How can we use AI to drive the sustainability agenda?
Earlier this year I kicked off this new blog series looking at how technology can drive the sustainability agenda. This time, I’d like to focus specifically on Artificial Intelligence and some existing and potential sustainability use cases.
Firstly, let’s delve into a little history. The study of Artificial Intelligence began way back in 1956 when a group of researchers got together in New Hampshire to discuss how machines could perform “intelligent actions”. Since then the landscape has gradually evolved, thanks to technological changes such as increased computational power and new discoveries in neuroscience.
How then should we describe AI today? AI refers to “any computer system able to perform tasks that generally require human intelligence”, according to the dictionary. Gartner defines AI as “technology that appears to emulate human performance, typically by learning and coming to its own conclusions”. How is that possible in practice? AI capabilities today derive from three major sources. First, data - the huge amount of information provided by the Internet of Things. Second, perception - that is, the ability to recognise voice and images. Third, cognition and problem solving - that is, the ability to continuously learn and improve.
No one can disagree that the potential of Artificial Intelligence is enormous. However, as Jedrzej’s article last week reminds us, although AI can mimic human intelligence, we decide what we want AI to be. AI systems absorb what we feed them. They take on our ideas of what is good and bad; our preferences and biases. The AI researcher Margaret Mitchell recently reported her experience of showing an AI the following image of a house burning down.
The reaction? “This is an amazing, spectacular view!”. Such a positive reaction is obviously not what we would expect from a human. Fear or sadness would be more natural. The machine’s reaction was based on previous experience, using colour and contrast recognition. Images that ‘matched’ this one in the machine’s ‘mind’ historically had all been positive images so this one provoked a positive response.
So, what does any of this mean in the context of sustainability? Simply, it means that if we want a more sustainable world, we must teach machines what sustainability is and how they can make a positive impact.
Sounds simple? But where do we begin when AI is such a pervasive technology, the application and impact of which span all sectors of the economy, from manufacturing and media to education to healthcare.
I’d like to look at some examples that touch upon how AI can be used to drive both environmental sustainability and smart social progress.
When the environment comes first: smart buildings and autonomous cars
AI has the potential to greatly improve environmental sustainability. On a small scale, machine learning mechanisms enable the constant reconfiguration of any system or infrastructure to use the lowest possible resources and eliminate waste. Data informs efforts to develop more efficient and greener systems, such as smart homes and factories.
For example, as shown in the picture above, Google’s DeepMind team has used machine learning to improve, by more than 15%, the cooling efficiency of their data centres, even after they had been supposedly optimised by human experts.
Autonomous cars will make an enormous ecological difference in the next few years. Pioneering companies such as Google and Tesla began researching and producing self-driving cars almost 10 years ago. According to the think tank RAND Corporation, these appear to offer a great opportunity - even when only 90% safer than human drivers - to save lives. As shown in the bar graph, a future without autonomous vehicles would have 2.1 million fatalities by 2070, against 1.5 million in a future with autonomous vehicles. Not only that, but autonomous cars also decrease pollution and make mobility overall more sustainable. A study by Nature found that transitioning to electric self-driving taxis could reduce emissions per mile by up to 94% by 2030 in the United States. And it’s already happening; the first self-driving taxi service, nuTonomy, was actually launched in Singapore in August 2016.
Imaging future societies: AI improving wellness and wellbeing
AI also opens new and exciting possibilities to address societal problems, ensuring healthy lives and promoting wellbeing.
In fact, one of the most discussed topics around AI is its potential impact on societal wellness and happiness. Specifically, around the shift in lifestyles that the technology will bring. Thanks to AI, it will be possible to automate many repetitive and dangerous tasks. Provided the socio-economic context is is place - and governments act in a way as to promote it - this could mean that people will have more time to devote to creative and high-mental-effort activities.
Many agree that AI will free us from the tasks we don’t like and allow us to focus instead on what we do like, ultimately making humans “even more human”. Some even image a future in which humans won’t necessarily have to work anymore. In a recent Guardian article, Toby Walsh, Professor of Artificial Intelligence at the University of New South Wales, states: “The irony is that our technological future will not be about technology but all about our humanity”.
Futurism aside, AI is already bringing valuable contribution to wellbeing by advancing disease research and prevention. Machine learning-powered knowledge allows for more rapid and accurate detection of early symptoms. Google’s Deep Mind has recently partnered with the British National Health Service (NHS) to deploy algorithms to help prevent sight loss. And it’s still fairly fresh news that a group of researchers has developed a machine-learning algorithm to detect signs of Alzheimer’s up to 9 years before its symptoms develop. That’s game changing in the research for the roots of this disease and especially for a cure, not yet identified.
Additionally, AI can contribute to making healthcare services more accessible in regions where medical personnel are absent or other constraints exist, saving valuable resources. Chatbots will soon be strengthening current telemedicine approach. Significantly, the AI-powered chatbot telemedicine app, Ada Health, recently won £47M in funding. By asking relevant personalised questions Ada Health identifies potential causes of symptoms as well as suggestions on how to proceed with care. This kind of app would have massive impact in developing countries, where there is often a lack of primary care.
AI’s negative side: not all that glitters is gold
Artificial Intelligence looks like incredibly promising technology. It’s inspiring and exciting but that’s probably also what makes it partially terrifying. There’s still too much vagueness around it. A fair and varied amount of criticism has been raised in relation to what extent AI can be used to promote development, touching upon the human elements that a machine lacks such as the ability to understand ethical and philosophical questions.
Among these, is the case of the scientist Joseph Weizenbaum. One of the fathers of AI, he created in 1966 - well before Ada Health - the computer program ELIZA that made users feel like they were speaking with an empathic psychologist. However, he progressively realised the technology’s pitfalls and finally became one of its leading critics. He set up a clear distinction between deciding and choosing. Whereas the former is a computational activity, the latter is a matter of judgement. Artificial intelligence can - Weizenbaum underlines - compute, but cannot judge, as it inevitably lacks crucial human qualities such as wisdom, empathy, compassion and so on.
He also reinforces that computers tend to be a “conservative” force. This allows them to continuously become more efficient, but ultimately hinders potential re-hauls of existing systems and, as such, social progress. Without a political and societal environment to support its development, AI alone certainly cannot foster sustainable transformation, if that transformation means a 360 degree cultural and operational paradigm shift towards sustainability.
All these thoughts have a sound foundation. Positive and negatives, opportunities and risks. Personally speaking, I share Margaret Mitchell’s view, that AI can and will make an impact in many different cases. The key to making AI contribute to a smarter, greener world is to take action now in order to visualise and direct how this technology could look like in 5 to 10 years.
Using an impactful metaphor, Mitchell clarifies that AI, under her definition, is not a “self-driving car”, but rather a “car that we are driving”.
Whatever your opinion, one thing is certain. We should certainly be broadening the AI conversation and inviting a wide audience to contribute to its development. This will allow us to make sure the outcomes of our investment in AI are in line with a sustainable future and a better place for all of us.