A couple of people wrote to me to say that they are annoyed that we built robots to make “art” (scare quotes) when we should be making robots to do laundry.
And aside from the fact that unless you do take your laundry to the river to wash it, you already have a robot that does 90% of the work (and I can respect wanting automation for the other 10%)…
We don’t have robots (chatbots) that make art or write code either! They do an average job of some things, poor at others, and you can’t rely on them at all. If this is “AI”, I want it nowhere near my laundry.
Our 2020s tech industry loves to automate the easy part and hand-wave away the part that matters. The chatbots can’t do anything; they can only do the pointless parts, the boilerplate, the busywork. This is not the work. Humans still have to do this and the chatbots make it harder because they have to sift through bullshit to do so.
Keep Sam Altman the fuck away from my chores. He’ll just break all my dishes and dye my t-shirts blue.
@samir "And aside from the fact that unless you do take your laundry to the river to wash it, you already have a robot that does 90% of the work (and I can respect wanting automation for the other 10%)…"
I guess the ironing might only be 10% but it still would be nice to get rid of having to do it ...
> Our 2020s tech industry loves to automate the easy part and hand-wave away the part that matters
Which itself is pretty much just what Brooks warned about in No Silver Bullet back in the 80s.
We’re pouring hundreds of billions into something that in the absolute best case scenario (for boosters) would only address accidental complexity, and expecting somehow essential complexity will disappear too
@samir enjoy...writing lots of tests.
you feeling alright mate?
Except that the code review is also given to the automate that programmed the shit ( and yes vibing code it is for you. )
And good luck for the maintenance of programs.
@samir Respectfully I think that's ... maybe the inverse?
We automated the bad bit, which is "generating the text", in a way that obviates the important part, of creating a community of comprehension around the problem we're solving.
@mhoye I don’t think you can separate these two things.
And I prefer my sharing through pair or ensemble programming, not asynchronous reading, which always misses the important part (e.g. “why are we doing this, anyway?”)
I think that's more true of emails, status reports, and PowerPoint presentations than of software that must actually do the work of running the company. Far more obviously dangerous in the latter.
In the former, the management and sales spaces were filled with ill-thought-out nonsense already.
@paul_edwin the purpose here seems to be to avoid learning or sharing at the review stage though
that's not what folks are doing, they're cutting that entire step and moving towards incomprehension
From the Department of "Just Because You •Can• Doesn't Mean You •Should•"
Also, LLMs are designed to fool us. That's their core essential feature. That's what they're designed to be best at.
So they not only increase the volume (both frequency and size) of code reviews, they make the process much more difficult and error-prone, as they're *designed* at their core to produce plausible-looking code that will fool code reviewers.
@samir Instead of getting a robot that does the laundry and washes dishes so I can spend my time being creative, I get a thing that fakes creativity so I can spend my time doing laundry and washing dishes.
This is the most disappointing dystopia possible.
@MyLittleMetroid It is true, I could go for more automation!
What I really want is a bathroom-cleaning robot. I hate cleaning the bathroom.