A.I. and Your Food
Can AI really help our food system by delivering better food?
It’s hard to avoid AI.
It’s everywhere we look and if you were to believe certain tech entrepreneurs, it can solve every problem we have today.
On the other hand, other experts are warning of a dire future if our regulatory systems don’t keep pace with the technological developments.
Cue visions of the world in the Terminator films for effect.
Beyond these justifiable concerns, entrepreneurs are trying to join the data handling power and deep learning capabilities of artificial intelligence with our food system. Their goal: find ways to grow food better, more sustainably, with greater precision and with greater predictability. Even food producers are jumping in, with the use of AI to develop and bring to market products faster than ever before.
The results so far, are intriguing.
Let’s start with a cattle ranch. A ranch is like any other organization today as they are facing labour shortages. Add to that the severe time crunches around calving time and the need for a sixth sense to determine if a particular animal in a herd is healthy or not.
Along comes Onecup AI. This small firm has developed an application called BETSY that depends on the collection of different data points to better assess the current health and disposition of a particular cattle herd.
The rancher installs a series of integrated cameras/processing units on poles that can overlook both outdoor pens and inside barns. The cameras can then be connected through WiFi routers with the initial data gathering done locally on each unit.
Once installed, the BETSY application — through the data it collects through the cameras — is then able to identify and track any animal within a herd through identification of each animal’s unique physical characteristics. Once identified, the data gathered on each animal is used to track what the they eat, their physical condition and when a calf is expected to be delivered. BETSY can track disease indicators, such as coughing, respiration and limping; activity of the animal, such as walking, running and lying down; the intake of nutrition and, over time, the growth of each animal.
All the data gathered is stored in a cloud environment that is accessible to the rancher wherever they are. The rancher can also add in data for each animal concerning interventions by veterinarians and all information gathered is available through a central hub in the application, that includes archived video feeds, video playback and a holistic view of the entire herd.
Another example is vertical farming. This is where certain vegetables or fruits can be planted on successive shelves in an indoor warehouse structure, where the environmental conditions are carefully managed to guarantee production yields.
Growing in such an environment requires access to a considerable amount of data concerning the artificial environment that is created, but also, on the plants themselves.
QuantoTech in Vancouver has leveraged its vertical farming expertise through the development of Light Emitting Diode (LED) lighting systems that can operate at different wavelengths, such as red, blue, ultra violet and white. Software can then vary the wavelengths to tweek the morphology (form and structure) as well as the phytonutrient quality of a particular crop.
This is layered on top of a typical hydroponic system that can then vary and control the excess heat that is thrown off by the LED lighting systems. Where AI comes in is in the management of all the data variables. These can be controlled via a dashboard as the system is programmed with proprietary grow recipes that QuantoTech has developed, giving the operator a seamless experience. An added plus: this can all be monitored remotely, which minimizes labour costs and maximizes the number of sites that a client can operate at the same time.
Then there’s food.
Kraft-Heinz is using AI to swap traditional ingredients in some of its key products to convert them to plant-based.
They’re latest stab at this is with Kraft Dinner, or KD.
Kraft-Heinz is in a joint venture with NotCo, a Chilean company that specializes in replacing animal-based components of food with plant-based components, all the while maintaining taste, mouth feel and appearance. They have three products they market: Not Milk, Not Burger and Not Chicken.
How does NotCo do this? By using AI! They have developed an AI platform that is essentially a database that knows — at the molecular level — the composition of animal-based food products. With this as a baseline, they then match that information with the same molecular information on 300,000 plant-based alternatives to find the closest match.
This is what they’ve done in partnership with Kraft. They analyzed the molecular composition of the regular Kraft Dinner and swapped out the animal-based components with plant-based ones, in an attempt to conserve the colour, feel and taste of the original product.
The more times the NotCo AI platform — called Guiseppe — does this, the better it gets at doing the swapping. It uses deep learning to be able to adapt and learn from each exercise in ingredient replacement. As it learns, it gets faster at completing the swap. NotCo started with a mayonnaise product that took the platform some 18 months to complete the analysis and recommend replacement products. NotCo’s next product — NotMilk — took 10 months. It’s most recent — NotChicken — took 2 months.
And Kraft Dinner? No one is saying how long it took, but this is the third collaborative project undertaken by the Kraft-NotCo joint venture. The two other products are Kraft NotCheese slices and Kraft NotMayo.
All this in an attempt to preserve, as much as possible, the same taste but with a different ingredient profile. After all, when you’re selling 1 million boxes of KD a day, you don’t want to muck with the quality or taste that consumers are used to.
Other examples abound. There is an AI-powered robot that is learning how to pick blackberries, and is getting better and better at it with each iteration.
The University of Florida is experimenting with an AI-powered ‘flavour connoisseur’ that is trying to identify the best chemical compounds that produce the best fruit flavours.
XpertSea is using AI to assist shrimp farmers in South America to better manage their stocks.
The examples are endless, as AI-driven tools become more accessible to smaller enterprises, driving down cost and increasing take up rates.
In the end, will the addition of AI tools in farming, transportation, retail and in the home give us a better food system?
It might. If farmers can harness the oceans of data that are generated on a single farm and increase the efficiency of their operations; if transportation companies can harness truck and driver data to increase efficiency (and, hopefully, better configured routes to reduce total driving time for drivers); if grocers can harness pricing, demand and supplier data to offer a better in-store shopping experience and generate less waste; and if consumers can use AI-driven tools to match health data to food requirements and preferences, then yes, we might have a better food system.
A food system that produces less losses and more income for the farmer, better driving conditions for the truck driver and less wastage en route, a better and less wasteful retail shopping experience, and, maybe, a healthier population if such AI-tools are widely available and not to just for an elite — these are all worthy objectives for AI.
They key is going to be the governance structure around AI. Right now, it’s the wild west out there, and companies developing the tech are largely unregulated by anyone else other than their own Boards of Directors. Look at the SNAFU that erupted for Open AI.
For such a technology with such astounding possibilities and dangers for human existence, it’s probably worth some supervision, and commonly developed frameworks. Then we can be sure it will be working for us and helping us develop that better food system we need.
