Getty Images

How artificial intelligence is playing a role in insect farming

An AI engine developed by Singapore startup EntoVerse is helping cricket farmers improve yield by optimising environmental and other conditions

Edible insects such as crickets require less feed and emit less greenhouse gases such as methane and carbon dioxide, making them a fast-growing source of alternative protein to feed the world’s population while minimising environmental impact.

While cricket farms have sprouted all over the world – from North America and Europe to Southeast Asia – to capitalise on the cricket boom, not all farms are run efficiently due to knowledge gaps in farming the insects at scale.

At Vietnam’s Cricket One, for example, about 25% of crickets could die due to unqualified labour and the challenges with maintaining control over millions of crickets, according to Dmitry Mikhailov, co-founder and chief technology officer of EntoVerse, a supplier of artificial intelligence (AI)-based systems for managing insect farms.

Mikhailov, an associate professor at the National University of Singapore, and his team at the Singapore startup have been collecting data on crickets and the factors that could affect the insects’ well-being and behaviour.

“These creatures move very fast; they change colours and there is a differentiation between girls and boys, so we use cameras to do image recognition and detect patterns,” he said. “We also use microphones to decipher how crickets communicate with each other.”

Using this data, EntoVerse can help cricket farmers optimise environmental and other conditions, avoiding situations where the insects can be stressed by the lack of food, weather changes or even the sex mix in a cricket colony. It also collects data on insect feed to determine what works best for the health of the crickets.

All of that data is fed into an AI engine, which has identified certain correlations that were not discovered until now. “For example, crickets will stop mating when they sense a typhoon approaching from 100km away, but nobody knew there was such a correlation,” said Mikhailov, who has a PhD in big data and AI.

Read more about AI in APAC

For larger farms with automation capabilities, such as those in Europe, Mikhailov said the data and insights can be used to automatically adjust temperature, humidity, ventilation and other parameters.

Smaller farms, like those in Southeast Asian countries such as Thailand and Indonesia, can also use the same insights to optimise processes and reduce human error, he added.

Having tasted success with crickets, EntoVerse is planning to develop AI systems that can support farming of other insects such as grasshoppers and mealworms, as well as black soldier flies whose larvae can turn organic waste into animal feed.

Responding to concerns from customers about sending insect-related data across borders, EntoVerse has started looking into federated AI, a neural network architecture that makes it possible to train AI models using local data. “We basically send small AI modules to the servers of our clients and the modules come back without the data,” said Mikhailov.

With more laboratories conducting genetics research on insects and modifying genomes in some cases to develop bigger crickets and black soldier flies that are resistant to low temperatures, Mikhailov said EntoVerse’s AI engine can also help to supervise genetic modifications and other tests.

In addition, to speed up the growth of crickets, EntoVerse took a mathematical model used to make targeted pharmaceutical products to develop additive recipes for cricket feed. Mikhailov said the company has completed tests with Indonesia’s Bandung Institute of Technology and started offering the capability to its clients.

Mikhailov hopes EntoVerse’s technology can help to increase the employment of women on insect farms by providing training and assistance through AI, and to help its clients reduce greenhouse gas emissions. “We are focused on our attempts to be a reliable partner for measuring how beneficial you are to nature by lowering emissions, but you have to measure it on a constant basis and prove that you are really doing it,” he said.

Read more on Artificial intelligence, automation and robotics

Data Center
Data Management