

You’re absolutely right that the computer is still a black box to a lot of people, but throughout the personal computing era, there has at least been a pathway to mastery for the tools it offers. Furthermore, the touchscreen/smartphone era has roped in mechanisms of touch and proprioception that make the devices a more intimate, if deeply imperfect, extension of the self. Up until sometime late last decade, the Steve Jobs “bicycle of the mind” concept was still a driving force in the field.
I still don’t think most people grasp what a subtle, but fundamental, break it is that these AI products demand you confront them as a wholly separate entity from yourself. The path to mastery, and the feedback loop that builds that path, is so obscure it may as well not exist. If you wish to retrain a model, you’ve got to invest huge amounts of time and resources, as well as what remains a specialized (and not well-specified, as Ed highlights) skillset… and since it’s a probabilistic process, you’re still not going to get consistent results.
I am more and more convinced that one of the damning core flaws of the current crop of AI technologies is that they are designed to incentivize use of centralized computing resources. Their designers are simply asking completely the wrong questions for the people the technologies are being imposed upon. But you can’t say that someplace like HN, or even some parts of Bluesky, because so many people’s salaries still depend on the rents from centralized computing.
For the record, none of these generated clips thus far have featured an appearance by Omega Tom Hanks