L’artificial intelligence is slowly entering everyday life. More and more often, when we need to have some information quickly, we turn to ChatGPT or Gemini. The Images we produce them with Dall-E, while we talk with the native virtual assistants of our smartphones or with company chatbots, an increasingly less frustrating experience.
Even in agriculture, artificial intelligence has its advantages potential to change the way farmers work in the field and this was discussed during the World Agri-Tech Innovation Summit 2024 Of San Franciscoof which AgroNews® is mediapartner. Two days of meetings and discussions which saw representatives of the major companies and research bodies active in the sector take to the stage.
As recalled by Elliot GrantCEO of Mineral (the company of Alphabetalias Google, dedicated to the agricultural world), when talking about artificial intelligence we must first of all distinguish two major applications. THE Llm, Large Language Model, which have the task of interacting naturally with the farmer, supporting him in making decisions or finding information. And the machine/deep learningused instead to analyze data, in order to categorize it, find relationships and extract valuable information.
Google’s Matt Hancher explained Google’s potential for managing georeferenced data
(Photo source: Tommaso Cinquemani – AgroNotizie®)
As discussed by Elizabeth Fastiggi (AWS), Jeremy Williams (Bayer), Feroz Sheikh (Syngenta) and by Grant himself, the challenges to face they are not few.
First of all there is the issue of data quality. Creating models that rely on data found online leads to disaster, as it is poor quality data. Large non-agro companies, like MicrosoftGoogle, AWS, etc., which have large skills in the IT field but scarce in the agronomic one, I am therefore looking for valuable datasetwhich are often the prerogative of research centers and agribusiness companies.
Then there is the issue of division into silos. Today, many of the data that make up knowledge in the agricultural sector are divided into watertight compartments, not interoperable. Machine manufacturers have their data, as do those who develop Decision Support Systems (DSS) or management software. The challenge is to integrate all this data and make it available to analysis algorithms, in order to extract value for the farmer and above all provide a unique interface to whom you can ask any question, from a fertilization plan to predictive maintenance of the tractor, up to the analysis of the company’s income statement.
What then raises some doubts is the phenomenon “black box”. Artificial intelligence algorithms operate according to logics that are not fully understood understandable from the human being. Therefore, the results they arrive at, which often turn out to be truthful, are obtained through unknowable processes. And this aspect leaves many operators perplexed. In fact, it is difficult to argue the validity of certain indications to the farmer without being able to explain the reasons.
Eric Horvitz, Microsoft’s chief scientific officer, illustrated the potential of artificial intelligence for scientific research
(Photo source: Tommaso Cinquemani – AgroNotizie®)
For others an additional problem is represented by bias which could hide within the datasets fed to the algorithms. However, this is a problem data quality and of how it is generated. If the data is corrupted at the source, the result can only be distorted. Ergo, even in the case of artificial intelligence it is necessary to carefully evaluate the origin and quality of the data.
Finally, an element that perplexes many of the speakers who took turns on the stage of the World Agri-Tech concerns theadoption. Today there are a lot of well-established technologies, at affordable prices and with smart interfaces but they are not adopted. The reason? For many the theme is of cultural character. Farmers, often at an advanced age, are reluctant to adopt technologies whose usefulness they do not perceive, which they consider unnecessarily expensive and complex to use.
One wonders what the reception will be when the first model designed to solve the problems arrives bureaucratic problems linked to the Common Agricultural Policy (CAP), to the Regional Complements for Rural Development (CSR), to the declarations toRevenue Agency and so on. In short, when artificial intelligence will be able to untangle the tangle of rules that bind farmers and about which farmers know little. So will she still be perceived as useless?
AgroNews® is a registered trademark of Image Line® Single-member LLC
