Interview with Prof. Dr. Frank Ordon, President of the Julius Kühn-Institute (JKI)

Q. How do you think AI (Artificial Intelligence) can help us improve agriculture for better food security? What are the potentials and the risks associated with the application of AI in agriculture?

That’s a good question. Unfortunately, I’m not an expert on artificial intelligence, but on genetics and breeding. But let’s say, artificial intelligence will help us to make plant production more precise in terms of irrigation, of fertilization, of the application of plant protection products and will therefore contribute to an environmental friendly, resource efficient agriculture. To some extent, we are already on the way to this precision agriculture, which is a big issue at the moment. But, production processes can become even more efficient with the help of artificial intelligence.

Q. Are there any risks associated with this application or do we have unlimited potential?

Like in other fields of technologies it is always a question for which purposes AI is used. Risks may occur if AI is handled as a “black box” and we are not able to understand how and why an algorithm comes to certain results. Another issue in my opinion is, that it is always of concern to whom the data which are generated belong and for what purpose the data will be used.  This has to be clarified.

Q. That brings us to our second question about sustainability in agriculture. From a point of view of genomics, how do we address the issue of sustainability in agriculture? Especially in light of climate change.

We are facing many challenges against the background of climate change and in general agriculture has to become more sustainable. This means for the future, that we have to produce food and feed by less water and by applying less fertilizers and plant protection products, but simultaneously have to increase or at least maintain yields. To achieve this, we need crops, which are more resource efficient with respect to water and fertilizers and are more resistant to pests and diseases as well as to abiotic stress. In this respect, breeding is most important. At the end, plants can only develop those traits in the field – interacting with the environment and the respective management – which are genetically fixed in seeds. So seeds are at the very beginning of the production chain and therefore plant breeding is of prime importance for adapting agriculture to future climate scenarios and to improve sustainability.

In this respect one should keep in mind that research to implement machine learning in the selection processes during plant breeding has just started. We now have very big genotypic and phenotypic data sets available. So the implementation of AI in plant breeding may be a next step.

Q. Do you have research on AI in plant breeding at JKI?

No, we are not conducting research on AI in plant breeding yet but next year in Germany a project on wheat will start to implement big data in the selection process. This project will be funded by the Federal Ministry of Food and Agriculture.

Nevertheless, we are working on the use of AI in the area of weed detection for example. The aim is to practice a side specific treatment of weeds in order to save herbicides in the future. This is done against the background of risk reduction and the increase of sustainability in agriculture. Further examples are robotic solutions for different agricultural cultivation processes like seeding, weeding, fertilization and spraying and for the purpose of phenotyping. Also in the field of monitoring, JKI is using methods from AI in order to gather information about plant development, diseases and pests as well as the change in biodiversity in agricultural landscapes.

Q. Do you already have a few plans for pilot testing?

Actually, we are working on an autonomous robot concept for non-chemical slug control for example. Slugs are quite deleterious to some plants like rape seed seedlings and may cause severe economic damages to farmers. Within the robot concept, the slugs are detected with sensors. After detection the robot saves their position and cuts them into pieces. Some slugs show cannibalistic behavior and are attracted by their dead fellows. The robot implements this in its strategy and comes back to the saved positions after some time in order to control the fellows. This is also to some extent a self-learning process. In addition, we can learn from the robot generated maps which parts of the field are more prone to slugs and under which conditions the slugs will most probably appear. In the end, we can use this specific information to continuously improve the control strategy.

Q. What do you think about vertical farming?

Let’s say for a city like Delhi, for example, there may be problems to bring fresh vegetables from the farms to the centre and therefore to some extent in these, let’s call them megacities, vertical farming of high priced products may really be an option. But, we will most likely not produce wheat, rice or maize, the crops which feed the world, in vertical farming. In my opinion for high value products like vegetables or herbs and so on, vertical farming may be an option.