5 Ways Artificial Intelligence is Making the Food Industry More Sustainable
The food industry is solely responsible for almost a third of global emissions . We need to get this under control – perhaps artificial intelligence is the answer?
This is the age of information. Everyone and everything are more connected than ever. It can be pretty overwhelming, and not just for us people! With the ever-ballooning number of devices and computers, paired with an increasing expectation for processing speed, conventional methods of computation are beginning to fall short.
Artificial intelligence (AI) may be the pivot essential to adapting to this age.
In a nutshell: AI is a broad term that encompasses all manner of computation that takes inspiration from the human brain. This ranges from expert systems used by medical staff to machine learning, the latter being what most think of when they hear the term.
The food and beverage industry is one where millions (if not billions) of irregular items go through processing every hour. Long gone are the days when humans were expected to wash, sort, categorise, process, package, and send off such quantities of produce. But, with these tasks delegated to machines, we have unavoidably lost the “human touch”. Mountainous quantities of food waste are produced each year as a result of the inefficiencies of large-scale logistics.
Just the kind of inefficiencies that AI is perfect at diminishing. Here are 5 ingenious ways that AI is making the food industry more sustainable.
Increasing agricultural productivity through smart farm equipment and machinery
AI has impacted nearly all areas of society, but it is yet to make a significant impact within the agricultural sector. However, it is now growing at a significant pace and the AI agribusiness industry is expected to grow to $2.6 billion dollars by 2025. The growth in this area of agriculture has been claimed to be the start of the 4th agricultural revolution, where data will play a significant role in the development of the industry. AI is already having substantial impacts on farming in some areas of the world through automated irrigation and crop monitoring systems. Around 70% of the world’s freshwater consumption is used in farming and automated irrigation; GPS, sensors and other software can help direct irrigation to the right areas. This can reduce excessive water usage and subsequent environmental impacts such as agrichemical leaking and soil erosion. Furthermore, it can reduce both water consumption and labour needed for farming. This improves efficiency and can ultimately save ever-critical resources such as water.
Automated product ordering and replenishment at supermarkets
Almost half of all food produced in the world ends up in waste. This is often a result of bad distribution and supermarkets and large food producers not efficiently understanding – and aligning – consumer demands with production. Consumer demand can vary significantly, and many factors can influence it, such as seasonal variations and major events, such as sporting events. Supermarkets can find it hard to understand the peaks and troughs of consumer demand for products, leading to both wasted food and losses. AI software, however, has the potential to plug this gap, by developing complicated algorithms where software can undertake ‘Machine Learning’ and develop an understanding of factors that influence demand and automatically predict variations. This AI software can then allow for automated ordering for supermarkets which closely matches demand. This can reduce the time-consuming aspect of ordering as well as reducing wastage across entire networks of distribution and retail outlets. For example, Blue Yonder, an AI software company, has worked with Morrisons Supermarket to optimise replenishment of supermarkets and reduce wastage. Morrisons has already seen significant benefits including the reduction of stockholding in stores by around 2-3 days. The more AI systems like this are used, the better the software can learn to understand factors affecting demand. This will consequently lead to greater efficiencies.
Growing underground – hydroponic systems and LED technology
33 Meters below the ground in Clapham, London, Growing Underground, has been producing micro greens and salad leaves in hydroponic systems and LED lighting in a former WW2 air raid shelter. They have only been growing plants since 2015, but they already supply retailers including Ocado, Farmdrop and M&S. This form of growing does not require any soil to grow plants, but uses a nutrient solution within water. Sensors are used to measure every aspect of the environment around the plants, including wind velocity, temperature and humidity. This smart monitoring allows the optimisation of growing conditions while also increasing energy efficiency. The site is constantly monitored in real-time and allows for quick adjustment if the environmental factors change. This not only allows for greater energy efficiency, but this method of farming uses 70% less water than conventional soil-based methods. No pesticides and fertilizers are needed, as the bunker is enclosed and the environment optimised for growing. This farming is changing where food can be grown and allows for otherwise disused areas, such as bunkers and old factory buildings, to be put to good use. This also reduces the cost of transportation, reducing GHG emissions further, as food has to travel less distance to reach consumers.
Decreasing resources spent on logistics, aka: making sure your food arrives faster and fresher! We’ve already talked how AI is improving things on both ends of the supply chain: automating and optimising agriculture and enhancing supermarket stock management. But what about everything in-between?
Every link of the supply chain generates food waste. Most originates from the food lost during farming and from being chucked out by consumers; however, a significant amount of wastage also comes from food in transit. A theme issue titled, “Reducing food losses by intelligent food logistics” by Jedermann, Nicometo, Uysal, and Lang (2014)  delves into the effects that different transit conditions has on the shelf life of food.
They discover that there are a troublesome number of variables that can impact the longevity of fresh foods: temperature, humidity, even seemingly inconsequential factors like atmospheric composition and the position in the truck, are just a few examples.
“Transport temperatures and the position in the truck led to tremendous variations in quality, resulting in 57% of the berries arriving at a packing house without sufficient remaining shelf life.”
With so many factors affecting the quality of food in transit and storage, it’s unreasonable to expect each item of produce to be catered for with our current practises. Current means only account for the bulk of food – a lot will not survive the journey, but most will and that’s the best we can do.
How can AI improve things? Computer vision can be used to sort items of produce individually, based on remaining shelf life. Doing this at each station along the food’s long journey will allow food that is determined to have the least remaining shelf life, to have priority for the best conditions. Researchers Yuzhen Lu and Renfu Lu of Michigan State University give their analysis of the recent advances in computer vision techniques for the quality assessment of apples in chapter 11 of the book, Computer Vision Technology for Food Quality Evaluation (2016) . Currently, colour imaging is used to categorise apples by type; however, this is ineffective at gauging the interior structure and quality of apples. Magnetic resonance, and hyperspectral imaging are proposed technologies that would be able to diagnose the internal quality of fruit and veg (Lu and Lu, 2016) .
- Every step of the food supply chain generates food waste. AI can locate and forecast this waste so that it can be intercepted by companies like FoodMaven before it becomes landfill.
- https://foodmaven.com/ (I’m not sure they use a model based on AI, but it’s still a good example of how it COULD be used)
- TED talk: https://www.youtube.com/watch?v=Sp8XTMS8Nx8
- Delivery to front door
Highly adaptive energy use for refrigeration
In 2016, Google published an article briefing its success at implementing a machine learning system in one of its data centres. The system, dubbed “DeepMind”, was able to consistently reduce the data centre’s cooling bill by 40% while activated. 
Figure 1. Power Usage Effectiveness (PUE) with and without DeepMind intervention. (R. Evans and J. Gao, 2016)[/caption]
The article explains why an AI solution out-performs their old system so drastically: there are too many internal and external interactions happening too fast to be predicted by “traditional formula-based engineering and human intuition”.
This same technology could, in theory, be applied to any electricity distribution network. Google state that they are working on making this tech widely available for use, and not just for other data centres.
“We are planning to roll out this system more broadly and will share how we did it in an upcoming publication, so that other data centre and industrial system operators — and ultimately the environment — can benefit from this major step forward.” (R. Evans and J. Gao 2016) 
An example of how this could be used in the food industry is refrigeration. Imagine the impact of the same energy reduction seen at Google’s data centres in all the big refrigeration storages across the world. The power consumption could be fine-tuned to a precise temperature for each perishable in storage. All factors, internal (quantity in storage, air flow) and external (weather, time of year) would be accounted for to optimise power consumption to super-human levels.
By Charlie Creech, Junior It and systems consultant and Jonny Sheldrake, Junior systems consultant
 Jedermann, R., Nicometo, M., Uysal, I. and Lang, W. (2014). Reducing food losses by intelligent food logistics. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, [online] 372(2017). Available online at: http://rsta.royalsocietypublishing.org/content/372/2017/20130302 [Accessed 9 Nov. 2018].
 Lu, Y. and Lu, R. (2016). Computer vision technology for food quality evaluaton – Chapter 11. 2nd ed. Boston, MA: Elsevier, pp.273-304. Available online at: https://www.sciencedirect.com/science/article/pii/B9780128022320000116?via%3Dihub [Accessed 14 Nov. 2018].