|
The problem with AI is that it's just an imitation of intelligence - it's literally just studying what actual intelligence has produced and then using statistical math to parrot it back. That gets you 95% there, or 97%, 99% etc. but never 100% correct and approaching the point uses up exponentially and infinitely more computing power, so people settle at 80-90% correct because it looks promising.
But then, when you put the 80-90% correct solution to work, it doesn't work. In trivial example cases, it's easy to fix the missing 20% and it looks like the AI did the bulk of the work for you, so it actually saves you time - but in real work that last 20% is usually the hardest part and what takes the longest. The bulk 80% that the AI did is just what some summer intern could bang out in a couple days, and the important 20% that the AI got wrong is what takes experts months to solve with or without the AI.
Worse still, it takes people a lot of work and time to learn what the AI has done, understand what it was supposed to be doing, what it did wrong, and then re-doing the broken bits themselves. Babysitting the AI is more work on top of the work you're already doing, and studies find that people spend 10-20% longer to accomplish the same tasks using AI than without.
What makes people report back that AI is more efficient is simple psychology: people don't judge their own efficiency by comparing the minimum effort vs. their own effort, because we don't know what is the minimum required effort. Instead, we think we're being more efficient when we're working harder, even though it's accomplishing less work. The effort we put into something changes the value we place on the outcome.
For instance, people who are used to writing long command line entries really fast feel like they're more efficient than people who do the same thing by clicking a button with the mouse for the same effect - even if it actually takes more time. Likewise, people who spend the extra effort to use an AI to "augment" their work feel more efficient because they're doing extra work, which is a complete illusion of efficiency. They're not comparing the outcome to any measured baseline, because no such thing is usually available, but actually over-valuing the outcome based on the amount of effort they're spending. This is also known as the IKEA effect:
https://en.wikipedia.org/wiki/IKEA_effect
>A 2011 study found that subjects were willing to pay 63% more for furniture they had assembled themselves than for equivalent pre-assembled items.
If people were being rational, they'd pay less for items they had to work for in addition, but since they over-value the outcome by their own effort, the situation is reversed. People are both willing to spend more effort AND pay more money. The same thing happens with AI - it promises to be the cheaper "flat-pack" option, but actually costs you more.
Edited at 2025/09/05 17:31:12
|
