In this first blog of a series of four on artificial intelligence (AI) and machine learning (ML), I
look at what it is and why it has the potential to transform the recycling and waste sector.
Other blogs in this series will include:
• Using AI and ML in MRFs and to understand your material composition
• How AMCS is developing AI and ML solutions for recycling vehicles
• The future of AI and ML in the recycling sector.
What is Artificial Intelligence (AI)?
Essentially, AI involves processes and algorithms that can replicate some human skills such
as problem solving and learning. In some ways, AI is already better at some tasks than
humans such as spotting patterns in vast data arrays. But we are not at the point where AI is
as talented as the human brain, and it may never be.
When used on certain tasks, AI can focus on processes that we might find laborious or
incapable of. For example, you wouldn’t place a member of staff on your vehicle to check
every item in the bins you collect, but cameras and AI are capable of this.
AI is also used an overall term, so within it, there are sub-sets such as machine learning,
reinforcement learning and deep learning.
What is machine learning (ML)?
ML is AI that takes data processing further. It learns by being trained. You tell the AI what is
wrong, and then it learns what is right using ML. In the recycling sector, it might be that you
want to train the ML to recognise PET bottles. So, you show it HDPE bottles, cardboard
boxes, steel cans etc and PET bottles.
As you are training it, most likely it will not always recognize PET bottle images, but when it
gets it right you confirm that it is. In the background, ML is using mathematical models to
code itself on the properties of PET bottles until it begins to identify them every time.
But increasingly, ML is advancing so that it no longer needs to be trained. This unsupervised
learning allows it to spot patterns in data and work things out itself. This is especially useful
in complex data sets, where there are no obvious trends that can be seen by humans.
Examples of this type of ML are the recommendations you might receive from Netflix or
YouTube that learns your viewing habits and then suggests other things you might like.
What is reinforcement learning?
There is also a type of ML called reinforcement learning, where it learns from the
environment it is placed in. This is useful for robotics where the device is constantly
improving. In the recycling sector, it could be that robots use reinforcement learning in
future to collect bins, understanding the challenges put in front of them such as gates,
plants and trees or items placed against the bin – all things that humans can deal with
easily, but robotics currently struggle with.
What is deep learning?
Deep learning is a form of ML where artificial neural networks are used to replicate and
surpass human analysis. It has been used for translation purposes, medical image analysis
and climate science.
Potential future uses for deep learning could include human-like ability to assess
contamination on sorting lines or identifying materials so that different paper grades could
be sorted.
Why is AI important?
We live in a society driven by data. That isn’t necessarily just structure data such as numbers
on a spreadsheet, but increasingly unstructured data such as images and tweets too.
This means AI is being used for everything from data processing to autonomous vehicles.
As an example, AI could be used to process customer data, to better understand the waste
and recyclate generated by customers. Patterns could be spotted that enable you to run
your business more efficiently, identifying where it may make sense for you to run a food
waste or paper recycling route,
Indeed, our AMCS route optimization software is a form of AI, as it analyses and works out
the optimal schedule based on the pattern of the customers you need to service.
When it comes to your vehicles, AI is likely to transform how they are used in the coming
years. Solutions already exist that can predict maintenance schedules, especially when
combined with telematics. (Read my recent blog on reducing transport costs with
telematics).
But self-driving capability is likely to be common on vehicles in the near-future, and AI is the
bedrock of this.
By analysing millions of images, recognising, and deciding on data such as traffic in front and
around you, pedestrians, cyclists, maybe even a bin blown into the road by strong winds,
the AI using sophisticated ML algorithms and artificial neural networks decides on how to
operate the vehicle safely and efficiently.
We aren’t quite at the point where self-driving will be used soon for recycling and waste
management, but it is probable that it will become normal over the next few years.
But AI is already making a difference for recycling and waste management companies, and
over the next few articles, I’ll show you how.
I’ll also look at how your business is likely to be a lot different in future thanks to AI and ML.