When we tell people about Whale Seeker, they’re often surprised to hear that we’re a for-profit company. It begs the question: when it comes to undoing humanity’s past incursions on nature — many of which are the direct result of corporations’ activities — what role do private companies have to play? What are the strengths and limitations of various governance structures in combating these problems?
In our line of work — monitoring whales — academic research institutions are one place where a number of AI initiatives have arisen. These institutions have the advantage of fostering cross-fertilization between biologists and tech experts. But none of the many university-based initiatives have developed into the kind of scalable approach we need to address the scale of the world’s vulnerable whale populations. One reason for this is that software, especially in a field as dynamic as AI, is best thought of as a service, rather than a finished product. The job is far from over once a piece of software is developed: it has to be able to accommodate new and ever-changing data formats, environments and target species. Being a for-profit company gives WS a framework for not only developing, but also maintaining and improving our solutions over time: our employees are laser-focused on the unique intricacies of whale detection — not just for the duration of a research project or grant, but year-in, year-out.
Governments have also been key players on this particular challenge: for example in 2017, when public outcry over a number of collisions with critically endangered North Atlantic Right Whales came to a head, the Canadian Government responded by imposing a slowdown in the Gulf of St. Lawrence, which seems to have been at least somewhat effective in curbing fatalities. However, the government departments responsible for marine wildlife monitoring often lack the technical expertise to not only create, but also maintain, a complex AI pipeline, and usually look to contract out this work.
Nonprofits are another structure that has a useful role to play in the ecosystem. Conservation groups undertake the important work of bringing awareness to environmental issues, lobbying to shift policy, and bridging the gap between scientific research and ecosystem recovery strategies. Environmental nonprofits are often early users of cutting edge technology, but developing new AI tools requires fast, flexible product decisions and financial investment that is often incompatible with the governing board and donations structure of a nonprofit organisation. Technology developed by nonprofits often operate at a loss and depend on the continued influx of fundraising dollars, which is a major barrier to scaling these solutions for widespread use.
One final question we get is why it wouldn’t be more efficient for a larger, domain-general tech company to tackle the problem. To which one response is…they have! A recent collaboration between NOAA and Microsoft yielded a publicly available dataset of over 14,000 seals in the Arctic. But in our last blog, our CTO Antoine Gagne revealed that Microsoft's automatic annotation had missed at least 1,500 seals in the imagery! So while valuable, the involvement of a larger tech company doesn’t seem to have cracked the problem in any definitive way. Partly this is because of the peculiarities of the problem, relative to the AI tasks most companies are used to solving. For example: marine mammals are scarce targets in imagery, appearing in a small fraction of the images collected during a survey and taking up an even smaller fraction of the total pixels at play. This makes acquiring training data difficult — and since every new dataset is different, new training data is needed often. By focusing on this problem for several years, Whale Seeker has developed a Human In The Loop solution, whereby we interleave phases of automatic and manual detection to re-train a model to a new environment using a fraction of the manual annotation time required for conventional AI datasets.
As a private company, Whale Seeker has a unique position within the diverse ecosystem of minds tackling the issue of whale monitoring. Like many other tech startups, we’re run by a lean, adaptable team that can shift with the constantly changing landscape of machine learning. As a for-profit company, we build long term viability into our tools and service models, and provide options for industry clients to adopt more sustainable practices while maintaining or improving their bottom lines.
We believe that a sustainable, ethical business can also be financially viable. As the first AI company in the service of wildlife to be a Certified B Corporation, we are held to the highest standards of responsible social and environmental performance. This includes company policies addressing AI-specific ethical challenges from the very beginning of our operations.
As with any global challenge, the private sector is just one component of the complex ecosystem of effort required to come to a solution — but a vital component nonetheless!