Can AI Feed the World?

There is one subject that is always part of any conversation about Industry 4.0: AI. But where do things stand regarding the use of artificial intelligence in agriculture and agricultural equipment? And can AI feed the world?

Text: Jörg Huthmann

When analyzing images to look for tumors, the detection rate of self-learning computers now beats that of experienced specialists. Social media platforms and search engines use our preferences to deliver tailored advertisements. Carmakers are developing self-driving vehicles. And Hollywood provides alternating scenarios of Terminators that save the world and pose a threat to humanity. So far, the imagination of screenwriters is the only thing behind movie plots, while artificial intelligence is powering everything else.

Springboard innovations

Politicians have now begun to address the subject. In July 2018, the German federal government ­adopted the cornerstones of an AI strategy. It calls for more academic educators and also the estab­lishment of an agency for springboard innovations that primarily deals with artificial intelligence. As a result, Germany – just like its European neighbors – hopes to improve its standing in the global com­petition for the best AI ideas. At the same time, this issue is nothing new, given that there was already talk of artificial intelligence more than 30 years ago. What is new, however, is that all of the technological and economic parameters for using AI have now been met. Computing power and memory are be­coming increasingly cheaper, while networks continue to grow faster and more extensive. What’s more, it is cost-effective to invest in these new technologies and integrate them into machinery that ­already exists or will be developed. They could take over work described as dull, dirty, and dangerous, for example.

Artificial intelligence and robotics

AI and robotics frequently appear as synonymous expressions, even though a robot control system does not necessarily have AI. It would be an accurate description for autonomous robotics systems like the kind currently being developed by universities and start-ups for agricultural uses. But AI is now increasingly replacing work performed by humans in the insurance, banking, and telecom­munications sectors.

Is AI the answer?

The supposed technology issue has therefore even provoked a response from philosophers like Richard David Precht from Germany, who takes a thoroughly critical view of digitalization and AI. Precht does not believe that all key problems can be solved through technology. “Much of the discussion on this matter is nothing more than an attempt to change the direction of the wind with an air pump,” he notes in an interview in the weekly news magazine Der Spiegel. Even though there has been tremendous technological development, attempts to feed humanity have failed despite overproduction.

He brings up a key issue here: Feeding the world seems to be primarily a matter of distribution that must be handled at the political level. Nevertheless, artificial intelligence helps strengthen the irreplaceable basis of global food production, namely the production capacity of farmers. In the long run, after all, farmers will only be able to deal with increasingly extreme weather conditions, dwindling land for cul­tivation, and economic pressure with help from reliable data, which is why researchers and companies like CLAAS are working to develop the agricultural equipment of tomorrow with artificial intelligence.

Science

Simon BlackmoreProfessor and Head of Engineering at Harper Adams University (Newport, UK)

“To feed the world with AI, it would require a worldwide mechanization of agriculture and a global education campaign.”

“To feed the world with AI, it would require a worldwide mechanization of agriculture and a global education campaign.”

“To feed the world with AI, it would require a worldwide mechanization of agriculture and a global education campaign.”

Industrie

Dr. Carsten HoffManaging Director of CLAAS E-Systems GmbH & Co. KG (CES)

“To feed the world with AI, it would require a worldwide mechanization of agriculture and a global education campaign.”

×

Science

Simon Blackmore laughs as I read him an old software development adage over the phone: “A fool with a tool is still a fool.” Blackmore, who is a professor of Robotic Agriculture at Harper Adams University in the United Kingdom, thinks the figure of speech is certainly a useful one. When asked if it also applies to artificial intelligence, he explains that “the effective use of artificial intelligence is comparable to the proper use of a hammer. Used properly, it can be extremely helpful.” However, because AI as a tool is quite complicated and difficult to program, it is currently not good enough to feed the world, Blackmore says. “This would first require a worldwide mechanization of agriculture and, at the same time, a global education campaign.”

The professor not only deals with developments in the field of AI on a scientific level, but also uses them for the development of agricultural robots. Machines can only act autonomously when performing certain tasks by way of AI in the form of self-learning algorithms. Blackmore’s approach is quite disruptive – instead of advancing the current generation of machinery in the form of self-driving equipment or tractor-implement combinations, he prefers to send significantly smaller, specialized robots onto the field. In the future, they could take over tasks such as sowing (drills), crop care and protection, and also harvesting. These machines already exist. Startups supported or assisted by Simon Blackmore, such as Earthrover or the Small Robot Company with their agricultural robots Tom, Dick, and Harry, and an AI named Wilma, are in the process of testing. The core element of AI – the self-learning algorithm – enables these robots to improve their work with increasing experience and to distinguish ripe from unripe crops, for example, or to fight just weeds when it comes to crop protection by using lasers, mechanical means, or precise minimum application of pesticides.

In Professor Blackmore’s scenario, all work would be possible as part of “farming as a service,” with autonomous machines working 24 hours a day, seven days a week, because it is difficult to imagine, for financial reasons alone, that a farmer would replace a larger part of his fleet with robots. Conceivable would be models in which contractors gradually expand their range of equipment and services with robots. However, the question of economies of scale always arises. In large-scale agriculture, high-throughput machines are more efficient than small systems in almost all operations, for example in the vast acreage of North and South America or Russia. The issue of soil conservation through plow-less farming is another point of discussion.

With his vision of the agricultural technology and agriculture of the future, Simon Blackmore succeeds in fulfilling the scientific mission to question existing conditions and rethink them. Meanwhile, these ideas are being tested as experimental vehicles in the field and are already in regular use in horticulture. It remains to be seen what opportunities and risks this will entail for sustainable agriculture and, above all, for the main players: farmers, contractors, and agricultural equipment producers.

×

Industry

Dr. Carsten Hoff is the Managing Director of CLAAS E-Systems GmbH & Co. KG (CES), where Dr. Boris Kettelhoit oversees Advanced Engineering. The two CLAAS management executives also see artificial intelligence as a useful software tool. As Carsten Hoff notes, while there has been talk of artificial intelligence for more than 30 years, AI has only been able to fully realize its potential recently thanks to fast computers and networks. The ability to learn is the key difference between traditional control technology and software based on AI algorithms, which is why the software specialists regard AI as a factor enabling the advancement of CLAAS equipment. For Carsten Hoff and Boris Kettelhoit, the disruption does not come from turning away from existing product lines, but rather in the way software is programmed and making CLAAS equipment smarter. One prime example of this is CEMOS AUTOMATIC, an assistance system for combine harvesters. The artificial intelligence used in the system learns from the constantly changing harvest conditions over the course of the day and adjusts the threshing. The ability to learn independently and to come up with ideas for what action to take is the key feature of AI. The role of the operator is already limited today to monitoring the automatic settings and the machine process.

The question of self-driving machinery of course arises with regard to combines, forage harvesters, and tractors. Carsten Hoff cites a figure from the automotive industry, according to which properly equipped vehicles will generate up to 20 terabytes of raw data per hour in the future. Data from images of the vehicle’s surroundings will account for most of it. Analyzing the relevant details from this mountain of information in real time in order to generate a command to brake, for example, can only work with the help of AI. Hoff and Kettelhoit note, however, that some problems still need to be resolved, especially when it comes to image recognition. Tasks that are easy for humans, such as identifying a partially concealed hindrance, still pose a challenge to AI systems.

At CES, the team is focusing above all on the robustness of the systems developed there. “We have to be in control of the technology – in every respect,” Carsten Hoff says, capturing this idea in a straightforward way. “If AI learns and then makes decisions, we have to be able to understand those decisions,” he adds. When asked if AI can feed the world, both men point to the proper sequence: Manual labor is followed by mechanization and then automation. The use of AI in agricultural equipment only comes after that. In the meantime, artificial intelligence has begun trading in foodstuffs on the world’s commodity futures exchanges, but whether this serves the interests of feeding the world is another matter altogether.

More articles