Access all digitized human knowledge.
Can AI help you expand your talents and find flow?
It appears that there is one effect many researchers are finding across multiple fields: generative AI has a significant impact on lower skilled and less experienced people. However, if we automate difficult tasks we cut ourselves off from the essential components for achieving mastery like flow.
- There is hype about AI's impact on individuals, but we should expect conclusions to change as research quality improves.
- People tend to overestimate how much time AI saves them on tasks—perception does not match reality. However, people enjoy work more with AI assistants.
- AI's failures are not obvious, so it's hard to calibrate appropriate confidence in its outputs. Overreliance leads to downgraded human performance.
- AI appears to significantly enhance the abilities of less skilled and experienced people—it is a "proficiency enhancer" when you know a little bit about something.
- If AI assumes difficult tasks before we have a chance to expand our talents, it risks short-circuiting human development and mastery.
- While AI may help streamline access to enjoyable, focused states of "flow", it cannot replicate the fulfillment and growth that comes from owning the hard-won process of mastery ourselves.
There’s a great deal of hype about the impact of generative AI on us as individuals. Studies are early, can be poorly designed, and are generally small. We can expect the data and conclusions to change, for the effects found to be smaller as study sizes increase. We can expect a wide range of anecdata as people share their trial-and-error experiences of using AI at work.
The perceptions people have tend to exaggerate generative AI’s productivity impact. When objective data is gathered we find that people think they save about three times more time on tasks when they use generative AI than they actually do. Data aside, people say they enjoy their work a lot more: for instance, coders are able to find flow states more easily and often when GitHub Copilot is on board.
More concerning is people’s “machine intuition” or lack of it. It’s difficult to calibrate confidence in AI’s output because how it fails is not obvious. And people get overconfident, downgrading their performance by over relying on AI. It’s a special kind of automation bias where AI is assumed to be correct, even when people know better than to trust it.
Having said all this, it does appear that there is one effect many researchers are finding across multiple fields: generative AI has a significant impact on lower skilled and less experienced people. It is a proficiency enhancer when you know a little bit about something. In some situations it can significantly accelerate people’s abilities, for example, in coding. Suddenly someone with very little experience can appear to have new expertise.