AI and agriculture

This Commentary reviews what Artificial Intelligence (AI) might be able to do for agriculture. It predicts that one fruitful area for this will relate to how professionals work with farms and farmers. It ends with some cautionary remarks on how farmers should be careful to control the data on which AI will be trained.

Defining AI

Artificial Intelligence (AI) enables computers and machines to use information to achieve things autonomous of human input.

Social media feeds use AI. So do smart kitchen devices. Therefore, dare I say it, AI already runs (much of) our lives.

Even if we seldom think of such apps and devices as AI.

Thus, a washing machine controller, which enables the machine to recognise a part load and then use half as much water as old models, thanks to sensors and algorithms, may not seem like AI. Even though it is.

Often the term AI is reserved for breakthrough ideas like beating chess grandmasters or driverless cars. However, once these have been introduced, they no longer feel so radical. Once we have driverless cars, we will no longer think of them as AI devices. They will become commonplace, just another autonomous tool that is helping us.

With this perception challenge in mind, this Commentary will explore, first, where AI is already making a difference in farming. And then imagine where AI may take farming next.

Do farmers use AI tools?

Many farmers now have a plethora of professional apps on their smart phone. These allow them to order supplies, monitor stock, place fertiliser, pay bills, measure soil moisture, measure grain dryness, check weather, market produce, schedule tasks for staff and so on.

Beyond smart phones, other devices on farms and orchards also process information autonomously. E.g. modern tractors, seeders, sprayers and harvesters are now packed with sensors and algorithms to help optimise outcomes.

Therefore, just like the rest of us, farmers already interact with much AI.

How much impact is AI having on agriculture?

Farming is in many ways a highly suitable medium for AI. It is a large industry and highly measurable. Although currently not all this data is collected. And very little of it is processed into valuable insights. It’s a bit like the chicken and the egg; AI tools need data sets, which are large enough to ‘learn’, but will not penetrate particular farming activities until enough (good) data is available. Our view is the egg is starting to hatch. Agriculture is moving from being data-lean to data-abundant.

Every farm season is unique, as are most farm production systems. This lack of standardisation makes many farms or the tasks within these farms less suitable for simple modelling. Although agriculture is very broad and therefore application opportunities of AI vary widely across farming sectors e.g. from a high-country sheep farm to an intensive apple orchard.

Even so there is no doubt that AI is steadily infiltrating agriculture, just as it is changing information flows in many aspects of everyday life. However, where might AI take farming and the agri-food industry next?

What is AI currently good at?

• AI tools tend to be good at:
• Text compilation (from auto-complete, to LLM)
• Solving complex puzzles
• Supervising robots
• Mining data
• Creating new information tools

How relevant are these for the future of agriculture and food?

Agriculture as an input-output system?

In my first attempt to answer this, I conceptualised agriculture as a series of input-output systems (such as input: nutrients+photosynthesis; output: wheat). However, this yielded underwhelming answers. Farming can be made more efficient, although mechanisation has already stripped out a lot of labour. So that most sectors are already pretty lean.

The horticulture sub-sector is an obvious exception. It does still employ large numbers of people. For tasks like weeding, pruning and picking. There is therefore significant potential for fruit and veg picking robots. However, these have been “5 or 10 years off” for at least a decade. The breakthrough moment for horticulture is coming closer. But slowly.

The reality is that mainstream farming itself is already a series of capital rather than labour intensive systems. AI will help the main input-output ratios of agriculture get more efficient however these gains may not be much faster than in the rest of society. With an exception likely to be fruit and veg operations, eventually.

A surprising discovery

My father died in 2004 and shortly afterwards I found myself re-learning farming. With so much ahead of me I planned simply to stay on the farm and get on with it. Just me, Bridgie, the kids and our staff. However, something unexpected happened. From that first week, helpers started calling me and showing up at the farm. And to this day they have not stopped doing this. What is going on?

It turns out that a farm is not really “run” by a farmer. Instead, he or she is a generalist assembler. No farmer has the skills to be their own agronomist, vet, seed specialist, machinery engineer, geneticist, or accountant. Instead, we outsource this expertise from a revolving door of “Ag Professionals”. On a piece-work basis.

Farmers and their salaried staff also don’t carry out some farm physical tasks. Almost every week, one or two expert contractors arrive on farm. To carry out sowing, spraying, harvesting, scanning, crunching, shearing, insemination (the original AI!) and many more tasks.

Might these eco-systems of professional and contract helpers offer opportunities for AI?

Can AI help ag professionals?

What I had not learned from Ag College, nor from my time as a shepherd, is just how many of these “Ag Professionals” there are out there helping farms. I now estimate there are seven to ten of these professionals per farm in a typical agri-economy. So, in most developed agri-economies, there are 300,000 to 500,000 of these good people. Most of them are white collar and therefore well-trained and open to using screens and data.

Might this army of helpers be made more efficient and effective by AI tools?

For example, with the right algorithms encapsulated in web tools, might each vet be able to look after 50 farms rather than, at present 30?

Might a field suffering from a disease receive more prompt diagnosis and treatment via agronomists getting alerts directly from sensors. Rather than having to walk the field?

Opportunities for autonomous machine support

It seems possible the very things that AI is good at may be what is needed by these professionals. Might:

• Vets, agronomists and farm advisors use text compilation to write first draft reports?
• AI parsing of symptoms help solve the puzzles of diseases?
• Robots accompany contractors to farms, making them more effective?
• Big data sets be mined by users to generate insights?
• New information tools be invented to further help farms?

My main conclusion from six months pondering AI and agriculture is that the farm service industries offer a ripe environment for AI to help farmers make better production and sustainability decisions.

Control of big data

One last issue that is too important not to mention: farmer control of data. It is no accident that the dominant tech companies are making the biggest AI advances. Firms like Google (i.e. Alphabet) and Facebook (i.e. Meta) have large private data sets of information on people like you and me, on which to train their AI. It would be a pity if a similar grand larceny were to happen to farming.

Farmers should ensure that all counter parties with whom they share data acknowledge, from inception, that farm data is the property of the farm, not the party who is offered access to it.

Fortunately, farm data standards such as the Australian NFF Farm Data Code; the UK’s Farm Data Principles; and Farm Data Standards / DataLinker (NZ) have been set up to address this. Firms who sign up to these standards commit that farmers and only farmers should be able to permission their data to AI and other projects. Please make sure organisations who get measurements from your farm commit to follow them.

Feedback and next Commentary – on inflation

It would be great to hear your views on AI I’ll aim to summarise them in my next Commentary, which will mostly focus on inflation and farm produce prices.

Forbes Elworthy.