Generative AI is here. There’s no way around that. It is affecting a lot of professions out there, and the list will only grow. At the same time, the examples of AI being used to the detriment of – at best – the user (and, at worst, everyone) seems to grow on a regular basis. There are many reasons for this. From my perspective, the single biggest contributing factor is that people don’t understand what can and can not be delegated.
Over twenty years ago, I served as a communications and cryptography specialist in the Royal Norwegian Navy. Among many other things, this involved assembling and properly formatting a general report that we sent off to all our ships. Whenever this task fell to me, I consistently completed them without error or issue. After completing them, I handed them off to the officer of the watch, who would check and approve my work before I sent it out to the ships.
At one point, I asked why they still needed to check my work, after I had performed this task without issue for six months. The answer taught me a lot about responsibility. It went more or less as follows:
Two reasons: First: By my going over it with fresh eyes, we reduce the risk of critical errors. Second: While I can, and am indeed more than happy to, delegate the task, I absolutely can not delegate the responsibility. Whenever a sailor reporting to me makes a mistake, that mistake is my responsibility.
I have thought a lot about this in the years since we had that conversation. It ties into so many things that I have done in my professional life. What’s more, it ties into my deeply held beliefs about safety culture. By adding in a control and verification loop, we make it safe to make errors in our work. Not because we shouldn’t try to do our jobs to the very best of our ability (and as such must not rely on others to catch our errors), but because making errors is a part of the human experience.
When I use AI, the practical application of this lesson means that I use AI to help advise me, or to perform tasks that I would rather someone – or something – else do. I then take the output, and critically assess it based on my own best judgement and knowledge.
Whether I leave AI to perform the entire task (such as bookkeeping for a club), use it as a sparring partner for writing blog posts, or use it to help me plan my workouts and track my progress, it is an assistant. Double checking the output is on me.
By posting a comment, you consent to our collecting the information you enter. See privacy policy for more information.