|
Not to be an AI apologist, but isn’t “should I buy or lease X car” kind of a great use case for the AI of today? “Here is the cost of this vehicle, new, at these dealers near you, based on their websites. Taxes in your state are this, and registration fees in your state are this, your upfront cost will likely be this. At these banks/credit unions near you, the rate for a new auto loan is this, based on the credit score you told me. Your total cost of car, fees, and interest would be this, and your monthly payment would be this.” It could pretty easily go through that exercise for new, used, and lease (provided lease details like buyout cost are available, I have no idea if they are) and give the user a much clearer picture of what they’re walking into with data compiled from a lot of sources quickly. It’s not going to answer the question directly, but with the right prompts could be pretty useful versus going from site to site like a ton of people do. Edit: Fresh page, so Meow Wolf story. I took an edible and went to Denver Meow Wolf after a few days of work stuff, and thought going into it that I might have been exposed to COVID in the previous days. I got super paranoid with all the people around and the experience in general and had a Bad Time in there.
|
# ? May 5, 2024 19:51 |
|
|
# ? Jun 9, 2024 10:07 |
|
The problem is no ai is remotely capable of that
|
# ? May 5, 2024 19:52 |
|
Schiavona posted:Not to be an AI apologist, but isn’t “should I buy or lease X car” kind of a great use case for the AI of today? It would be useful if that's how it worked, but there is no information retrieval or data compilation happening. The AI is predicting what a human would say in response to the prompt, and there are no guarantees that it's going to accurately state the dealer's terms.
|
# ? May 5, 2024 19:53 |
|
Schiavona posted:Edit: Fresh page, so Meow Wolf story. I took an edible and went to Denver Meow Wolf after a few days of work stuff, and thought going into it that I might have been exposed to COVID in the previous days. I got super paranoid with all the people around and the experience in general and had a Bad Time in there. I'm sorry that was your experience; for better or worse you'll find that all of MW's installations are really popular so you'll never fully escape crowds there anytime soon. Try again when you're straight and level, I've found them to be a really fun experience and could spend all day in there.
|
# ? May 5, 2024 20:01 |
|
AreWeDrunkYet posted:https://www.fastcompany.com/91114065/meow-wolfs-co-founder-is-building-a-trippy-spa-of-the-future-in-austin My friends and I have already promised each other we will be there as soon as it opens. (So I've got about 25 months to make my abs visible.) Meow Wolf seems to be in a weird place right now. They're expanding to Houston this year, and also announced a Los Angeles location. Meanwhile, they also laid off something like 10% of their workforce a few weeks ago. The experiences are pretty fun, and worth the high price, but I don't know how sustainable the model is. How big is their customer base, really? And how many times will people re-visit a location?
|
# ? May 5, 2024 20:02 |
|
Bird in a Blender posted:Every online chat systems is already run by ChatGPT or some other AI bullshit. Eventually all phone calls will be the same, I just don’t think they’ve quite gotten good enough to do it for every application yet. I know it’s being used for some things already where you at least think you’re talking to a human, not the usual menu stuff. Talking to an AI is already a vast improvement over talking to someone in Indonesia who is reading cards to you.
|
# ? May 5, 2024 20:02 |
|
AI is great because you can like now never actually get answers to anything and you’re just caught in an endless loop of misunderstanding. Also, 0 no longer works to boot to an actual representative.
|
# ? May 5, 2024 20:22 |
|
Eric the Mauve posted:Talking to an AI is already a vast improvement over talking to someone in Indonesia who is reading cards to you. At least the person is reading the cards and not making up what the cards might say based on some cards they stole from the next call centre over.
|
# ? May 5, 2024 20:24 |
|
ultrafilter posted:It would be useful if that's how it worked, but there is no information retrieval or data compilation happening. The AI is predicting what a human would say in response to the prompt, and there are no guarantees that it's going to accurately state the dealer's terms. This is not true. (Some) LLMs are capable of retrieving data on the internet. I am literally helping train a LLM model right now in-between shitposting where the model is grabbing information from the internet in realtime. The problem is that the models aren't giving me sources for their findings and I am not allowed to penalize the models for giving wrong information (because I don't know where it got it from). So… keep that in mind in the future when you are asking LLMs stuff like: Schiavona posted:Not to be an AI apologist, but isn’t “should I buy or lease X car” kind of a great use case for the AI of today? Because they are not being trained to discern what information on the internet is real and what isn't.
|
# ? May 5, 2024 20:45 |
|
Schiavona posted:Not to be an AI apologist, but isn’t “should I buy or lease X car” kind of a great use case for the AI of today? This is not the level of maturity of current LLMs. Not even close. In fact, this scenario with a guarantee of accuracy an no AI "hallucination" is basically still far enough away to be science fiction. You're believing marketing hype. It's not reality.
|
# ? May 5, 2024 20:48 |
|
Eric the Mauve posted:Talking to an AI is already a vast improvement over talking to someone in Indonesia who is reading cards to you. Whether someone or, now, something, is talking confidently and coherently has no bearing on where it is an accurate or useful thing to listen to. The person in Indonesia reading cards at least has mostly accurate basic information, which you can actually use on some level. The AI is just loving guessing at literally every aspect of communication, and just because it's okay at guessing what a good sentence looks like doesn't mean it's good at guessing anything else. Don't fall for the same poo poo that got your great great great grandparents to buy Dr Clark's Miracle Elixir because he had a well painted cart and a loud voice.
|
# ? May 5, 2024 20:58 |
|
Yes I understand all that, that was more about voicing my contempt for modern "customer service" whose purpose is to make you give up and stop bothering them. Like, the person reading cards to you is giving you no information at all that you can't google for yourself. Of course companies can't wait to cut even that cost in favor of automating it all, and who really cares if the information the customer gets is factually accurate or not if there is literally no way for the customer to actually reach you or any human who works for you?
|
# ? May 5, 2024 21:03 |
|
Schiavona posted:Not to be an AI apologist, but I think the issue most people are going to have is that in the best case you're using AI to fill out this webform. https://www.edmunds.com/calculators/lease-vs-buy-calculator/, worst case its making up numbers that seem right to the algorithm. In my limited experience ChatGPT is surprisingly bad at straightforward math. VV- Yea, I understand that but try explaining that to the AI. A computer that can't do math is like a Full Self Driving car that drives into firetrucks. nomad2020 fucked around with this message at 21:29 on May 5, 2024 |
# ? May 5, 2024 21:19 |
|
Boris Galerkin posted:This is not true. (Some) LLMs are capable of retrieving data on the internet. I am literally helping train a LLM model right now in-between shitposting where the model is grabbing information from the internet in realtime. The problem is that the models aren't giving me sources for their findings and I am not allowed to penalize the models for giving wrong information (because I don't know where it got it from). So… keep that in mind in the future when you are asking LLMs stuff like: Congratulations, you're working on rear end in a top hat Technology that only has rear end in a top hat Uses. nomad2020 posted:I think the issue most people are going to have is that in the best case you're using AI to fill out this webform. https://www.edmunds.com/calculators/lease-vs-buy-calculator/, worst case its making up numbers that seem right to the algorithm. In my limited experience ChatGPT is surprisingly bad at straightforward math. That's because it doesn't actually do maths, it just reaches into a big cauldron mix of everyone else's maths homework, so it might grab questions from one place and answers from another, or even someone else's completely wrong calculations. It doesn't even understand maths enough to actually be bad at it. It understands nothing.
|
# ? May 5, 2024 21:26 |
|
My fave thing about AI being bad with money (aside from a bunch of companies desperately pumping tens of billions of dollars into the industry to distract everyone from the circles they've been going in for the past decade) is that the boosters are claiming exponential growth will happen in the field that gave us clippy, Alexa, and Siri. None of which are known to have grown at all, let alone exponentially. I still have to spell out the names of half the songs in my music library to Siri if I want to play a specific song. Same experience as I had back in 2012. The thing is confident, smarmy, and the biggest mark in the universe. Nobody is saying anything at all about training LLMs to reject false information, and that's a rudimentary request that is going to kill it stone dead in any field with a hint of accountability. gently caress, I'm the old lady yelling at clouds now. oh well.
|
# ? May 5, 2024 22:01 |
|
I feel like most companies training LLM's are probably doing it as a cover for whatever they're actually going to do with the massive amounts of scraped data.
|
# ? May 5, 2024 22:41 |
|
Midjack posted:Pluteschool nobleman peeps with figgie dealios. ah, I see you've finished training the BDID LLM
|
# ? May 5, 2024 22:49 |
|
Lyesh posted:Same experience as I had back in 2012. This is the truth, and the best example of "none of this poo poo actually got to the point that it's truly useful for any purpose". It's gotten incrementally better, but I still don't see any real exponential growth use cases for its current state that aren't "stealing content and regurgitating it, potentially at the correct time(s)" which is not gonna fly for basic legal reasons. Desert Bus posted:I feel like most companies training LLM's are probably doing it as a cover for whatever they're actually going to do with the massive amounts of scraped data.
|
# ? May 5, 2024 22:51 |
|
Desert Bus posted:I feel like most companies training LLM's are probably doing it as a cover for whatever they're actually going to do with the massive amounts of scraped data. But yeah, the space is populated with true believers, the credulous, and those who have no scruples against making bank by selling the other two a bill of goods. Cugel the Clever fucked around with this message at 04:05 on May 6, 2024 |
# ? May 6, 2024 00:34 |
|
I think I know someone who got scammed by Pauzible. They thought they just “paused” their mortgage with them or someone like them, and is now having to pay expenses just to sell and walk away from their house. This thread is still the #2 google result for pauzible scam.
|
# ? May 6, 2024 01:00 |
|
Mr. Nice! posted:I think I know someone who got scammed by Pauzible. They thought they just “paused” their mortgage with them or someone like them, and is now having to pay expenses just to sell and walk away from their house. and #5. two hits in the top 5. looks like the top hit came after they actually put a calculator on their site and people could run numbers
|
# ? May 6, 2024 02:04 |
|
Eric the Mauve posted:Yes I understand all that, that was more about voicing my contempt for modern "customer service" whose purpose is to make you give up and stop bothering them. Like, the person reading cards to you is giving you no information at all that you can't google for yourself. Of course companies can't wait to cut even that cost in favor of automating it all, and who really cares if the information the customer gets is factually accurate or not if there is literally no way for the customer to actually reach you or any human who works for you? On the other hand, half their clients are either incapable or unwilling to do that googling themselves.
|
# ? May 6, 2024 02:18 |
|
nomad2020 posted:I think the issue most people are going to have is that in the best case you're using AI to fill out this webform. https://www.edmunds.com/calculators/lease-vs-buy-calculator/ yeah i feel like the biggest thing about ai hype is when people or companies are marketing some advanced AI tech (that the technology isn't even there for yet so it doesn't work reliably) where even the end result, with a theoretical perfect 100% hit rate AI, is no more convenient than it would've been before For example, the "AI shopping" startup Nate, whose whole value proposition was that you could, like, copy-paste a link to an online-store product page into their app and it'd use ai to "automatically" click the buy button for you and I guess fill in your credit card details. And for this fantastic convenience they'd charge you $1 per transaction. Of course they never actually got the ai part to work: https://www.theverge.com/2022/6/6/23156318/artificial-intelligence-nate-app-ecommerce-go-read-this posted:If you can’t be bothered to fill out your credit card and address details when shopping for jeans online, the Nate app sounds like a service you might want. The company bills itself as an “artificial intelligence startup” that uses AI to auto-fill customer information for $1 per transaction, saving shoppers a few minutes when completing purchases through the Nate app. So they were selling a useless service that added 0 value or convenience, and even that useless service was actually completely nonfunctional to the point they had to fake it. Is paying $1 to make a call center worker in the philippines type your credit card number out for you instead of using your web browser's autofill feature BWM?
|
# ? May 6, 2024 02:32 |
|
It turns out the real hype was artificial artificial intelligence all along.
|
# ? May 6, 2024 03:01 |
|
Mr. Nice! posted:I think I know someone who got scammed by Pauzible. They thought they just “paused” their mortgage with them or someone like them, and is now having to pay expenses just to sell and walk away from their house. Maybe we should change the title of the thread back to "Pauzible is a scam, they will steal your house and your money"?
|
# ? May 6, 2024 03:26 |
|
Pauzible stole my house and kicked my dog
|
# ? May 6, 2024 03:27 |
|
RPATDO_LAMD posted:yeah i feel like the biggest thing about ai hype is when people or companies are marketing some advanced AI tech (that the technology isn't even there for yet so it doesn't work reliably) where even the end result, with a theoretical perfect 100% hit rate AI, is no more convenient than it would've been before Wtf a credit card number is like 16 digits. It's not that hard to type in.
|
# ? May 6, 2024 03:29 |
|
Deformed Church posted:Whether someone or, now, something, is talking confidently and coherently has no bearing on where it is an accurate or useful thing to listen to. As the purveyor of the finest patent medicines available on this side of the Mississippi, I encourage you all to ignore the above post and purchase your tinctures from the man with the well painted cart and loud voice; these are the hallmarks of a man with the vim, vigor, and pep that come of taking Dr. Volmarias's Radium Code, the only tincture of Radium that will solve nearly any problem you encounter on these awful forums!
|
# ? May 6, 2024 03:33 |
|
I was wondering how to liven up my Sunday evening and injecting myself with Radium Code sounds like a capital idea!
|
# ? May 6, 2024 03:58 |
|
Hotel Kpro posted:I was wondering how to liven up my Sunday evening and injecting myself with Radium Code sounds like a capital idea! I’d suggest searching within you for the answer but, well, you know
|
# ? May 6, 2024 04:01 |
|
mrmcd posted:Wtf a credit card number is like 16 digits. It's not that hard to type in. Also, most major web browsers can save that info for you and autofill it. Like - wat.
|
# ? May 6, 2024 06:24 |
|
RPATDO_LAMD posted:yeah i feel like the biggest thing about ai hype is when people or companies are marketing some advanced AI tech (that the technology isn't even there for yet so it doesn't work reliably) where even the end result, with a theoretical perfect 100% hit rate AI, is no more convenient than it would've been before That's why a common joke is A.I. stands for Actually Indians. Like how Amazon mass surveillance no scan grocery story checkout was not AI based visual product identification but over 1000 Indians looking at the camera footage.
|
# ? May 6, 2024 15:40 |
|
'ai' is not one technology. it started off as a scientific grant sales tactic and now it's a general sales tactic. it's associated with like 15 different technologies, all of which have pretty decent claim to being 'ai' and which have like 5 orders of magnitude difference in effectiveness.
|
# ? May 6, 2024 15:49 |
|
However valid their claims to being 'ai' are, none of them can claim to be actual artificial intelligence.
|
# ? May 6, 2024 15:58 |
|
ai is in an alchemy stage of life, looking for the homonculus and the fifth element of quintessence and the paracelsuses are out doing their thing. but no boyles, no lavoisiers and no mendeleevs. the search for intelligence i think is basically comparable to the search for the philosopher's stone or the panacea, we'll finish it when we figure out it doesn't exist
|
# ? May 6, 2024 16:02 |
|
bob dobbs is dead posted:'ai' is not one technology. it started off as a scientific grant sales tactic and now it's a general sales tactic. it's associated with like 15 different technologies, all of which have pretty decent claim to being 'ai' and which have like 5 orders of magnitude difference in effectiveness. There is some overlap in that most of those technologies benefit from parallel processing that in turn is more efficient on similar underlying infrastructure. A lot of the video card shortage and nvidia bubble is data centers trying to ramp up GPU capacity, whether it's for machine learning, large language models, computer vision, generative models, or a bunch of other stuff being branded as "AI".
|
# ? May 6, 2024 16:24 |
|
there's no such thing as any applied math in the modern day without some matrix multiplying getting stuck in there somewhere, somehow, anyhow its just that the trend has gone to absolutely gigantic fuckin amounts of matrix multiplying
|
# ? May 6, 2024 16:26 |
|
bob dobbs is dead posted:ai is in an alchemy stage of life, looking for the homonculus and the fifth element of quintessence and the paracelsuses are out doing their thing. but no boyles, no lavoisiers and no mendeleevs. the search for intelligence i think is basically comparable to the search for the philosopher's stone or the panacea, we'll finish it when we figure out it doesn't exist Kind of a weird thing to say about Frankenstein's creation, but go off I guess. I'm kind of impressed even if he did it totally by accident.
|
# ? May 6, 2024 17:14 |
|
bob dobbs is dead posted:'ai' is not one technology. Sorry Bob, AI means chatbots now. Maybe it'll mean something else in 2025
|
# ? May 6, 2024 17:42 |
|
|
# ? Jun 9, 2024 10:07 |
|
it also means image generators !
|
# ? May 6, 2024 17:43 |