PR and the Five AI Roles
AI owner, model architect, engineer, technician or training data. PR can get a piece of every one.
(This is inside-baseball stuff related to the future of PR and communications... skip it if that’s not your bag.)
Snipped from a Hacker News comment related to Open AI’s new release: “We’re moving toward a world where every job will be modeled, and you’ll either be an AI owner, a model architect, an agent/hardware engineer, a technician, or just.. training data.”
I know many folks think big swaths of PR are—and will remain—immune to AI. But as a thought exercise, I can see the industry getting a piece of every one of these roles:
AI owner: Highly technical or specific domains will absolutely require custom-trained models; a smart domain-focused PR shop could train those models and build a business (or at least a profitable line item on the invoices) out of renting it to clients. Probably makes less sense right now because the rate of iteration is too fast to easily keep up.
Model architect: People are modeling the work of PR professionals into a tailored AI—I don’t have to guess at that; I know folks doing it. The more interesting question: Why aren’t PR people/shops modeling the behavior of journalists and influencers into AIs? Decades ago, we needed a hit in the NYT for a certain level of shock-and-awe coverage. Today? The right model and API framework could launch 20 niche Substacks, cross-pollinate them, promote them via earned and paid channels on social, and likely get the same impact or more. And we could do it with more data-driven certainty about our process and results. Ethical? Not everyone will think so. Is that gonna stop the train? Nope.
Agent/hardware engineer and/or technician: This is where a lot of individuals and shops are experimenting right now, futzing around with how to make AI work for specific clients and highly individualized/quirky workflows. It’s a start... but it’s not a way to build a business with a moat around it.
Training data: This is the one I never hear people talking about. The world is awash in new AI models, many of them open-source. How long before some PR person figures out they can impact the output of those models by flooding the zone with biased training data? As with the model-architect issue, it’s something many folks would find unethical. But the sheer amplification it would offer is too tempting a target to think someone won’t do it.
Greg W. Brooks. First published on Facebook, May 17, 2024.