Adoption of Artificial Intelligence in the Chilean Salmoniculture

Many companies do not have the capabilities to consolidate, harmonize and update information from sensors - and other data sources - in centralized databases. Thus, much information is not used in the analysis.

👉 see the original column here

Translation to English:

Adoption of AI in the Chilean Salmoniculture

Artificial intelligence or AI (by Artificial Intelligence), in general terms, tries to emulate cognitive processes of human beings, such as problem solving, pattern comparison and creation of new patterns, using machines. The process of learning and self-correction used by artificial intelligence is based on the collection of large volumes of data (Big Data), use of rules to use information and generation of results (which can lead to the generation of new rules) that can lead to one, or a set of actions.

Artificial intelligence can be designed to solve specific problems, usually called weak artificial intelligence. It can also solve complex cognitive processes, for example, solve problems given the learning gained from solving different problems. Usually, this is known as deep artificial intelligence.

The use of robots that carry out specific tasks, the monitoring of crop health through the use of sensors, and the use of machine learning to assess and predict environmental impacts and productive practices in performance, are three areas keys associated with the development of artificial intelligence in food production systems.

An article published this year in the scientific journal Precision Agriculture, revealed that 18 studies reported positive returns given the adoption of automation of processes and/or robotic technologies in agricultural systems in the US. These studies basically focused on costs and returns of investments based on prototypes. The authors of the article emphasized that more analysis based on real situations is required, in addition to estimating integrated and long-term effects, such as changes in cropping systems, market impacts, demand for specialized labor, etc.

The development of new sensor technologies has allowed the collection of large amounts of data. However, commonly the use of these sensors is limited to performing a specific action (e.g., stop feeding when there is a fall of pellets to the seabed) and the information generated is not integrated with the rest of the production chain. An article published this year by McKinsey & Company indicated that 60% of business leaders surveyed acknowledge that the use of sensors can deliver relevant information for their companies, although only 10% state that the information generated by these sensors is used extensively within their companies. Many argued that it is difficult to extract, manage and analyze data in real time. This is consistent with an evaluation made by the same firm to a gas company in the US, where it was reported that managers used only 1% of the data generated by 30,000 sensors to make planning and maintenance decisions.

As a company dedicated to the management and analysis of data in aquaculture systems, we have observed a similar situation in the salmon farming industry in Chile. There is information generated from many sources, many of them sensors located in sea farms, processing plants, etc., but unfortunately it is underutilized. Many companies do not have the capabilities to consolidate, harmonize and update information from sensors - and other data sources - in centralized databases or Data Cloud. Therefore, much information generated is not considered in the analysis.

The application of artificial intelligence in aquaculture must go far beyond the isolated implementation of technologies that help solve a specific problem in the production chain. The generation of integrated analysis of big data is the main value of artificial intelligence. Finding patterns in the data to learn and even reveal hidden information that offers solutions to complex salmon farming problems in Chile is the next challenge for this industry. For this it is key to understand that there are many tools (sensors, production software, etc.) within the production process that generate valuable information, only when these information flows are consolidated and then be correctly and timely analyzed.

Pablo Valdes Donoso, DVM MPVM MS PhD
Pablo Valdes Donoso, DVM MPVM MS PhD
Assistant Professor of AI in Veterinary Medicine

My research interests include using different data sources to address issues regarding health and economics in food production.