Artificial intelligence (AI) is already part of our daily lives. Google Maps, Amazon recommendations, Facebook ads, facial recognition, automated translation, answers from voice assistants—all these technologies are powered by some kind of AI.
Machine learning is one of the most promising types of AI. Basically, a computer is coded with an algorithm or model that enables it to continue to learn and improve its predictions with more data overtime. That’s how Google Maps can recommend optimized routes with incredible accuracy.
In the seafood industry, we’ve seen AI used for species recognition (see FishFace), tuna quality inspection (see TunaScope), and electronic monitoring of fishing activity (see SnapIT). But it’s tech startups in aquaculture that are really blazing AI’s trail. AquaByte uses AI to count sea lice and growth rates of salmon. XpertSea uses image recognition to count shrimp and predict growth rates. And Aquaconnect and Aquaponics uses AI to improve shrimp farm management and productivity.
The reason we’ve seen limited application of AI in other parts of the industry, especially seafood processing and supply chains, is due to a lack of digital data. Paper records still dominate the industry.
However, more seafood companies are going digital and those that do could reap significant rewards from the “AI wave.”
Here’s my take on why AI could be a true disrupter:
First, some seafood companies continue to balk at digitization, not entirely convinced on the return on investment (ROI) or worried about staff training and internal capacity. Most companies recognize digitization is inevitable, but some are slow to change. AI is about extracting more value from existing data and so could help drive digitization in the seafood industry by boosting the ROI.
Second, seafood has more natural variability than most food commodities. Species, fish size, seasonality, sexual maturity, catch method, storage time—all these factors affect quality and yields. As a result, it’s often hard to accurately predict production and quality outcomes. That’s a perfect problem for machine learning, whose core strengthen is its ability to take an incomprehensible number of variables to accurately make predictions.
And third, for AI to work, you need clean, comprehensive and complete data. False or inaccurate data will undermine AI’s ability to make good predictions. AI motivates companies to ensure their suppliers are sending them accurate data, thus improving supply chain traceability. Technologies such as blockchain don’t solve the “garbage-in, garbage-out” data dilemma. AI does.
In fact, AI can be used to predict whether data is “dirty” or fraudulent, which is why the FDA has launched a pilot project to use AI to screen seafood imports. In 2019, the FDA screened 15 million food shipments imported into the United States and is using its massive, historic data-set of inspections to better predict which shipments may be problematic or fraudulent and subject to further inspection.
AI could incentivize the good actors and catch the bad ones.
Over the coming months, we’ll be announcing more news about our partnerships and research on AI. If you’d like to learn more or get involved, please reach out.