.
Therefore, in this matter, it is necessary to clarify how exactly people can help. Machine learning technologies can calculate a person's carbon footprint (the sum of all greenhouse gas emissions that they create in the course of their daily activities) and make small changes that will reduce it. For example, the system might suggest using public transport more often than personal transport; less often buying meat in the store; or reduce electricity consumption at home. Each of us individually creates a small carbon footprint, but if you take all of them at once, the numbers are much larger. Changes in our attitudes towards consumption and the addition of all the individual actions aimed at this can have a large cumulative effect.
AI will improve the efficiency of meteorology and climatology
Many of the most significant climate change impacts in the coming decades will be associated with highly complex natural systems, such as changing cloud or ice sheet dynamics.
These are the very issues in which AI has high hopes. Accurate modeling of these processes will help scientists to better predict extreme weather conditions (such as hurricanes and droughts), which in turn will help states develop methods of protection against the worst effects of these events.
Artificial intelligence will help with geoengineering
At this stage, this use case for AI among all the ones presented above is the most speculative, but high hopes are also pinned on it, at least from some scientists.
If we can develop ways to make the cloud cover of our planet more reflective, or even create artificial clouds based on special aerosols, then we can reflect more sunlight from the Earth. But this issue requires serious investigation. AI can help with this, but the authors of the report note that this method of using artificial intelligence is a very distant issue that will require the cooperation of all governments in the world. For example, experts from the Canadian University of Waterloo agree with this position, who believe that this unreasonable approach to the issue of geoengineering could start a third world war.