Efficiency My Arse

Lately every time I read about AI it's in the context of efficiency. Like someone's made another new button that lets you do a 10% worse job 10% faster than before. And as there's money to be made in this new trade-off we're asked to accept that it's some kind of moral imperative to press that new button as often as possible regardless of how contextually appropriate that trade-off is. I'm not buying it.

An annoying problem I have is that I remember things. One thing I remember is that five years ago, before all this, mob programming was all the rage. Under that trend, we were asked to trade away efficiency in exchange for better sociotechnical outcomes. The goal was knowledge sharing, better code, and a shared sense of ownership.

I think my memory of this trend is still so vivid because it could be annoying in its own way at times. For one, I've always been one of those irritating speed demons who like to have three tasks on the go at once and use a million keyboard shortcuts to avoid wasting a moment. Sitting and watching someone use their mouse to select a line of code can admittedly challenge my patience.

Plus, it was super commonplace for mob programming to fail at its stated objectives. Unless you're utterly determined and meticulous about facilitating it properly, it rapidly degenerates into a disfunctional dynamic where several participants are sidelined for hours at a time as a passive audience to a dominant in-group. That out-group's engagement approaches zero. Paying the full price of the increased engineering time and capturing almost none of the sociotechnical upside was what opened my eyes to the reality of how much mob programming was purely trend-driven. The technique had become an end in itself and the outcomes were now secondary.

Now that the polar opposite – speed at the expense of understanding – has become trendy all I see is a mirror image of the mob programming fad. It's become so trendy to chase efficiency using the one specific technique of generating code using AI that the technique has become an end in itself. It doesn't matter if the efficiency goals are reached because the real driving force has become FOMO.

Another memory I'm sadly unable to shake is that the term "AI" used to refer to an entire field of computer science that existed prior to ChatGPT. And apart from large language models, AI's other big contribution to society has been recommendation systems.

So I think about all the hours each day that people spend scrolling TikTok, and it makes the idea that AI is about efficiency difficult for me to swallow. What exactly is so "efficient" about a social media addiction that swallows people's entire evenings? If AI is tricking people into living in a neverending lockdown instead of going out and living meaningful lives, then is it even a net positive for the economy? Or does "the economy" really just mean "rich people's bank balances"?

From 9am to 5pm, AI is all about making you more efficient, and then from 5pm until whenever you pass out on your sofa with your phone still illuminating your sleeping face, it's about making you do as little as possible. And because there's money to be made on both sides of that contradiction, we're asked to accept it as if it were a coherent plan for society.

All makes me wonder what lessons can I apply to the AI trend from any of this. One from the mob programming trend is about that disfunctional in-group dynamic. I think it arises so frequently because the people with the most social standing within a given team tend to find themselves in the in-group where the disfunction is less apparent. So I reckon the relative social status of each organisation's AI maximalists and skeptics probably has a big effect on what working there looks like right now.

Another conclusion might just be that tech is incredibly trend-driven. That can be pretty annoying and sometimes I'm even nostalgic for the love that went into some of the detail work when TDD was trendy or when TypeScript was new and people were still excited about it. But a possible upside is that, for all we know, we might only be 18 months out from some YouTuber with a zillion followers coining a hot new term for using a text editor to write code and flipping the entire trend on its head.

Fingers crossed.