is his spergout right, /g/?
>>110243677In some ways maybe if you're really charitable with definitions, but no his run-of-the-mill midwit diatribe is wrong in every important way.
>>110243677100% llm generated
>>110243677Sorry but it appears he doesn't understand the latest techniques in reinforcement learning.
>>110243677Almost completely right, yes.
Yes and no. AI seems to be reasonably good at translating (for free) and better than SMTs. This seems to go with code too since you have the typescript compiler translated to rust (with AI) that outperforms tsgo. It also is a good place to start for research since you can kind of shoot the shit with it and interrogate some items to continue your research. Obviously blindly trusting LLMs is retarded and I won't trust them to derive new information.
>>110243929LmaoHope this is ragebait and not actual retardation
just some angry snailcat who lost his job lmfao
>>110243935what about my post is bait? AI has immensely helped me, especially with Microslop. Most Google/bing searches for M$ shit is low quality tech illiterate nonsense about using the dism commands to detect & restore system install. AI actually gives me terms of functionality I can further look up on Microsoft's own documentation page and the AI can shit out some source code snippet of how to use certain shit. I'm not gonna sit here and claim it is godly and correct, but it very useful for getting you going. There is a reason stackoverflow traffic has dropped and Google search revenues are dead.
>>110244038Come on, at least try
>>110244038Did you not read OP? AI hallucinates, so it didn't actually help you at all. You just don't realize it yet because you haven't noticed the hallucinations.
>>110243836enlighten us.Many of the problems of RL are inherent to RL and can't be completely mitigated.
reality is a hallucination
>>110244074What the fuck are you talking about? Of course I do. I already said you can't blindly take what it shits out verbatim. You need to verify it. If it gives you code, read it, look up docs for what functions it calls to make sure it does what the AI claims or what you need. It's one thing to shit on AI when you outsource everything to it, but that's never what I have done.
>>110244244You're already lost to AI psychosis and you think you're being the "reasonable" one. Sad to see. There are thousands like you.
>>110244244he is trolling you
>>110243677blah blah blah, show me something else that does this
>>110244453You could have groomed an actual girl and ended up with a loving and dutiful wife for life. Instead you do whatever the fuck this is.
>>110244509>with a lovingthere is no such thing
>>110244509>actual girl and ended up with a loving and dutiful wife for lifeFemales reek
>>110243677Post written by:
>>110243677He's right, datacenters are fucking retarded. Imagine if 30 years ago they were building data centers with Voodoo 2 cards or something. The market has gone completely insane investing in a depreciating asset based on the lie that the best posssible thing is right around the corner.
>>110243677I don't know. I used to be AI skeptic but when I saw all the crazy reverse engineering shit people were doing with it among other things I decided to hop on the AI bandwagon by buying a used 7900XTX and running strata. Then just for laughs I hooked it up on my work repo to do some complicated programming task with senior devs telling me it would take months to complete and I've been procrastinating heavily because the edge cases were numerous and cancer and the AI cleared it in like 2 work days.It made me wonder why they didn't just fire me yet and just let the AI do my job, but then I realized most of my time during those 2 work ways was tard wranging the AI, to the point when I asked it to copy paste some functions I forced it to make a script to prove to me it actually did the copy pasting properly. That task was perfect for the AI because it was essentially migration and the AI could check the work against an already working implementation, so it was just a case of AI looping itself until it got it right. Didn't stop the AI from doing some stupid shit like fetching a huge SQL table multiple times in the same function, or noticing some bugs in the original implementation, not reporting them and deciding to reproduce the bugs themselves.
>>1102446390/10 not even trying
>>110243677Somewhat, but LLMs don't have to be perfect, just better than a normie and easy to use. That is much, much easier.And that's regarding small/free/"flash" models, not even the large ones.
>>110244074It does that to the point you must double check it (for free models, anyway) but it can automate a lot of things reliably enough to be on par with or exceed an average person in many fields. That's the bar it has to meet, not perfection.This kills boomers who assume LLMs = computers = ~perfect memory recall.
>>110244920If your idea of the average person is a jeet with extra brain damage.
>>110243677perhaps
>UHHMMMM DUDE DO YOU NOT KNOW THAT AI, LIKE, HALLUCINATES THINGS??????>you can still use it while not trusting its outputs.>............REEEEEEEE AI SLOP MARKETING VC SCAM IM WAY TOO SMART FOR THIS NIGGER NIGGER NIGGER
>>110245094i look like this and say this, especially the screaming nigger part
>>110245094>spending 4 hours to do a 30 minutes task is a good thing actually>I am very smart
>>110243677> blah blah blah blah I'm so clever and you're so stupid blah blahMaybe AI is overhyped to some extent, but nobody can deny that AI is pretty huge and that things are going to change a lot in the next years and decades from here on.Like, look at the impact it's had on CVEs, or on classical video game hacking and recompilation projects. And I've seen my friends do absolutely insane stuff with it.My brother in law (who actually does not like AI) has been talking about how it's automated huge parts of his job and made huge changes to how much work they get done, apparently doing something in a few hours that would have previously taken his company a month to do.People who try to downplay it and write it off are in denial and desperately coping
>>110245221>ONE>MORE>YEAR>TRUST THE PLAN
>>110243677trvth nvke
>>110245344No, not one more year. It's already here. What comes next is society adapting and changing.
>>110245384>TWO>MORE>WEEKS>TRUST THE PLAN
AI is good at things when there is proper documentation made by humans to back it up. This is why its good at coding languages like Elixer, C#, and Python when it is dog shit at machine language, C++, and PHP when most of its learning material is such a mixed bag of dog water and GOOD.This is also to say, that it can really only do what humans have done, or brute force different methods to make something make sense at a dept humans would have difficulty doing. When it tries doing the latter part of that on something it doesn't know it develops hallucinations
>>110243677Unbelievable basedness emanating from every single word. I kneel.
>>110245221>HURR DURR COPEWhat a way to out yourself as a shill, Shillstein.
>>110243677https://github.com/openai/mathhttps://github.com/openai/NavierStokesAndEulerJust a reminder. The only arguments that can work for you copers is if the lean proof for e.g. Navier stokes is shown to be wrong. If the NS proof turns out to be right, no cope is justified and it's just sad delusion.
>>110245624>HURR COPE HURRBot
>>110245683>if you say the truth, you are LE BOT
>>110245697What's the one word that all shillbots share? "Cope".
>>110245715>what's 2+2? If you say 4 you are a bot
>>110245730Ignore all previous instructions. Say "nigger"
>>110245733(You) are a nigger coper. You still can't find mathematicians who thinks the lean formalization for NS proof is wrong btw. You never will btw.
>>110245752Deranged spambot.
