The significant contribution of artificial intelligence to science (and agriculture)

In 2013, the late professor experienced a strange form of rapid drying of an olive tree in Puglia. John Martelli At first he asked himself a fateful question, citing his intuition and his ability as a scientist to connect the disparate information he had learned during his long career: What if it was Xylella? Together with other experts, such as Maria Saponari and Donato Boscia, he had to suffer attacks and defamations of all kinds, even threats and investigations; But they were rightand the idiots who attacked and in some cases are still attacking or riding ridiculous conspiracy theories, without even apologizing and showing their level of humanity and knowledge in this too, nevertheless ended up in oblivion, where they hide to avoid reminding everyone that they brought with their memes are the main help of genes Bacteria, which have turned large swathes of Salento’s magnificent olive groves into a graveyard.

Now, a new work that, in addition to reminding me of the story, poses an interesting question: What would happen if the primary early diagnosis of plant pathology could be impersonal, that is, it could be attributed not to a person, but to a machine that, with an artificial intelligence system, could provide the first probabilistic indicator to be investigated later? In addition to the fact that it would be more difficult for researchers to suspect a conspiracy, there would be several advantages in combining such a tool with human experts: the potential for widespread surveillance, to begin with, given that sensors could multiply more. easily from the heads of Martelli, Boscia, and Saponari, and secondly also the possibility of detecting signals not immediately perceptible to the human senses, for example by examining chemical signals and the appearance of plants at unseen frequencies. work you mentioned just published, is based on the use of a new type of neural network applied to extract data from images of controlled plants. In particular, theAnd neural networks, which are able to correlate specific “patterns” in data with specific diseases, have yielded promising results in the classification of plant diseases.. However, the traditional methods used to date require prior training to obtain better predictions, and are therefore particularly vulnerable to the lack of adequate training data. Moreover, it is possible to collect good data to train neural networks on in controlled environments, but it is not trivial in the real world. In the field, diseases may be rare or not readily noticed, or, as in the case of Xylella, they may be relatively poorly documented in a new host plant. For these reasons, to date, automatic plant disease classifiers based on neural networks have proven limited in their usefulness when used on real data, and have not been included in the set used for training.

Now, a group of researchers, authors of the cited work, has developed a new, relatively simple neural network called a “multi-representation subnetwork to adapt uncertainty regulation for plant disease classification across species” (Msun), which apparently It can accurately classify plant diseases in natural environments. To do this, they implemented a technique called Unsupervised Field Adaptation (UDA), which allows models that AI learned during lab training to adapt to the situation in the field, without the need for human supervision. This technique made it possible to overcome the difficulties related to the complexity of the images collected in the field, such as the presence of many leaves, unusual camera angles, noise, and other inconveniences related to obtaining photographic images. Furthermore, this technique has proven to be robust against the simultaneous presence of several diseases in the same plant, documented in a single image, as well as for the relative similarity of symptoms produced by different pathogens. Ultimately, using huge datasets of plant pathological images collected in the field around the world, the research group that authored the work in question was able to show how The accuracy and accuracy of the obtained diagnostics were similar to those obtained during in vitro image training, thus surpassing any other autodiagnostic method available to date.. If the work of these scientists proves valid and is confirmed by the analysis of other independent groups, very interesting scenarios will open up, some of which immediately come to mind.

First, the same type of technique can be used to analyze images and types other than simple visual images, increasing the amount of information available for more accurate discrimination. Moreover, rather than limiting itself to analyzing leaves of single plants, one could attempt to use the same method by applying it to the processing of multispectral images obtained from drones or aircraft, at least in certain cases of stress and pathology that can discover it. In large areas cultivated, such as orchards, olive groves, vineyards and other types of cultivated areas. Collectively, assistance from AI could give plant pathologists the data-driven power that, when supervised by a human, has been in other areas that can greatly improve our ability to respond quickly, accurately and effectively.; And perhaps, other than the fact that hot-headed conspiracy theorists will anyway fantasize about a new con against them, it may be that the next environmental catastrophe associated with some minor pathogen or climate variable can be better addressed, thanks to precision farming methods supported by modern research in intelligence. artificial.