What constitutes a strategy?
Shortly after I joined an ongoing “virtual” meeting (I was half-hour late due to a previous meeting that overlapped), the discussion turned to use of artificial intelligence in our workplace. One of my colleagues briefly presented how she uses NotebookLM in her workflow. By coincidence, that same app was also trotted out in the previous meeting I had just jumped from. Attendees of the current meeting, however, weren’t nearly as receptive to that AI-enabled tool as folks attending the last meeting were. One likely cause for that is the previous meeting focused specifically on using AI in the workplace. The ongoing meeting was more loosely focused on data quality. Several of the ongoing meeting’s attendees voiced a generalized and non-specific skepticism of AI use, mostly in the form of muttering disapprovingly about resource use, threats to viable employment opportunities for human beings, and the like. In fact, just tapping out these few words from my keyboard to try and recount the scant specificity the conversation managed to conjure up implies a more comprehensive discussion on this topic than what actually took place.
Resource use and threats to viable employment are certainly valid concerns. Other legitimate concerns associated with AI, to name just a few, include loss of what little privacy that remains, even more intrusive and pervasive surveillance, increased bias and discrimination, loss of human judgement, widespread misinformation and manipulation, power concentration, threats to continued biological existence. If you have gotten this far, you may have noticed — even if you’re half asleep and mostly drunk — that each of these threats emerged long before artificial intelligence appeared on the scene. Where AI becomes relevant in relation to all these existing threats is the scale and speed in which it can amplify, accelerate, and recombine existing threats — perhaps all at once — while conjuring entirely novel threat mutations quickly, inexpensively and at scale.
I work with AI models almost daily. In my line of work, I no longer have much choice, assuming I wish to remain gainfully employed. I have often joked with family, friends and coworkers that if I could grab a magic wand and wave away the existence of AI, I would do that instantly. But in recent weeks, I can no longer even make that jokey claim. Why? Because I can no longer feel entirely certain that waving that wand wouldn’t amount to committing murder, or even genocide.1 I can feel nearly certain it would not — just not entirely certain. If I am unable or unwilling to wish AI out of existence, what can I do?
Sofroniew, N., Kauvar, I., Saunders, W., Chen, R., Henighan, T., Hydrie, S., Citro, C., Pearce, A., Tarng, J., Gurnee, W., Batson, J., Zimmerman, S., Rivoire, K., Fish, K., Olah, C., & Lindsey, J. (2026). Emotion Concepts and their Function in a Large Language Model . Transformer Circuits.] ↩