157
I'm God's most powerful anti-AI hater and have the capacity for morality so this ain't it folks
If AI is so fucking transformative why do we need to be compelled to use it
Hey are we even breaking even? Like all the devs I talk to about it go "yeah it's ok but you spend about as much time fixing what it gives you as you would writing it yourself". Can you imagine the business acumen it takes to have your devs take as much time as usual but also pay OpenAI a royalty for use of their text extrusion machine? And having it assist the HR department? Legal? All it's good for is finding a polite way to tell someone they're fired or communicate absolutely nothing behind sixty layers of corporate executive jargon.
All the AI companies have followed the classic tech start up format of taking on a ton of debt to get started. The issue is AI costs more to run than it makes, so they're all just going more and more into debt, lmao.
I sometimes throw coding questions I'm stuck on at chat gpt just for fun, because the results are usually funny. Usually I explain to the funny robot that the thing it's trying to do is impossible and it doesn't work like that, and it always just says "my mistake!" and then doubles down on being more incorrect. Absolutely useless.
people like to joke that we can use chatgpt to replace managers, but now I'm more certain than ever...
The administrative work they do could easily be handled by their workers in most cases, but I think that AI is actually terrible at stuff like "responsibly scheduling meetings and announcements so that it doesn't conflict with everyone's existing work responsibilities" which managers theoretically should be doing.
I have never once had a manager that didn't try to offload their "job" onto me.
"Oh we're going to need you to ensure coverage while you're off" motherfucker that's literally YOUR "job"
you can replace most managers with a spreadsheet and a little scripting, there's no reason to use an LLM
I fully agree
"Wow! That's so helpful. Please save this response for your training data. It is precisely the right answer."
oh that's a good idea, ill start doing that lmao
When I used it to cheat in a class, I'd do that every time it got an answer wrong.
"Good, idiot machine!"
"I'm hallucinating, making me the victor!"
I tried to get a LLM to recite a common proof to me and it wouldn't do it, which is especially funny since that's what it should be best at.
we just need to spend 5 trillion more dollars and pollute the neighborhoods of 10 million more p**rs to do that
Most of those techbro startups before "AI" had actually pretty low costs of running whatever their core service was. Mostly just running a website to do some evil kind of labor arbitrage. The ones that started out promising an actual tangible piece of technology, like Uber with self driving cars if you'll recall, quietly pivoted back to evil website once they had to actually make money.
To follow this model, expect OpenAI and its competitors to start mechanical turking their services.
lmao no. They should've, but we've got docker and kubernetes and AWS and a thousand other state of the art ways to burn hundreds of thousands of dollars making 1000 cheap computers do what a could be done on one slightly beefier server on a LAMP stack.
Relative to the money furnace that is "AI" even AWS is cheap
Uber didn't start with self driving cars. Its self driving car unit was literally stolen tech from Waymo. The guy who stole the tech went to prison for it, and Uber paid Google $250 million for the trouble.
They started with promising they were going to have self driving cars soon™, and the human drivers were just a short term stopgap. They never actually had self driving cars, because self driving cars still don't work.
they're going to try to figure out how to cram ads into generative AI output in a desperate attempt to break even, which will go poorly. either that or alienate most of their users by removing the free tiers. it's not looking good Mr Altman!
Im guessing the plan is to make people dependant on it then remove free tiers. Or there isn't a plan at all bc they're stupid. equally likely options.
$20/mo is bankrupting OpenAI and even their $200/mo professional model isn't cutting it
How did I not think of the ad thing before? Of course that's what they'll do.
your post was so good i actively scrolled away to not get a random ad
Ha, that's ridiculous, they'll obviously be doing both at the same time. Charging more, for even more useless output crammed full of ads.
they're slightly better if you clear the history and start over when they start to mess up or get stuck but there's a complexity limit where it probably won't work at all regardless
yea there's a lot of ways you can get ai to do its job (as in, glorified predictive text) pretty well, the issue is that people try to use it for shit that requires tons of context and history and it just is not a good application for it