The way we design new products and services has scaled in ambition to include our entire economic system. New tools such as artificial intelligence can ensure that the only limitations we experience are those of our imaginations and the world’s finite resources. This learning path explores what artificial intelligence is and how it can be used to accelerate the transition to a circular economy, focusing in particular on the opportunities for AI to:
The science of making inanimate objects smart.
AI is an overarching term for a collection of technologies. It deals with computer models and systems that perform human-like cognitive functions such as reasoning and learning. AI software is capable of learning from experience, differentiating it from more conventional software which is preprogrammed and deterministic in nature.
Artificial intelligence (AI) doesn’t necessarily mean giving intelligence or consciousness to machines in the same way that a person is intelligent and conscious. It simply means the machine is able to solve a particular problem or class of problems.
AI helps to solve problems through performing tasks which involve skills such as pattern recognition, prediction, optimisation, and recommendation generation, based on data from videos, images, audio, numerics, text and more.
What’s the difference between general AI or specialised AI?
General AI is a general-purpose system that exhibits intelligent behaviour comparable to a human’s full range of cognitive abilities. Narrow or specialised AI is good at performing specific, well-defined, tasks normally associated with human cognitive abilities, such as image and voice recognition, trend prediction, and pattern spotting.
There is a common misconception that AI algorithms are ‘smart’ by themselves. In fact, AI is dependent on humans to clearly establish the inputs and outputs for a model (piece of software) before a machine can solve it.
An example is the voice assistant such as Siri, Alexa or Google Assistant - which needs to be able to understand speech and respond with a sensible answer or action. However, in order to effectively train the algorithm and adjust the input data accordingly, humans need to know what type of questions they expect it to be asked and what a sensible response would be. The lifecycle of AI development typically follows a process of data collection and ‘engineering’, algorithm development using the engineered data, and refinement as the data input is tweaked to achieve the expected outcome. Once the expected outcomes have been achieved to an acceptable level, decisions can be made based on the algorithms output. As the quality of the data improves over time, the quality of the algorithms output will also increase.
Take the simple example of developing an AI algorithm which can recognise pictures of cats.
The right kind of data has to be collected (in this case photos of cats and other animals) and it has to be ‘engineered’ - that is, reformatted and labelled so the algorithm can understand what it is looking at. The photos with cats and other animals will have to be tagged as ‘cat’ or ‘not cat’ so the algorithm can learn what type of features are unique to a cat. The data engineering process often requires a lot of manual work to manipulate the data into the right format. What you put in is what you get out, so good quality data is a very important consideration for applications of AI.
The algorithm is trained using the labelled photos of cats and other animals, and refined until it can apply what it has learned to unknown photos. Humans need to know what they expect to see as a result of the algorithm performing its task so the results can be sense checked. The results, for example, may include both photos of cats and photos of cat toys. The algorithm will then be refined to recognise the difference between a real or fake cat using additional images and information.
Each algorithm is trained to perform a very specific function, such as object detection for autonomous driving, identifying fraudulent financial transactions or delivery route optimisation.
Understanding this AI development process is important background knowledge for considering how we can apply AI to a circular economy, enabling us to think about key questions such as:
Which of the following statements about AI is true?
AI is not a new technology, nor do its origins lie in robotics.
There are numerous myths surrounding AI. One of the most common is that AI is a new technology. In fact, the first academic project investigating AI was in 1956 when a small group of mathematicians and scientists gathered for a summer research project on the campus of Dartmouth College. The reason it feels like a new field is because what we call ‘AI’ keeps changing. Clever things like automatic number plate recognition for cars (developed by UK police in the late 1970s) are now taken for granted. What we’re seeing today is simply the next step in the long-running evolution of developments to make computers better at analysing data.
Data is absolutely crucial to the development of AI, but the quality of the data is much more important than the quantity. Developing AI does not necessarily require huge amounts of data, but well labelled, clean data sets. Labelling involves translating messy real world data into a format that the AI algorithm can understand, for example, tagging an image of a car with the label ‘car’, which could involve a lot of manual human work.
Now that you understand AI a little better, how do you think it could be used to enable a circular economy?
Creating regenerative systems by introducing AI to design, business models, and infrastructure.
AI can be a hugely powerful tool. Imagine if it was being used to accelerate the transition to a circular economy and create new opportunities for large scale positive change. In the following sections we will explore how employing AI in our design, business models, and infrastructure could increase our ability to create new, regenerative systems based on the principles of circularity.
Informing designers’ choices.
The circular economy puts a strong focus on design. Indeed, the Circular Design Guide asks: what if you could redesign everything?
Designers working with AI can create products, components, and materials which are fit for the circular economy. Employing AI can account for better designs faster, due to the speed with which an AI algorithm can analyse large amounts of data and suggest initial designs or design adjustments. A designer can then review, tweak, and approve adjustments based on that data. AI gives designers a more informed insight into the most effective designs to create and test to make the best use of their time and expertise.
The following three case studies illustrate how AI is being used right now to enable better, more circular design.
Enabling new business models with AI.
Since the industrial revolution, the linear economic system has become gradually more optimised and efficient, most recently using digital technologies such as AI. Similar techniques could be applied more widely to circular business models to increase their competitiveness.
The case study below showcases how one company, Stuffstr, is using AI to dynamically set prices in its fashion resale marketplace.
In the video, Martin Stuchtey - co-founder and managing partner of SYSTEMIQ, paints a picture of a future mobility system in which AI plays a key role.
Streamlining the circulation of materials in the economy.
Products at the end of their life are not as uniform as they were when first manufactured, so they are harder to automatically disassemble, sort, and separate. Their condition typically needs to be manually inspected and then treated based on what damage or wear and tear they have sustained.
There are many opportunities for AI to help streamline the infrastructure needed to circulate materials in the economy - many of them focusing around the ability for AI algorithms to recognise and identify objects using cameras and other sensors.
The three case studies below demonstrate how AI is already being used to improve and optimise processes such as waste sorting, recycling, and sorting of food produce.
Is AI the answer to all of our questions?
While there is huge potential for AI to be a force for positive change, it also raises questions about building fairness, interpretability, privacy, and security into these systems - which are currently active areas of research and development.
AI applications need systems designed to follow best practice, alongside considerations unique to machine learning. With the potential to be fairer and more inclusive than decision-making processes based on ad hoc rules or human judgments, comes the risk that any unfairness in such AI systems could incur wide-scale impact. Thus, as AI increases across sectors and societies, it is critical to work towards systems that are fair and inclusive for all.
Ultimately, AI can be a very helpful tool to achieve circular economy ambitions, but it should be held to account by humans, driven by human values and principles, avoid creating or reinforcing unfair bias, and account for privacy and security of data.
There has never been a more exciting time to be an innovator.
Some AI systems are already in the market and just need to be used more extensively in a circular context, particularly for circular business models. This could facilitate everything, from predictive analytics (such as setting the optimal service and repair schedule for durable equipment), to dynamic pricing and matching for the effective functioning of digital marketplaces for second hand goods and by-product material streams.
As the technology develops, in a similar way to how it is being used today to improve traffic flow through cities, AI could be integral to the redesign of whole systems, which create a circular society that works in the long term. If you’d like to know even more about the potential for AI to enable the shift to a circular economy, the scoping paper, Artificial Intelligence and the Circular Economy - written in collaboration with Google, and with analytical support from McKinsey & Company - provides more detail.
LEARNING HUB FUNDED BY
Eric and Wendy Schmidt Fund For Strategic Innovation
We have detected that you are using an older browser. Please update to the latest version of Google Chrome, Mozilla Firefox or Microsoft Edge to improve your user experience.
If you are unable to upgrade your browser, please see our Technical FAQ page to get tips on how to improve your user experience.
