Elektrine lite

← Feed

@tshirtman@mas.to

2026-11-12 19:06 UTC

@glyph@mastodon.social (not sure if the thread is done) For that matter, i agree with all that, and it's been a concern to me for a long time, and it was even before AI, as i automated company processes done by real humans by hands before, and these humans could explain what could go wrong before, and the dependency chains, but if (when) we automated, how long would would it take until they couldn't explain that anymore, and rely entirely on the automation being correct? I agree AI makes this problem much − and potentially infinitely − bigger, and even if we end up scaling that ladder all the way, i want to scale it slowly, purposefully enough, that i have time to understand what i gave away, and check how things are going even if i handed them off, if i *really* still understand, what safety or higher level process is needed for control, etc, before moving on, and consider if i *want* to forget about it, or if operational control is still needed. Any activity, we stop doing, we forget how to do.

Replies (1)

  • @glyph@mastodon.social 2026-11-12 19:08

    @tshirtman@mas.to Yep, exactly. Almost every "AI problem" is an existing problem which AI is making catastrophically worse :). At my last job we were wrestling with a ton of non-AI automation problems, Taylorists were fretting about it in the 1910s, automation is a process you need to approach with discernment and intention, otherwise lots of kinds of cognitive offloading can go off the rails. You have to ask these questions of any technology and they have been quite bad with some tech in the past.

    Open ##5548635