In a few weeks, I’m starting another course on AI and creative work. I already know that some of the people taking it are enthusiastic about AI, while others are strongly opposed to it.

AI has become a significant part of my professional life. I run a company built around AI, I teach it, experiment with it and spend a lot of time thinking about how it will change the way we work. But the closer I get to the technology, the less interested I am in being either “for” or “against” it.

After reading about Jacob Coxon, the researcher who recently resigned from Anthropic after previously working at OpenAI, the feeling grow even stronger. Coxon has raised serious concerns about the race towards increasingly capable AI systems and whether the companies developing them are taking the risks seriously enough.

His warnings concern frontier AI and possible future systems, not the tools most of us use at work today. And there is considerable debate about how likely the most extreme scenarios actually are.

But I find it significant that some of the people asking the hardest questions about AI are people who have been deeply involved in building it.

The more I work with AI, the harder it is to have a simple opinion about it

When you first start experimenting with generative AI, it’s easy to become fascinated by capability. A task that took three hours suddenly takes twenty minutes. Something you couldn't do yourself becomes possible.

When you work with it every day, you also start noticing other things. How confidently it can be wrong. How easily we trust something because it sounds convincing. How quickly “we can automate this” becomes “we should automate this” without much discussion in between.

I can be fascinated by what AI makes possible and concerned about where its development is heading at the same time.

I don't think those positions contradict each other.

AI literacy should include criticism

This is why I’m actually looking forward to having people in my upcoming course who don't necessarily like AI.

Teaching AI shouldn't be about converting skeptics into enthusiasts.

Someone who is concerned about what AI might do to creative professions, copyright, employment, power or the environment has legitimate reasons to question it. But I also think there is value in understanding the technology you are criticizing.

The same responsibility applies to those of us who are excited about it.

I used to think AI literacy primarily meant understanding what these systems can do and learning how to use them well. I increasingly think that's only half of it.

AI literacy should also mean understanding their limitations, questioning the incentives behind their development and being able to make informed choices about where we do — and don't — want to use them.

We need the skeptics in the room

The development of AI is too important to become a conversation only between people who are convinced it will transform everything for the better and people who want nothing to do with it.

If everyone who is uncomfortable with AI decides that it “isn't for them”, we risk losing some of the voices that should be part of deciding where this technology goes next.

Perhaps some of the most interesting people in my upcoming classroom will therefore be the ones who arrive not wanting to use AI at all.

Not because I want to change their minds. They may leave still opposed to parts of this development.

But hopefully all of us leave with a more informed and more complex view than the one we entered with.