AI for Good — How Artificial Intelligence can Help Sustainable Development
Transcript of Keynote at Startup Energy Transition Festival 2018, 16.04.2018, Berlin
The development of Artificial Intelligence is one to the most important events in recent human history. The final outcome of it, is still to be determined. At a moment when many might see the development of AI as potentially more threatening than beneficial, a growing coalition of researchers and innovators around the world tries to make sure that the opposite is the case. This movement towards “AI for Good”, is gaining significant momentum and brings up relevant questions just at the right time. It also drives business and innovation in areas where Artificial Intelligence is used as a promising tool for sustainable development.
In only a few thousand years, while becoming the most dominant species on the planet, we have fundamentally transformed our environment. Human activity has long become the most defining influence on our global ecosystem. The pace of change has rapidly accelerated in the last 250 years since the invention of the steam engine and the resulting first industrial revolution.
Now, we are at the brink of a new revolution, which is driven by the transformative power of emerging technologies. Technologies like Artificial intelligence, genetic engineering, AR/VR, robotics, and 3D printing will fundamentally change our lives in the future. They are incomparably more powerful and develop at a much faster pace than any technology before. While, previously, it may have been possible to learn from our mistakes when testing out new inventions, one fatal error could now be just too much. This means that, more than ever, we have to engage in strategic thinking to avoid as many threats as possible and make sure these technologies are used in a beneficial way.
Technologies are tools, and can only provide value when applied for a meaningful goal. It is our core responsibility to constantly evaluate the direction we are headed and identify desirable future goals. In 2015, the United Nations published the 17 Sustainable Development Goals as a roadmap to guarantee a sustainable future by 2030. These goals range from ending hunger and poverty over realizing sustainable energy and gender equality to preserving our biodiversity.
Over the last few hundred years, technology and innovation have lead to a vast improvement of our living conditions. While surely, technology is also among the culprits of many of our problems, it is clear that technological development and innovation will be at the heart of moving towards a sustainable future. In many instances we have simply just passed the point at which a pure limitation of our human impact will do the job.
AI can play an important role in forwarding sustainable development. This is not because of the hope — or fear — of the advent of a hypothetical future super-intelligence which will solve all of our problems — or be our worst and final problem all together. The hopes are based on the application of some key capabilities within the domain of AI. These are capabilities in the area of Natural Language Processing, like Speech Recognition, or Machine Translation; or in Computer Vision, with Image Recognition and Classification; or in general in the field of data analytics and pattern recognition.
Most of the recent advances in Artificial Intelligence are based on the application of different — more and more sophisticated — Machine Learning techniques. Using such algorithms, computers are equipped with the ability to learn without being explicitly programmed. What you need to train the algorithms is all the time the same: you need a suitable set of training data, and you need sufficient computational resources to do the training. The more sophisticated algorithm you use, the more of these resources you need. The fast paced growth of both of data and GPU — together with the development of better algorithms — is now bringing forward ever more astonishing applications of Artificial Intelligence.
So how and where can we use this for sustainable development? The hopes and ideas are many, ranging from applications in healthcare and energy over agriculture and education to environmental protection. Let’s look at some initiatives that are putting these ideas into practice.
An exciting field of AI application lies in the use of satellite data. The US startup Planet Labs was founded in 2010 by 3 NASA scientists. It now operates the largest satellite fleet in history, with over 200 mini-satellites — not bigger than the size of a brick — orbiting our planet. The first goal of the startup was to scan the entire landmass of the planet once a day. They reached that goal in late 2017. Companies can now access that data and even order custom features. That means that private access to close to real-time monitoring of the earth’s surface has become reality.
One interesting AI use case of satellite data is Global Fishing Watch. The initiative was launched in late 2016 as a transparency platform with the goal to protect the world’s fisheries. Every day, itprocesses over 22 million position messages from more than 200,000 ships to detect patterns that signify which vessels are fishing, when and where. Doing so, it allows anyone with an internet connection to see fishing activity anywhere in the ocean in near real-time, for free.
Another use case is a research project at Stanford University. They compared high resolution images of rural areas in Africa in order to identify infrastructure and characteristic features and then predict how rich or poor certain areas are. It showed, that their predictions matched remarkably well with other available sources. With their research, they want to provide an easy tool to track progress towards eliminating poverty.
The agricultural sector is surely bound for an AI revolution as well, with hopes for higher efficiency and lower consumption of vital resources.
Montreal-based startup Nectar developed a fully integrated beehive management system. After their custom sensor has been applied, it allows beekeepers to monitor their beehives in real-time over their smartphones. That way they get early warning signals when something is wrong — without any disturbance of the colony.
Recently, the San Francisco based vertical green-house company Plenty made the headlines for closing the biggest seed fund round in agtech ever. The world´s biggest tech fund, SoftBank Vision, invested over $200 million in the company. Plenty uses IoT sensors and machine learning to grow crops vertically indoors using only light, water and nutrients. The company claims it uses only 1 percent of water compared to traditional farming techniques while at the same time growing up to 350 times more produce on one squaremeter.
In the energy sector, a lot excitement is around the implementation of intelligent networks, that connect producers, consumers, and storage, and supply energy just in time when needed. AI is used to predict energy consumption peaks and help with real-time optimization of operations settings.
The German company Gridhound targets network providers as customers. It is a spin-off of the Institute for Automation of Complex Power Systems at the RWTH Aachen. The company developed a cloud-based Advanced Distribution Management System. It allows network providers to monitor and optimize their distribution grids in real time through a pay-per-use machine learning based software solution.
On the private consumer side, Ecoisme is smart home application and energy monitoring system which is connected to your mobile phone. It detects all major home appliances, measures power consumption and analyses the user behavior. By sending out reminders and suggestions for positive behavior (like closing the door of your fridge that you may have left open), it promises to save up to 15% of electricity consumption.
There is also great potential on the research side, for example through machine learning enhanced prediction of new energy materials. One example of this is Berkley Lab’s Materials Project that combines machine learning with density functional theory calculations to identify promising new materials compounds. It has calculated over 50.000 electronic bandstructures by now.
With so much potential for sustainable AI applications at hand, it is crucial to provide the right incentives for innovators to move into this direction. Competitions like the Startup Energy Transition can play an important role by rallying up their ecosystem and providing shining examples of what is possible.
In the case of AI for Good, the XPRIZE Foundation teamed up with IBM Watson in 2016 to lay up the IBM Watson AI XPRIZE. The AI XPRIZE offers $5 million in prize money for teams to demonstrate visionary examples of how to use AI to tackle the worlds grand challenges. The prize is now in its 2nd round with 59 teams competing from 14 countries. A wildcard opportunity in December this year will allow aspiring teams to enter into the competition for the last time. Together with the already existing teams, they will compete for the grand prize to be presented at the TEDGlobal stage in 2020.
Promoting sustainable use of AI technologies is not the only field of AI for Good. Globally, politicians, researchers and companies are coming together to develop common principles that should make sure that all AI technologies are applied responsibly. Recently, the innovation consultancies accenture and PwC have both included the pressure towards responsible AI among their key technology trends for 2018.
A very important area certainly is the topic of Explainable AI, which wants to crack open the black box of machine learning algorithms and provide ways to understand the reasoning behind machine decisions. Only this in turn will provide a route towards accountability of AI systems. Just look at the recent case of the tragic and fatal Uber accident, that ended in the death of a pedestrian and the suspension of all of Ubers testing of self-driving car activities. The aftermath of the tragedy shows how important it will be in the future to be able to assess the reasons behind certain decisions of AI systems.
Other concerns lie in the field of digital ethics, which tries develop guidelines to make sure that autonomous agents comply with our moral and ethical standards. Data privacy will be become even more important as we need to protect individuals from the growing desire from corporations and intelligence agencies for data as the most valuable resource for the development of more powerful algorithms. Wide-spread AI education and accessibility is crucial in order to prevent that AI will increase the global divide and lead to an unfair advantage in the hands of very few countries and organizations. Making sure that AI is not used in a directly harmful way is a very important concern. By now, several thousands of AI researchers around the world, including leading thought leaders like Tesla’s Elon Musk and Deepmind’s Demis Hassabis, have signed the petition for a ban of autonomous weapons systems.
In January this year, the German DIN Institute has started a working group on Artificial Intelligence, focussing on standardization as a way to more AI safety as well as addressing issues like ethics and data privacy. Last week, Deutsche Telekom invited industry experts for a pre-release of the company’s first AI Guidelines. Industry initiatives like Partnership on AI, or Audi’s Beyond initiative try to pro-actively engage with the topic through fostering collaboration and targeted projects.
Another important initiative has been started last year with the AI for Good Global Summit, hosted by the United Nations in collaboration with the XPRIZE Foundation. It creates a neutral platform for exchange and brings together UN officials with AI experts and industry to build common understanding for the opportunities and challenges of the implementation of AI technologies. In one month, the second AI for Good summit will present new, practical results of breakthrough teams demonstrating the use of AI for achieving the Sustainable Development Goals.
Using the words of of US historian Mervin Kranzberg: “Technology is neither good, nor bad; nor is it neutral”. This holds true also for Artificial Intelligence. It is 50 years ago that Stanley Kubrick released its science fiction opera “2001: A Space Odyssey” where the autonomous computer HAL attempts the hostile takeover of the spacecraft Discovery and its crew. While these doomsday scenarios are certainly not at immediate stake, it will be vital to ensure that a most powerful and transformative technology like artificial intelligence will be used in a beneficial way.
This is what AI for Good is about. In order to do so, we have to create efficient incentives for innovators to develop AI applications for sustainable use. We have to ensure the responsible application of all AI technology, everywhere. We have to promote AI education, literacy and accessibility in order to prevent a threatening digital divide. And we have to strengthen global exchange and cooperation around the topic to make sure that the right ideas and solutions find their suitable targets.
