What’s the best way to adapt to AI? A new book offers ideas

As AI tools become more sophisticated, IT professionals find themselves speared on the horns of a dilemma: AI might make them more productive…but it also might automate them right out of a job.

IT Brew sat down with Christopher Mims, tech columnist for the Wall Street Journal, to discuss his new book, How to AI, which delves into the evolution of AI and how it has impacted the modern workplace.

Mims seeds the book with practical “laws” for AI, including “AI is a feature, not a product,” and “Don’t trust it, and always verify its work.”

While his examples of AI integration often focus on people in non-technical roles, much of his advice applies just as effectively to sysadmins, developers, and others busy figuring out how the technology can improve their workflows.

This interview has been condensed and edited for length and clarity.

One cool thing in your book is a comparison between AI and a collective intelligence, like a fungus-based network system that exists under tree roots.

Christopher Mims: I think I compared it to slime molds, actually. It’s really interesting to me because the world is full of potential substrates for computing, and I feel like what these other types of weird intelligence show, or just the huge variety of way nervous systems are built and organized in the animal world, is just that we shouldn’t be surprised that you can create a sensing and reacting system out of all kinds of shit, you know?

I think one of the things that’s interesting to me in how these transformer-based systems work is that, if you look at it a certain way, it resembles a so-called swarm intelligence where you have independent actors that are sort of communicating with each other laterally. So, that was just a fun bit of quasi-speculation I threw in there, really just to emphasize how weird these things are.

There’s a lot of hype on the side of vibe-coding being able to do what expensive CRM packages formerly did, and then a lot of folks doing the building and testing who’re saying, “Oh, this may not live up to the hype.” How do you feel about that?

Everybody always wants to say, “This time it’s different.” It’s always a little bit different. But it’s never fundamentally different, not on the time scale of a few years. So, for the history of technology, over most time horizons, yes, there is displacement. But is this going to nuke all of the SaaS subscriptions you have? Probably not. And there are a bunch of reasons for that, from who owns the data to who can secure this, to who can provision it for your whole company.

The idea that your employees are all just going to vibe code alternatives to every bit of software they use tomorrow? I mean, okay, go for it. I dare you…What is the “yes, and” part?...For early adopters and the people who are developers inside of companies, you’re going to have more, new types of totally custom apps and software that allow you to do things that were previously impossible. In turn, that’s going to put pressure on these cloud service providers to do all kinds of new things.

You talk about how the Skynet idea is really not going to happen, at least not anytime soon. But there are a lot of lower-level cybersecurity concerns around AI, guardrails, data leakage, etcetera. Do people seem pretty sanguine when you’ve been talking to them about it, or freaked out?

In the past few weeks, people are freaking the fuck out, right? Because you can now use Anthropic to, if you’re red-teaming a company’s security infrastructure and you know what you’re doing, and you have AI working alongside you, it will supercharge your ability to penetrate an organization. We’re already seeing it as an adjunct to attacks happening in the wild. This can be a bonanza for cybersecurity companies, because penetrating a network or a company or a piece of software or whatever else requires persistence, and this is something these systems excel at. They don’t have to be very smart as long as they can be tireless.

So, it’s a huge cybersecurity threat. We’re in a very unstable time in the evolution of technology, where the pendulum is going to swing back and forth on different issues, weekly or monthly. People are going to go from “everybody should vibe code” to some huge disaster where somebody vibe-coded something and it was full of insecurities, and they pushed it live…there’s just going to be so many pendulum swings.

Given recent high-profile tech layoffs like at Block, many IT professionals are anxious about AI automating their jobs. How should they future-proof their careers?

A world in which a software developer’s job goes from primarily writing code to specifying software, thinking about system architecture, evaluating code even that’s alongside an AI. That’s for some people a big transition. Others are already kind of operating like this, and it’s a different skill set, and I think people should be thinking through what other skills they have.

I know it’s really brutal out there, and I’m not trying to be a Pollyanna about this, but I think in the long run, the arc of history bends toward more programmers and more IT people, even in a world in which AI is doing most of the actual coding.