Why are LLMs, the most useless and ineffective of AI systems, pushed the hardest?
Humanslop is already mainstream, might as well make AIslop mainstream too
it scales as long as you have enough training data. There is nothing else that scales that well and still keeps the wide problem solving skills.
>>107780924Because capitalism offers us the best allocation of resources!
>>107781144>it scales as long as you have enough training datadoes this mean we've officially run out of trainign data? LLMs have barely improved for a while
>>107780924How would LLMs be the most useless AI?If you don’t know anything about a subject, you can ask it a deep question to get a laymen’s term answer. This saves so much time and gives people access to knowledge that would have been inaccessible.I’ve used it as a physics tutor to explain concepts I wouldn’t have been able to understand.I also use it in my codebase every day. You might not like it, but if used properly, it can be very helpful.
>>107780924normgroids love words and talking
>>107781281>How would LLMs be the most useless AI?Basically every other AI system, from chess bots to Doom monsters to Expert Systems computer vision has some use case. What's an objectively good use case for """LLMs"""?
>>107780924theyre cool i dont understand why people are so against being able to have a conversation with your computer, half the time i see these people making posts where they trick llms to show how dumb the llm is by asking it to solve riddles, maths or word games then boast about how dumb the llm is because they cant get it right. these people just seem scared to me why do people need to try so hard to show a cool thing is bad. yes they cant solve maths or whatever because thats not how they work theyre godtier search engines of their trained data
>>107781317>What's an objectively good use case for """LLMs"""?talking labile people into suicide
>>107781317the usecase is general purpose problem solving?? want to know about something or how to do something you ask it and it will give you concise information or instructions. the alternative is using a search engine to look for the info on websites or looking it up on youtube and hoping someone has explained but theres always tonnes of bloat info that doesnt really matter
>>107781317Try searching on internet how does a router recognize it received a packet.All you'll get is people explaining general workings of how routing works, while an AI will actually explain to you the bit check routers do to recognize packets.That's how it's useful. It's capable of giving you very specific information that you are looking for that would otherwise require you to spend hours of reading useless sources hoping it contains the answer buried somewhere in there.I legit wish I had this while I was in college, I spent hours searching for specific information because I couldn't understand the full picture.
>>107781317> What's an objectively good use case for """LLMs"""?Chat bot, coding helper, physics tutorIt’s honestly unfathomable you cannot find a good use case for LLM’s
>>107781356>the usecase is general purpose problem solving?That's not a use case. It can't do that.>>107781372>Try searching on internet how does a router recognize it received a packet.>AI will actually explain to youSo your use case is for retards who can't do a google search to be mislead by LLM hallucinations?>>107781375>Chat botUse case for a shat bot?>coding helper, physics tutorIt can't code its way out of a shoe box and it hallucinates too much to be trusted as a tutor.
>>107781399>So your use case is for retards who can't do a google search to be mislead by LLM hallucinations?Why don't you try it little faggot and show us the number of the pages it takes for you to find it https://letmegooglethat.com/?q=how+does+a+router+know+it+recieved+a+packetMake sure you link from this page because I don't want your lying faggot ass cheating
>>107781399kek funny how you do exactly what i said in my post here >>107781343 thats not the type of problem it solves seriously i dont get why people post stuff like this to show how its bad its like youre scared of a computer being good at something people do.
>>107781375>honestly unfathomableyes it's ridiculous hes probably retarded
>>107781281If you need a laymans answer you simply aren't capable of asking a "deep" question. To be able to ask deep questions you actually have to the required knowledge base to even possess the ability to do so.And the internet before LLMs already had plenty of good laymens explanations available for pretty much everything under the sun. The only thing chatbots really give you is allow you to speedrun AI psychosis by being having it confirm your shizo "questions"
>>107780924>pichivemind opinion detected>haha funny drawingonly artcucks make those, never techfags
Sure it sucks now but in two weeks we'll have AGI bro.
>>107781433>>the usecase is general purpose problem solving?>>107781433>thats not the type of problem it solvesYou're mentally ill and incoherent, just like the rest of your cult.
>>107781432https://letmegooglethat.com/?q=what+happens+when+a+router+receives+a+packet
>>107781375>physics tutorbut all info it provides is unreliable lol
>>107781547As someone who has spent a lot of time tutoring myself in physics/maths, I can tell you it’s infinitely better than nothing.Just being able to ask questions as you go is so helpful when you’re reading a dense book.Some stuff it can get wrong, but it’s definitely better than nothing, and it gets better all the time.
>>107781481Indians will never be more than slophumans and they most certainly 100% without a doubt will never be white
>>107782190>i don't know anything about physics and use a mindless token guesser as a "tutor">let me tell you how good LLMs are at teaching physicsYour lack of self-awareness is very much reminiscent of the token guesser itself.
>>107781547It can help you reason through things. You still have to check its work.
>>107780924It's the closest to a universal solution out of all AI types. Everything else is laser focused in its uses, but llms work for Shandra in accounting as well as it works for Pedro the gooner. I mean, it's still shit for both of them, but they both get to use it.Personally, I see this more as a defect than something good. Focused AIs would perform better and give out better results, and LLMs will never reach AGI, it's just not the right tech for it, so it's dumb to spend so much effort trying to pass this square peg through a round hole.>>107781281imo, using LLMs for things you don't already know the answer to is literally the worst possible usage for this tech. It hallucinates too much for you to use without being able to filter out the garbage.You need to already know the answer but be too lazy to type it out, that's the best scenario for using this mess.
>>107782234You think humans get everything right every time? You think humans don’t hallucinate, misremember, etc? Is all AI useless in your view then, because virtually everything that uses a neural network doesn’t get it right every time?Or do you just have some weird issue with LLMs?
>>107783404>imo, using LLMs for things you don't already know the answer to is literally the worst possible usage for this tech. It hallucinates too much for you to use without being able to filter out the garbageIt just has to be marginally better than the alternative to be useful. I cannot afford to hire a Ph. D. Physicist to tutor me, and it would require blocking off my schedule. This I can do at my leisure. You still have to understand what is going on. I have a problem/solution set (going through Schaum’s 3000 physics problems set), so I will know if the information is incorrect.And it’s not like I have no idea what is going on, and I’m asking the LLM to teach me. It’s just helping me reason through the problems. It would be fairly obvious if it were hallucinatingYou still have to discern what is true and what isn’t even if you aren’t using an LLM, and it hallucinates in physics much less than in other things.I already have the solution, and even an explanation. Just when the explanation isn’t entirely clear, I rely on an LLM to help me figure it out.
>>107781281>Hallucination machine teaching physics>Maintains codebase>Reddit spacingThis one might be a bit too subtle for most people to get
>>107781281>If you don’t know anything about a subject, you can ask it a deep question to get an answer which probably includes bullshit. However, since you don't know anything, you will believe the bullshit as if it was real, trustable, honest info. Thus, becoming even more retarded over timeFTFY
>>107781190There are now companies that are in the business of creating new training data for AI companies by employing humans from various domains to generate and label this data
>>107781317>What's an objectively good use case for """LLMs"""?Web search from multiple different sources from a single query
>>107783500>You think humans get everything right every time? You think humans don’t hallucinate, misremember, etc?No human qualified in their field of expertise bullshits at all. Unless he/she is a piece of shit person.
>>107783646>>Hallucination machine teaching physics>>Maintains codebase>>Reddit spacing>This one might be a bit too subtle for most people to getyes, he's a dork
>>107780924nothing else is going on right now, and the line must go up
>>107780924Jews
>>107783871>No human qualified in their field of expertise bullshits at all. Unless he/she is a piece of shit person.https://www.youtube.com/watch?v=5eW6Eagr9XA
