AI slop is good for business if you know what you're doing Amazon founder Jeff Bezos is known for saying, "Your margin is my opportunity." Substitute "irresponsibility" for "margin," and you have a formulation that applies to AI-driven software development. AI models or coding agents make mistakes, and many of the people using them multiply those errors by asking their AI helper to …
So what do you end up with if you feed one AIs code-slop into another AI and ask it to fix the input garbage?
Do you double the nonsense (or square it) or do you actually get something closer to what a human would have produced.
I am sure this has been tried. I wonder what the result was.
People probably have tried more professionally than I did. When I did it, it was very random. Some things got a lot better, one massive inefficiency was eliminated, a snarl of code was removed, some new weird bugs got introduced, and everything got moved around so even learning what changed took way too long. Your experience will probably differ, but I don't know how.
So what do you end up with if you feed one AIs code-slop into another AI and ask it to fix the input garbage?Do you double the nonsense (or square it) or do you actually get something closer to what a human would have produced. .... Pete 2
Yes, quite so, Pete 2, that is exactly the result whenever the fix to input garbage asked of humans is nowhere near as perfect as is possible and as AI can float ....... and as could and should be produced by humans if they were able and not riddled with so many imperfections.
The simply complex question nowadays to ask yourself of these emerged and surging AI states with their myriad impressive experiences with immersive experimental virtual spaces is can they be bothered to fix all the nonsense? Will they or won't they?
And there are no trick questions testing levels of intelligence asked here. Methinks a tally of likes and dislikes to what is shared here can answer that well enough to be an accurate enough indication of the fundamental truth.
> I wonder what the result was.
I shouldn't think you would need to ask.
I imagine repeated application of AI would converge on a fixed point or fly off to infinity. In any case you would get total nonsense either concisely or endlessly.
The wrong technology for the wrong problem - like using a spell checker to check grammar… only infinitely worse.
Cynic that I am, I would suspect these "one stop deslop shops" are first shipping their clients' sloppery off to outsourcers on the subcontinent where real programmers (of sorts) knock it into some sort of shape before returning it to the principals who then run an eye over it. Which probably brings the client back to where he was before he embarked on his "AI journey."
All rather "There's a hole in my bucket, dear Liza"
> I shouldn't think you would need to ask
Well yes, I do need to ask.
And when I did, this was the response:
Yes, you can use a second AI to fix code made by a first AI. This method works well because different tools have different strengths. One AI might find bugs or fix logic errors that the first one missed
Although since I asked an AI, I refer you to Mandy Rice-Davies now famous remark
"The wrong technology for the wrong problem."
Indeed, although long ago I worked for an organisation where the Finance team complained that Excel was "rubbish at word processing" but they refused to use Word as they only wanted to use Excel, Outlook and their accounting software. Eventually they saw sense and started using Word (ok, Word isn't great but at least it is intended for word processing).
"I imagine repeated application of AI would converge on a fixed point or fly off to infinity"
I suspect it's like mixing all the plasticine together - the code will all end up brown. And that's probably an apt description of it, if nothing else.
Yeah, it's like when you give yourself one of those AI (so-called) entirely automated RoomBaZor™ haircuts at home, superhuman-vibed up and all, shaving time off yer busy schedule, you always end up looking like Verity Stob facing execution or suchlikes, double-crested mohawk, leaopard spots, one missing eyebrow ...
It's good to have these outside expert barbeshop AI meatbag outfits at the ready then, to put'em follicles back into production-ready business logic deduped order, in some right tear-free no-more-tangles fashion imho, adroitly shavin' the 2ⁿᵈ eyebrow for symmetry, deftly adding fake eyelashes for style, plus masterfully layering on a few multiples of the regular kind (of lashes) for spanking good measure!
One'd shudder to think of not using a cutting edge LLM's Mixture-of-Clippers or Chainsaw-of-Tresses agentic skill for such activities in the first place, at all, randomly even ... I mean, really, who wouldn't?!! ;)
I can see some people viewing fixing AI slop code as an opportunity, but to me such drudge work and maintenance of someone else's screw-ups is just a pain-in-the-arse nightmare to be avoided at all costs, the same as I avoided taking on jobs that would have required fixing manually created slop code (and I did have a few opportunities cross my table over the years to do just that - I only took on one such job for AT&T cleaning up their billing process, because even a surface skim of the code showed that 99% of the problems were caused by repeated queries to scan the data for each step of the process instead of scanning once and using smarter logic to do the calculations. I was rather proud of the results: run time under 2 hours instead of over 24, which had been causing the OS to kill the job before completion.)
