The US navy also went through this 50+ years ago. They had the ability to automate nearly everything, but the admiral in charge of the project wisely said that a machine could suggest what to do, but there must always be a human in the loop to actually carry out the action, or we would lose our ability when the machines failed.
The line that got me: "you don't feel it happening." That's the scary part. It doesn't feel like losing a skill. It feels like being productive.
I've been calling the fix "human at the helm, not in the loop." Approving what the AI suggests isn't the same as steering. Your point about human-on-the-side vs. human-in-the-loop puts words to exactly that.
Saving "deliberate inefficiency" for my next team conversation. Thank you for writing this.
"We are definitely not rejecting AI tooling; we are adopting it and integrating it stronger than ever before, because that’s the only way forward."
I think this is the weak point in the article. Is it really the only way forward? If so, why? It bothers me how often I see AI adoption as a foregone conclusion. Must it be so? We have used our brains to write software just fine in the past, writing targeted tooling to increase productivity where needs arise. What prevents us from doing so now?
Also in novel incidents you are not left on your own with just your (rusty) skills. It is still AI plus you, even if AI cannot solve it on its own, you can still iterate during the incident, making observations, using AI to ask questions and dissecting the problem to eliminate root cause candidates. This is how it goes for me at my job. I am much more confident going into a SEV response knowing I have well integrated AI on my side.
Really amazing read. It reminds me of another research, im not quite sure it was that which was addressed in this post but il like to reference it either ways.
I couldn’t agree more… the impact of AI ranges from the Exec saying, “Oh give it to the AI” to engineers telling an AI to do their stuff for them, but there are still vast numbers of people too scared to engage as well. That excellent middle ground you suggest needs to be normalised FAST
The US navy also went through this 50+ years ago. They had the ability to automate nearly everything, but the admiral in charge of the project wisely said that a machine could suggest what to do, but there must always be a human in the loop to actually carry out the action, or we would lose our ability when the machines failed.
This one was a lightbulb moment for me.
The line that got me: "you don't feel it happening." That's the scary part. It doesn't feel like losing a skill. It feels like being productive.
I've been calling the fix "human at the helm, not in the loop." Approving what the AI suggests isn't the same as steering. Your point about human-on-the-side vs. human-in-the-loop puts words to exactly that.
Saving "deliberate inefficiency" for my next team conversation. Thank you for writing this.
"We are definitely not rejecting AI tooling; we are adopting it and integrating it stronger than ever before, because that’s the only way forward."
I think this is the weak point in the article. Is it really the only way forward? If so, why? It bothers me how often I see AI adoption as a foregone conclusion. Must it be so? We have used our brains to write software just fine in the past, writing targeted tooling to increase productivity where needs arise. What prevents us from doing so now?
Also in novel incidents you are not left on your own with just your (rusty) skills. It is still AI plus you, even if AI cannot solve it on its own, you can still iterate during the incident, making observations, using AI to ask questions and dissecting the problem to eliminate root cause candidates. This is how it goes for me at my job. I am much more confident going into a SEV response knowing I have well integrated AI on my side.
No lies detected.
Really amazing read. It reminds me of another research, im not quite sure it was that which was addressed in this post but il like to reference it either ways.
https://arxiv.org/pdf/2601.20245
Good article
I couldn’t agree more… the impact of AI ranges from the Exec saying, “Oh give it to the AI” to engineers telling an AI to do their stuff for them, but there are still vast numbers of people too scared to engage as well. That excellent middle ground you suggest needs to be normalised FAST