women.of.fisheries

Navigating AI in Fisheries

Posted on

We are living in a world that, at times, feels to be moving too fast to process. Like anything else, how we individually respond to this changing world varies. I have been around long enough to live through several major technological advancements – personal computers, the internet, and cell phones to name a few. Every one of these times felt like unchartered territory and a journey I didn’t ask for but one I had to make nonetheless. Today, I can’t imagine life without any of these things. They drastically changed the way I interacted with others and the world, for good and bad. When I look back, I see a similar pattern where some were more resistant to the new technology, others who embraced it wholeheartedly, and many falling somewhere in between. 

So, let’s talk about AI.

There are so many facets of AI to take into consideration when using it (or not) and how we use it if we do.

  • What exactly is AI? 
  • How can I minimize its environmental impact?
  • What are the ethical concerns? 
  • How should I use it responsibly or is that possible?

AI (or artificial intelligence) is an umbrella term that incorporates different kinds of data learning or “training.” The American Association for the Advancement of Science provides a video and overview to help make sense of it all. 

Here, I am going to focus on generative AI (GenAI) tools, such as ChatGPT, Gemini, Copilot, and Claude, that produce text, images, video, and code from user prompts. In full transparency, I am not an expert of AI or these tools. I am simply sharing my personal journey and how I am navigating this new technology. I also want to provide some resources that can help you find your best path forward. 

As a fisheries scientist, I want to harness the power of GenAI in a way that maximizes the benefits while minimizing its environmental impact. 

Much of my journey so far has focused on gaining AI literacy, in other words learning what AI can do, how to use it effectively, and how to evaluate what it produces. With that came a whole new vocabulary that felt a lot like I was reading something from a science fiction novel. Terms like “AI model” (what AI tools use to generate output), “AI slop” (low quality output that has not been edited or checked), “AI-gerism” (AI output used without attribution or containing copyrighted material), and “AI hallucinations” (AI output that is false or made up). Wait, what do you mean that AI can just make up stuff?! The excitement of AI and its capabilities was quickly tempered by the reality that although a tool, it is an imperfect one. It requires us to take individual and collective responsibility to learn about its pitfalls and take deliberate steps to avoid them. This means taking the time to check our AI output and disclose when we use it and how. In my experience, as an example, I have found AI output to be redundant, lack substance, contain unnatural wording and images, and be downright wrong at times. I do the foundational work first and then use AI to fine tune that work where needed, but even then, there is almost always additional editing and checking that need to be done. It is important for each of us to continue building critical skill sets, so that we can become better writers, thinkers, and scientists – and now also better judges of AI output.   

For many, myself included, energy consumption and other environmental impacts are another top concern. Fortunately, there are steps we can take as individuals and more broadly in the AI industry to minimize these impacts. By some estimates, using “green prompts” can reduce energy consumption by as much as 48%. The general premise is to 1) use the least number of prompts; 2) use prompts that are concise, specific, and easy to understand (think plain language over field-specific jargon); and 3) provide necessary context. What tasks you use and what AI models you use to perform those tasks can also make a difference. Text-related tasks require much less energy than image or video generating tasks. Likewise, using specialized AI models designed for specific tasks generally require less energy than general purpose AI models. There are other efforts to help make AI more sustainable, termed “green AI”, that include more efficient software, changing how AI requests are routed, investments in clean energy initiatives, and development of energy consumption ratings for users. 

How we each navigate this new world will vary from person to person based on our values, perspectives, professional roles, and relationships with technology in general. These differences can be pretty wide and polarizing, and that’s okay. Our journeys do not have to look or feel the same. We can praise the ways AI can help make our jobs easier, while also acknowledging the many concerns associated with it. On the other hand, we can choose to advocate for not using AI until it becomes more sustainable and other issues are addressed. Listed at the bottom of this article are some of the benefits, concerns, recommendations, and resources that I have found useful in my journey with AI. I hope they can also help guide you along on your journey as well.

At the end of the day, AI may be able to assist us, but there is no substitute for human creativity, curiosity, and experience for moving fisheries science into the future. We still need YOU.

AI Benefits

  • Can increase efficiency, productivity, and quality of work 
  • Can expand our knowledge on a topic and help identify knowledge gaps
  • Can help with skill building (e.g., coding)

AI Concerns

  • Disclosure rate is low among scientists, even though it is increasingly being used
  • Lack of disclosure can prevent others from evaluating the work and can reduce accountability
  • Can be a “black box” that hinders or prevents reproducibility of methods
  • Can produce biased viewpoints and misleading or incorrect output
  • Overusing or overreliance on it can negatively impact skill development
  • Electricity and water demands can contribute to our warming climate and disrupt local ecosystems

AI Recommendations

  • Understand what AI can do and what its limitations are
  • Adhere to AI guidelines, including those provided by places of employment, funding sources, and publishers
  • Use it to assist in your work, not replace your work
  • You are responsible for the quality and accuracy of AI output, so take steps to check and edit output accordingly
  • When presenting work that was created with the assistance of AI, disclose when and how it was used
  • Use effective prompts that are clear, concise, and specific and use as few prompts as possible
  • Use GenAI sparingly for tasks that involve image or video generation that require much higher energy consumption than text-related tasks
  • If using for image or video generation, use the lowest resolution and shortest duration to meet your needs 
  • Use AI tools that are specific to a task (like text editing or coding) rather than general-purpose AI tools 

Resources

  • AI principles by the Organization for Economic Co-operation and Development (OECD) (2024)