AI-griculture

When you think of the fruits of artificial intelligence, do you think of farms? No? The average person wouldn't but a certain team of researchers at Media Labs at MIT have shown that they're not average in any way whatsoever.

Media Lab's Open Agriculture report claims that using Artificial Intelligence to control the environment in which the plants grow by placing the plants in what is a shipping crate and creating mechanisms to control the temperature, the humidity and the light, researchers managed to isolate flavourful molecules and train a program that could adjust the living conditions of plants on a farm to maximise their production of those "metabolite chemicals".

Flavour is just one example of a metabolite chemical, other metabolite chemicals include vitamins, anti-oxidants and nutrients. In general, a metabolite chemical is one that is produced by the plant in response to its environment. The aim of this AI is to learn what environmental settings cause which metabolite chemicals to be produced and fine-tuning the environment to maximise their production.

The type of farming that this initiative uses is what is commonly known as vertical farming or urban farming, which is growing a lot of plants in a small dense and finely controlled area. This is a rising technique in farming, but there have been many failed attempts since these companies don't tend to share much information, this initiative aims to change that, by making all the software, hardware and techniques used opensource and for all to learn from or use.

When you tell your grandchildren about your trips to your granddad’s farm, this is what they’ll imagine.

When you tell your grandchildren about your trips to your granddad’s farm, this is what they’ll imagine.

Cyber-agriculture as a field has just been born and Media Lab's Open Agriculture initiative is the first of its kind and almost certainly more companies will come; it is a highly lucrative innovation. The plants are grown in ideal conditions to maximise taste and minimise cost, all the while being perfectly ethical and no requirement for modification of genes.

The goal of the farms so far has been to maximise taste, but the focus of the Artificial Intelligence can be shifted to make healthier plants that are less vulnerable to diseases or even to the changing climate due to global warming.

Media Lab's goal in this initiative essentially boils down to utilising the strides the field has made in machine learning in agriculture and AI that can explain in detail the environment of the plant and the plant's interactions with the environment.

A big advantage to this study is that neural networks can take a completely different approach when learning than humans, for example, to maximise the taste of basil, the AI realised that sunlight correlates with the production of the specific metabolite chemical and extrapolated with this data to try out leaving basil out in light conditions all the time; it was then discovered that the basil plant had a lot more of the required chemicals than a plant not being in light conditions all the time.

A traditional farmer would never realise this since the sun isn't up 24x7 in all places except the poles, but knowing this, farmers can adjust their farming conditions to maximise the duration of the sunlight the basil receives as much as possible, possibly even adding artificial lights. At the end of the day, this would lead to much tastier basil.

By using chemical techniques to measure the amount of each different chemical, the researchers also noticed that this significantly elevated the amount of other healthy chemicals in the basil, this entire innovation was brought by a computer; you could look at it this way: a program has made the healthiest and flavourful basil possible with those set of genes with a simple training method.

The training method included feeding data of previous experiments conducted and the chemicals produced by each experiment into a network and then letting the program experiment. This method can be used for other plants as well and hopefully, in many cases the skill between plants is transferrable and at the end of the initiative we are left with a general intelligence that can farm for taste or even make the healthiest plants possible.

The next step in the research includes letting the program learn how adding different plant hormones, the colour of light the plant receives and the nutrients it receives can affect the chemicals produced, with the aim of giving the AI more variables to maximise to get the healthiest and tastiest produce.

This method can also be used to increase the yield of medicinal plants and not just ones for consumption. There are certain plants across the world that produce chemicals useful for the treatment of different diseases. 80% of the world relies on medicinal plants but they're almost universally in decline due to high demand, in India, a rapidly increasing number of species are over-exploited and are endangered. Widespread usage of OpenAg's techniques could potentially reverse this since fewer plants would be needed and the demand could be met without over-exploitation.

As the climate changes, we are going to need to be more and more clever about how we grow plants so it should come as to no surprise if we see this technique return widespread in the not so distant future.