If the recent wave of AI (Artificial Intelligence) hype hasn’t grabbed your attention, then you may be surprised one day to find a T-800 Terminator riding a Harley up your driveway asking for the whereabouts of Sarah Connor. Or at least that’s what some in the media will have us think. Although the potential of AI is significant, the tech accessible for us ordinary folk is not quite at the ‘enslaving humanity’ stage.

According to IBM, at its simplest, artificial intelligence combines computer science and robust datasets, to enable problem-solving. The length and breadth of AI is too great for an article like this, but we can muster a shallow dive into the highlighted part – robust datasets – as it relates to Ag.

The recent AI exhortation is with respect to language based generative tools such as Chat GPT. Basically, these tools enable an analysis of War and Peace to be written in seconds by just asking it to write an 800 word essay on 'War and Peace'. This highlights its great strength (War and Peace clocks in over 1,200 pages) and touches on a fundamental weakness.  AI needs to have access to a robust data set that has already considered the topic otherwise garbage in = garbage out. Because War and Peace has been the subject of literary examination for well over a century, there is no shortage of material for Chat GPT to riff from. But something useful for WA farmers, such as, “Should I plant hybrid canola in the deep sandy duplex road paddock on 50mm of rainfall by Anzac day?” requires data sets so specific that they are not in Chat GPT’s sphere, if they have even been created at all.  

Farmers generate a lot of data that is specific to their farm (more than most realise) and tools are being developed that will eventually allow the massive computational power of AI to access your own data to solve problems.  Although farmers can currently use existing tools to bring together disparate data sets, they are clumsy and time consuming and the farmer has to provide the brain power to turn this into useful information. With AI grunt, answers to problems will be clearer and long held traditions challenged, thus driving better farming outcomes. Will a paddock that has been seeded to one commodity at a time for 100 years be cleaved apart because the algorithm determined that the valley area would perform better with a different approach than the elevated area? Almost certainly, and that will only be the start of it.

So where is this data?

Every farm in WA is under the scrutiny of satellite data algorithms whether they like it or not. They are used primarily for crop estimation throughout the season and are useful for storage and handling, service planning, market pricing and even banks assessing credit risk by gauging long term farm profitability.

Meanwhile, machinery data flies off to the cloud and head offices in the US and Europe and the myriad of smaller operational gizmos such as pump management systems collect a heap too. (Even a modern fridge does, but that data could only help answer the problem of an expanding waistline ). Then there are service providers in the form of CBH, fertiliser companies, chemical providers, advisors and so on which also hold a lot of data that effectively comes from the farmer.  

Farmers also punch data direct into the likes of Agworld, Agrimaster, Laconik (coincidentally, all WA based software companies) and daily rainfall measurements collected in an Excel sheet. Also useful are external datasets such as GRDC’s NVT results, the local grower group’s trials and CBOT market prices.

At the moment, bringing all these together is a costly and difficult exercise but, one day, the power of AI will be able to chew through this and then some. Working out the most profitable system for a piece of dirt will become more and more mathematically precise and the practice of successful farming will involve harvesting data – specifically robust data  – as much as the crop.