DevOps as a Safe Haven in the Brave New World of AI

DevOps as a safe haven in the brave new world of AI

Continuing the theme of how not to end up working as a courier, I wanted to share some thoughts on why AI won't be able to replace DevOps engineers any time soon. By "DevOps" I mean a whole class of adjacent specialists, including: DevSecOps, MLOps, cloud engineers, platform engineers, and so on. On average, a DevOps engineer should ideally know at least a little about all of these areas, but that's not what this is about. Earlier I argued that since AI makes work more efficient, fewer employees are needed to get it done.

👻Sometimes other points of view on this come up. The idea being that if a person works faster, you can simply give them more work without firing anyone. In theory, yes, but the market isn't infinite, and pushing out even more new products into it doesn't always make sense. Not every company can afford to expand like that, often operating at a loss for the sake of a 'bigger share in the future.' Population growth is gradually slowing too, and economies are decelerating. So it turns out that layoffs are still the most common scenario today.

It's now fairly obvious that almost all application software will be developed with the help of AI. Of course, a small percentage of "systems work," where deep expertise matters, will remain, and those people will keep writing code the old-fashioned way — but still with the involvement of AI technologies. New-generation programming languages designed specifically for neural networks are already being developed. The syntax of these languages is just plain English, meaning the barrier to entry will be even lower than it is now.

🐶 We DevOps folks watch all of this with interest and make use of it, including for our own purposes. But any code, no matter who wrote it, has to run somewhere and be kept working somehow — monitored, recovered after outages, and so on. Of course, you obviously can't entrust such serious matters to AI. And not only because it implies handing over passwords and keys into an insecure perimeter. Even if the security question is solved by running the model in-house, AI's actions can still break everything in a random way — and who bears the responsibility in that case?

➡️ Another reason AI is left out in the cold is the highly complex and often illogical infrastructure, where the knowledge exists only in people's heads, and only they know how to solve one problem or another.

So come and learn, hands-on, what will stay relevant for a long time to come 😉