Tel-Hai Magazine 2022

came the joint research goal of Prof. Shir and Dr. Gamrasni. The sequential experimentation cam paign took place at MIGAL’s Postharvest Center throughout 2021, and involved the extensive efforts of both the Gamras ni laboratory team and of members of the Shir computational group. Once har vested, fresh cucumbers were delivered to the laboratory, measured and then treated according to the detailed instruc tions of the AI-driven protocol. They were placed in storage for 4 weeks, and then evaluated. The success-rate of the AI-driven treatment was evaluated as the degree of minimizing their post-4 weeks quality-loss, which was manifested as col or deviation, as well as a reduction in both mass and stiffness. Overall, the AI obtained satisfying results by locating a diverse set of protocols, which outperformed the best-known practices, including the so-called ‘in house protocol’. Some of the protocols obtained possess a surprising nature that will require fundamental postharvest re search. Furthermore, protocols that proved successful upon evaluation, post 4-weeks, were placed back in storage for an extended period of time (for a total of nine weeks altogether). The figure below presents a gallery of an intermediate stage. The cucumbers treated by the best protocol were placed back in storage at the 4-week milestone for an overall period of 9 weeks. The fruit exhibited a surprising postharvest quality, while the cucumbers treated by the ‘in-house protocol’ ended up com

pletely rotten (photography is excluded). To the best of our knowledge, such a postharvest accomplishment for cucum bers has not been reported elsewhere. In the future, as more experiments are AI-guided, as reported here, the roles of scientists and engineers will shift from locating solutions and designs to explain ing the nature of the results attained, while aiming for mechanistic under standing. In this respect, it is a dream coming to fruition for the young genera tion of scientists. As for food security, the current campaign constitutes another successful step in the effort to optimize the entire Farm-to-Fork process. Notably, climate change has a strong impact on food loss, and, therefore, developing new and efficient methods to improve food security is much needed. Our innovative AI-driven method can significantly im prove fast and efficient development of postharvest protocols for a variety of crops, and it has the potential to contrib ute to the global efforts to prevent hun ger. This joint research has been supported by the Ministry of Science and Technol ogy, and also by an internal grant from the MIGAL Institute. Extensions of this work have also been approved for fund ing by the Ministry of Agriculture and Rural Development. 1 https://ec.europa.eu/food/horizontal-top ics/farm-fork-strategy_en 2 https://doi.org/10.1146/annurev-envi ron-101718-033228 3 https://postharvest.ucdavis.edu

A broadly accepted hypothesis is that AI will drive more decisions in future scientific activities. Is this a scientist’s worst nightmare, or a dream come true?

Figure 1 (from left to right): untreated fruit, post-4-weeks; human practiced protocol (‘in-house protocol'), post-4-weeks; AI’s best results, post-4-weeks; AI’s best results, post-9-weeks.

Tel-Hai Magazaine | 2022 13