this post was submitted on 29 Mar 2026
124 points (100.0% liked)
Fuck AI
6540 readers
1083 users here now
"We did it, Patrick! We made a technological breakthrough!"
A place for all those who loathe AI to discuss things, post articles, and ridicule the AI hype. Proud supporter of working people. And proud booer of SXSW 2024.
AI, in this case, refers to LLMs, GPT technology, and anything listed as "AI" meant to increase market valuations.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
To be fair: why should you search for the calculator app on your phone when you have ChatGPT right on your first page? (Assuming the AI would be able to do maths without errors)
Edit: People here really think that everyone is totally aware what the flaws of AI are. Of course that's a stupid thing to do. But you some people don't really think about such "small" decisions.
Because you’re trashing the environment, wasting compute cycles, unnecessarily transferring data, wasting bandwidth (however small), etc?
Just use the calculator app bro.
Or do some mental math. It's not a particularly hard one to do.
Some people will do anything to prevent having to do math. Part of my job includes counting tally marks from a sheet and entering it into a file. My coworkers use click-counters (which we all have because we use them for other tasks) and will sit there clicking the thing over and over again just to write down the final number. Meanwhile, I scan the page and use mental math to quickly come to an answer. I count twice to be sure, and I'm still done quicker than the click-counters.
It's funny, when I started I was given a click-counter and told I could use it for this task. I responded that I'd get too hung up thinking I accidentally clicked it too many or too few times, and would probably end up in a loop clicking and resetting the thing over and again because of it. They're tally marks, they're not complicated, it's much easier to just count them in my head than introduce a tool that will provide a new margin of error.
If a room of >6 people can't work out 1500/6 in their heads, they don't deserve to have their drain fixed.
I think this is only a valid question because the tech industry has worked so very, very hard to conceal all of the hidden costs of AI.
A question that's analogous to your original but takes all of the ridiculous costs into account might be something more like "Why drive 20 minutes to the store when you can get there in a helicopter in 5 minutes?"
Like, yeah... one is faster. But speed isn't the only variable, and there are plenty of obvious reasons why riding a helicopter to go get groceries or whatever is ridiculous. Soon, when AI firms can no longer afford to subsidize their costs, I think people will start to view AI this same way.
Sora at 15 million USD/day burn rate and 3 million USD in yearly revenues buttresses your point. I did see a guy on here (maybe it was a reddit repost or HN repost bot now that I think of it since lemmy hates ai) doing the math and for his plan that was $150 a month or so his compute costs were between $60 and $120 at the published prices for compute hours. His point was also that the big players would have to stop subsidising eventually.
Yeah, the numbers are kind of crazy if you take the time to dig into them. I'm forced to use Claude at work, but the company has bought all of us in at the $200 a month tier. That subscription gives basically unlimited usage... assuming you stay within Claude's ecosystem (Claude code, basically).
However, if you want to use Claude with some other tools you have to generate an API key, and that gets billed by token usage (on top of the $200 per month). I know devs who are burning through $100 PER DAY on API usage, and I suspect their usage patterns are not at all unusual for heavy users. But even if it is, you'd still only have to be spending $7 a day in compute for the math on a $200 monthly sub to make sense for Anthropic, so it's pretty obvious that their subscription customers are losing them money at an insane rate.
I'm pretty sure even doubling the subscription costs wouldn't be enough for them to break even on compute, which is insane given how expensive a monthly sub is already. Once the VC money runs out and a sub price rises to $1000+ per month, the whole landscape is going to change extremely quickly.
That guy is assuming that the Government won't bail them out with borrowed tax dollars, which they will. Unless the credit card is already maxxed out, that is happening it's just a question of when, 37 trillion and counting.
An AI that can answer any math question flawlessly is undecidable.
If you restrict the flawless answers to simple add subtract multiply and divide, then you had might as well use your calculator app. Putting the prompt into the calculator is faster than prompting the AI.
To decide to use the calculator app, you have to know two things first:
AIs are bad at maths. Some people just never heard of that because they don't read any tech news and never questioned the results of AI.
Where your calculator app is. If you rarely use it and take several seconds to find it, prompting "Calculate 1500/6" might actually be faster.
So assuming the person this thread is about never questioned those things, using AI could be considered a logical decision. Not good, but logical.
Why would you waste the resources of a massive data center to solve a math problem that can be answered more reliably by a calculator that fits in your pocket and runs on a coin cell?
Why should you install AI and link it to first page as that's the place for calc app.
Are you lost?