The seventh and final installment in the series “Happy Fish – How Aquaculture Operators are Growing Better Fish.”
While we will only be able to scratch the surface of this topic, there is great potential in machine learning and artificial intelligence applied with aquaculture and fish farm systems. We wanted to ask leading industry experts Matt Clarke of Poseidon Ocean Systems, Tim Sjostrom of Kuterra, and Libardo Estupinan of Oxygen Solutions about this application.
To start, we wanted to explore the idea of monitoring fish behaviour. The idea is that if farmers could monitor their fish continuously, problems could be addressed faster and fish could grow with better health. Matt Clarke started the conversation, “Within the industry, there’s a number of start-ups and technology providers we’re working with to apply this technology to our industry. Google is working on developing artificial intelligence to make fish farming more effective and more efficient. For a company like ours, it’s actually very exciting because we’ve been involved in aeration and oxygenation since the beginning of our company and it’s a rapidly growing aspect of aquaculture. But, if you go out on a modern fish farm, what you’re going to see is some incredibly advanced technology on the feed systems: there are remotely operated underwater cameras, there are wireless sensors that are underwater and transmitting data wirelessly back to the central command areas, and we can see offices that look like Mission Control at NASA for farming fish around the world.”
A specific example Clarke dove into referred back to our previous conversation about automated feeding systems, “Tim mention there is feed detection or pallet detection software that communicates with cameras. This allows the computer to pay attention to the amount of feed being eaten. A human operator can now allow a fully automated computer to dial back the feed rates if food isn’t being consumed. Those types of technologies are already starting to be applied- Cermaq, an aquaculture company with operations in Chile, Norway, and Canada, has a pretty public “i-Farm” initiative, where they use photo recognition software to literally identify every individual fish within the farms, from what I understand, and use that to and individualize if every fish is happy.”
The conversation veered back to oxygenation and aeration systems, “Libardo mentioned this earlier,” continued Clarke, “that control integrated aeration systems and oxygenation is really just coming into its own. But for the most part, aeration systems are not integrated into these control systems. What will be integral to the future of these machines will be the ability to apply the aeration and oxygenation systems into that central data stream, giving another variable that can be remotely or automatically controlled. Again, these systems are still controlled by the human farmers using this data to review their farms progress. However, we could have a more automated approach, where we can allow a lot of computer algorithms in the AI that makes decisions for us. An automated process like this would be able to give whoever is in control that ability to control the environment remotely and automatically, whether it’s through AI or machine learning, ultimately giving an advantage to the farmer.”
Though the advancement of machine learning and AI technology will help the farming process, Clarke wanted to emphasize the importance of the people in the industry, “I mentioned finding and training good people takes time. Therefore, if we can help overall farming operations by closing that discrepancy gap with an automatic system or backup who is able to keep the fish happier than ever before, then you could have the real human touch to tweak the system to that optimal efficiency and state of happiness for the fish. This has real economic benefit for the operations and really just translates back into that quality of fish- quality of life for the fish leads to better output. We’re building compressors that had AI and ML (machine learning) built into them so that the compressor can automatically respond to changes in the environment or changes in the fish. Whether the base technology is identifying with a human or computer that determines the fish are not as happy as they could be, the hardware, such as the oxygen generator or the air compressor, will automatically kick in and deliver the right conditions to make those fish or those animals happier. Fish happiness represents a major step forward in our ability to raise these animals better.”
Tim Sjostrom pointed out the importance of the large amount of data that these AI and machine learning elements could gather for the farmer automatically, “You could also have something such as local weather or local forecast streaming into your system.” Matt Clarke agreed to the importance of this element. After all, any variable that that affects or impacts a system could potentially be monitored by AI and machine learning, which would adapt itself and adopt those conditions without human intervention. Clarke and Sjostrom outlined earlier that in the event that there are challenges with staffing, AI and machine learning producing automated systems and processes and machines can resolve issues that would otherwise need a person standing there for 24 hours a day.
Clarke pointed out how 24-hour monitoring could prevent terrible disasters in your production, “You could take this a step further. Even with the staff and the experience people have today, compare that to a harmful algae bloom in an aquaculture situation. In 24 hours, a harmful algae bloom can wipe out the entire biomass of a farm. This is an incredible threat. So, if you have better management programs and remote sensors monsters, they help to predict and reduce those instances. But any event or in a situation where you don’t have access to these systems or if they fail, consider that in 24 hours your staff will only be there for half that time. The other half, they won’t be paying attention as they’ll be going to sleeping or eating dinner. To have a harmful algae bloom roll in at 9:00 PM and the first the first staff member looks out the window or check the check the readings at 7:00 AM, you’re looking at 10 hours unmonitored. That’s 40% of the threat time. So if you can tie AI and ML automated systems into your production, you can have sensors nearby or built in so that if the system recognizes a problem at 9:00 PM and engages alarms. Our systems are designed with an override for the silent mode on your phone, so that it’s going to wake you up in the middle the night. In my mind, that’s what’s necessary – they’re animals that need to be taken care of 24 hours a day. As humans, there’s only so much we can do, so to provide that that capability is really just an obvious next step.”
Check out the video linked below to see the whole conversation.
We would love to hear from you – If you have some AI or machine learning elements applied to our operation, please comment down below. We would love to hear more about how technology simplifies and improves your farm. Moreover, subscribe to our blog to get updates whenever we publish new articles!