The recent AI renaissance has been asking many how this technology can help one of the biggest threats to humanity: climate change. A new research paper, authored by some of the area's best-known thinkers, will answer this question and give a series of examples of how machine learning can help prevent human destruction.
The proposed use cases range from the use of KI satellite imagery to better monitor deforestation and to develop new materials that can replace steel and cement (whose production accounts for nine percent of global greenhouse gas emissions).
Despite this diversity, the paper (which we discovered about MIT Technology Review ) always returns to a few broad uses. Prominent among these is the use of machine vision to monitor the environment. Use of data analysis to identify inefficiencies in emission-intensive industries; and using AI to model complex systems such as the Earth's climate so we can better prepare ourselves for future changes.
The authors of the release, which includes DeepMind CEO Demis Hassabis, Turing Award Winner Yoshua Bengio, and Google Brain Co-Founder Andrew Ng, said AI could be invaluable in mitigating and preventing the worse effects of climate change. Note, however, that this is not a "silver bullet" and urgent political action is needed.
"Technology alone is not enough," write the authors of the paper, led by David Rolnick, a postdoctoral fellow at the University of Pennsylvania. "[T] Technologies that would reduce climate change have been available for years, but were largely not adopted by society on a large scale. We hope that ML will be useful in reducing climate change costs, but humanity must also take action. "
In all, the paper proposes 1
- Build Better Electricity Systems . Electricity systems are full of data, but too little is done to use this information. Machine learning could help by forecasting power generation and demand, and enable suppliers to better integrate renewable resources into national networks and reduce waste. The British Google lab DeepMind has already demonstrated this kind of work and used AI to predict the power generation of wind farms.
- Monitoring of emissions and deforestation in agriculture . Greenhouse gases are not only emitted by engines and power plants – much of it is caused by the destruction of trees, bogs and other plant life that has bound carbon through photosynthesis over millions of years. Deforestation and unsustainable agriculture cause this carbon to be released back into the atmosphere. However, using satellite imagery and AI, we can see where this is happening and protect these natural carbon sinks.
- Create new low-carbon materials. The authors of the paper note that nine percent of global greenhouse gas emissions come from the production of concrete and steel. Machine learning could help reduce this number by helping to develop low-carbon alternatives to these materials. AI helps scientists discover new materials by allowing them to model the properties and interactions of unprecedented chemical compounds.
- Predict extreme weather events . Many of the biggest impacts of climate change in the coming decades are driven by very complex systems such as changes in cloud cover and ice layer dynamics. These are exactly the problems that AI can deal with well. Modeling these changes will help scientists predict extreme weather events such as drought and hurricanes, which in turn helps governments protect themselves from their worst effects.
- Making transport more efficient . The transport sector accounts for a quarter of global energy-related CO2 emissions, of which two-thirds are caused by road users. As with electricity systems, machine learning could make this sector more efficient, reduce the number of wasted trips, increase vehicle efficiency and shift freight to low-carbon options such as rail. AI could also reduce the use of cars through the use of shared, autonomous vehicles. However, the authors note that this technology has not yet been proven.
- Reducing Waste of Energy from Buildings Building energy consumption represents another quarter of the world's energy-related CO2 emissions and is one of the "lowest-hanging fruits" for climate action. Buildings are durable and rarely retrofitted with new technology. Adding a few smart sensors to monitor air temperature, water temperature and energy consumption can reduce energy consumption in a single building by 20 percent, and major projects to monitor entire cities could have even greater impact.
- Geoengineer a more reflective earth . This use case is probably the most extreme and speculative of all, but some scientists are confident. If we find ways to more strongly reflect clouds or create artificial clouds with aerosols, we could reflect a greater portion of the sun's heat back into space. However, this is a great if and modeling the possible side effects of schemas is enormously important. AI could help, but the authors of the paper note that there are still significant "governance challenges" ahead of us.
- Give Individuals Tools to Reduce Their Carbon Footprint . According to the paper's authors, it is a "common misconception that individuals can not take meaningful action against climate change". However, people need to know how they can help. Machine learning could help by calculating a person's carbon footprint and displaying small changes that could be made to reduce it – such as the increased use of public transport; Buy meat less often; or reducing the power consumption in their home. By adding individual actions a big cumulative effect can be achieved.