The insect farmers turning to AI to help lower costs
- Published
Felix Collins' job would not be suitable for anyone who is squeamish about insects.
The founder and boss of a company called Full Circle Biotechnology, he is in charge of 20 million black soldier fly larvae.
Based at a small, indoor facility on the outskirts of Bangkok, the firm rears the insects to produce animal feed for the country's shrimp and pig farms.
The larvae, which live in a dark, warm and humid environment, feed on fruit and vegetable waste sourced from food and drink manufacturers, before being harvested and combined with probiotic bacteria and mushrooms.
Mr Collins says the product is a more environmentally-friendly alternative to the soybean-based animal feeds that dominate the market. The latter have been repeatedly linked to deforestation in South America.
He also claims that the firm's insect-based feed has a carbon footprint that is "100 times lower than that of soymeal".
Studies agree that insect-based feed generally does have a lower carbon footprint, but only if the larvae have been fed natural food waste. If the insects have been given a processed food source, then one report says that soybean-based feed produces less carbon.
Mr Collins adds that the feed made by Full Circle, contains up to 70% protein, compared with soy, which has less than 50% protein - making the insect feed more filling and nutritious.
Founded in 2019, Full Circle now employs 14 people, and supplies 49 farms across Thailand. It wants to increase this, but faces a hurdle - soybean-based feed is currently substantially cheaper.
Soybean feed is presently around 460 euros ($490; £400) per tonne, compared with €1,400 for insect-based, according to one European comparison earlier this year.
To try to reduce the price of its feed, Full Circle is now turning to AI to help maximise production at a lower cost. To do this it is training an AI system to study all available past and present data on insect farming to determine and then continuously fine tune the best methods.
This can include everything from temperature to food quantity, the optimum space that the larvae need, quickly and accurately counting thousands of flies, and whether to introduce new strains or species.
"AI can help us speed up the process of trial and error, helping us overcome obstacles, and develop a thorough enough understanding of insect production, to be reasonably confident that our production is optimised," says Mr Collins.
"AI can help us process, record and understand every attempt at insect farming to date, and expand the potential for a larger and more nutritious crop every time we grow one."
Some 5,000 miles away in Lithuania, insect farm software provider Cogastro is also working on an AI-based system. It currently sells monitoring software that automatically collects data for users to analyse, but the AI upgrade will enable to system to learn, adapt and make changes inside an insect farm for itself.
Cogastro says it is not rushing the launch of the AI, and instead plans to launch it commercially in the next three years.
Its founder and chief executive Mante Sidlauskaite, says she is wary of companies in the space that claim they already have AI systems readily available. "We know from our example that it takes time to develop a layer of software, and we have been here already for five years.
"It has taken us time to work with as many companies as possible globally, to understand the differences and similarities of their processes, so that we would be able to standardise our data models to cover their overlapping needs. So we now have a baseline to build an AI solution.
"But when we see start-ups coming up just yesterday, and then announcing that they have something related to AI, they cannot have an accurate AI model before they start to train it, and to train it, you need to have some data, and to have data you need to have some customers first."
She adds that she is concerned about the overuse of AI as a buzzword, without firms explaining exactly how their AI supposedly works.
Back at Full Circle, it is developing its AI system with a Singapore-based expert in the field called Simon Christofides.
Mr Collins says that as black soldier fly larvae have only been farmed commercially over the past decade and a half, there is still much to learn, and that the use of AI can rapidly speed up this process.
"Rice farmers are still trying to create the perfect crop after 13,500 years of passed down generational learning," he says. "Black soldier fly larvae have been farmed for 0.1% of the time rice has been grown, and that's reflected in our current understanding."
Mr Collins adds that the firm's strategy will be to use AI to glean insights "no human could ever achieve", by using multiple sensors and crunching millions of data points.
Yet he adds that sometimes you just need to leave the flies to get on with things on their own. Such as the breeding process, whereby mature flies mate on green ribbons laid out in such a way as to account for their territorial tendencies.
"You need to consider their social and behavioural dynamics as animals," says Mr Collins.