this post was submitted on 10 Jul 2026
33 points (94.6% liked)
Low Effort Memes
705 readers
3 users here now
Some communities have standards. We do not.
This is a low-stress zone for sharing silly memes. Whether you made it yourself in MS Paint or found it in some weird corner of the internet, share it with us!
- Reposts from corpomedia are 100% fine, fuck them.
- Ancient memes that still hit are always welcome.
- If it made you chuckle, it's good enough for us.
- Be nice to each other: No hate speech or drama.
TLDR: meme gud
founded 1 year ago
MODERATORS
Y'know... I did find the argument rather compelling that "the only use-case for AI is to automate the kinds of work that would probably be better off never done in the first place" referring to fakework, busywork, filler work, make-work. "Look Busy" kinds of work.
the only shit AI is good for is the shit you never needed anyone for in the first place -_-
AI is also great for replacing the useless (micro)managers who impose all that busywork on people to begin with in order to feel like they're doing something, rather than wasting their time making other people waste their time.
That's a pretty narrow interpretation of the meme.
I was kind of on the AI is garbage train because it kinda is and I spent a lot of time cleaning up the impact of people using AI to make decisions they didn't have the technical knowledge make. But then I changed what I was using it for.
One example is that it's really good at searching for constants. I wanted to model some things rq in Excel as a general sanity check and instead of having to go though a bunch of books at my desk or having a subscription to Perrys or whatever, I asked if to find my constants with citations. A quick look at the sources to make sure it wasn't making things up and I had them!
It's actually much faster than online databases because these constants in particular are extremely specific to the conditions so I could just ask it to account for that vs looking at several databases (they don't always have them all) and having to read a bunch to figure out how they are sorted. It also searched for newer publications and in this process I got new ideas on how to improve my model just because something triggered a thought.
So, AI had vastly improved search and I assure you that very little is filed properly so this massive improvement over word matching indexed files is going to make society crumble a little more slowly ✨
AI isn't necessary for that kind of search, however. A simple vector database lookup would do, and rather than passing the output through an unreliable narration layer, it would simply return the source texts directly.
Semantic search is a great technology, but it is separate from and independent of 'AI'.
Okay so is that available to me
Edit: I told the robot to just return values and sources it doesn't write me novels
my takeaway from this is that the AI is bypassing the "gatekeeping basic access to information resources by charging people subscriptions" business...
...which is, I note, again: something nobody should be doing in the first place.
AND FURTHERMORE, is literally what the business model of these AI firms will be if they actually made people pay at least as much for the tokens as they are spending in order to process those tokens.
because, as it turns out, running these gorillion-parameter LLMs actually costs money in terms of energy and cooling and the upkeep cost of the facilities these machines are running in, and it costs A SHIT TON MORE THAN WE HAVE BEEN LED TO BELIEVE - that's why AI companies are losing money so fast.
Didn't we all SEE what happened when these companies actually start BILLING for token usage?
Micro$lop has been finding out (having FUCKED AROUND) that their engineers are each racking up - this is the cost PER ENGINEER - of THOUSANDS of dollars extra every single month on Claude tokens. Spending as much as giving every last one of their engineers a $24,000 raise except with way less productivity resulting.
A user can accidentally burn through half their monthly 'token budget' allotted by their subscription in a DAY even without 'heavy usage' because IT TURNS OUT WE DON'T EVEN GET TO CONTROL HOW MANY TOKENS IT USES WHEN WE SUBMIT A QUERY! it can get stuck in a logic loop and churn through millions of them on the SIMPLEST queries.
god DAMMIT, just because they aren't BILLING someone yet doesn't mean there isn't a COST, and people are thinking it's affordable as a result WHEN IT REALLY FUCKIN AINT