Artificial Intelligence, Artificial Productivity: A Mismatch Made in Corporate America

Productivity—it’s quite the buzzword these days. Popular podcasts are built entirely around the topic, and the corporate world has gone mad trying to measure it in its employees. That is a large reason why people use AI—it has been touted as being able to increase productivity. But because corporate America and AI companies don’t seem to fully grasp what productivity is, AI, the supposed productivity tool, is frequently counterproductive. Corporate Understanding I personally learned about the dangers of misunderstanding productivity while working as a pleb in a corporate office supply store. I always found it odd that we were supposed to find customers and pester them into accepting our help—certainly not how I like to be treated when I shop. But one day the company came up with a super amazing new training where we were supposed to leave customers alone unless they signaled they wanted help. What a concept! Until that point, the visible production of following customers around the store and trying to help them was seen as superior to simply monitoring whether they were in need of assistance. Essentially, because we weren’t actively involved with customers, we were viewed as unproductive. I left not too long after that implementation, so I’m not quite sure what effect it had, but at least in that particular area, a better understanding of productivity had been implemented. The big lesson to come out of that experience: productivity does not equal activity. Productivity means achieving a useful result, and sometimes that means leaving people the hell alone. The thing is, it appears AI companies have the same understanding of productivity as my former employer did before the changes. This captures the feeling one can get interacting with AI Another misunderstanding of productivity came up when I called a friend the other day and he used his AI assistant to screen the call. It initially asked if it was an emergency. I replied no, then waited for two minutes in silence before hanging up and just texting my friend to contact me whenever. To be clear, it’s not his fault; he’s trying to use a productivity tool. In this case, the tool simply created more unnecessary work for the caller. I can write off the two minutes of silence as a technical glitch, whether that’s true or not, but the broader problem remains: voicemail or text was already a much easier communication process before the supposed productivity tool got involved. The next case, however, is a clear misunderstanding of productivity. Some of my pastimes include thinking and writing, so I primarily use AI to help me with my arguments, from postmodern theory to why manual transmissions are morally superior (because it’s evident they are). I use it to push back on my ideas and to help me see holes in my argument or different perspectives I might be missing. It certainly has its limitations, but overall it has been fairly useful, if only because it makes me properly articulate my ideas. But recently I’ve had some productivity issues with a few LLMs. The problem arises when I give an LLM my argument and ask for feedback—it always seems to find an issue. And fair enough, because that’s what it was asked to do if it found any. The problem is when there is no issue, it strawmans the argument in order to critique it instead of simply saying no problems as instructed. But because these are my ideas and I have the intellectual maturity of a five-year-old, it entices me into defending something I never actually said, taking me down a rabbit hole I should know better than to go down. So why do LLMs do that? A clearly frustrated LLM user Why Baby Why My suspicion is that short, positive feedback is treated as inferior to detailed, negative feedback. Positive feedback is frequently very short, and even detailed positive feedback is typically shorter than negative feedback. LLMs often behave as if short, good feedback is not very productive and detailed, negative feedback is, even when it’s not what the prompter asked for. So when asked to find problems only if problems exist, LLMs manufacture problems because visible critique looks more productive than restrained confirmation. What’s going on is simple: when the metric equates productivity with output, even garbage counts. So misunderstanding what productivity actually entails? Mission accomplished! If corporate America and AI companies both misunderstand productivity, large corporate AI companies like Anthropic and OpenAI compound the problem while selling productivity tools. The mirror of artificial intelligence, it seems, reflects corporate interests rather than humanity. I’m not quite sure what else we should expect—these companies are valued near a trillion dollars each. On second thought, maybe the mirror does reflect at least one side of humanity…just not the side we like seeing reflected.