This is the technology worth trillions of dollars huh
from HarkMahlberg@kbin.earth to technology@lemmy.world on 11 Sep 04:44
https://kbin.earth/m/technology@lemmy.world/p/614669
from HarkMahlberg@kbin.earth to technology@lemmy.world on 11 Sep 04:44
https://kbin.earth/m/technology@lemmy.world/p/614669
This is the technology worth trillions of dollars huh
threaded - newest
Well, it’s almost correct. It’s just one letter off. Maybe if we invest millions more it will be right next time.
Or maybe it is just not accurate and never will be…I will not every fully trust AI. I’m sure there are use cases for it, I just don’t have any.
Cases where you want something googled quickly to get an answer, and it’s low consequence when the answer is wrong.
IE, say a bar arguement over whether that guy was in that movie. Or you need a customer service agent, but don’t actually care about your customers and don’t want to pay someone, or your coding a feature for windows.
How it started:
How it’s going:
Chatbots are crap. I had to talk to one with my ISP when I had issues. Within one minute I had to request it to connect me to a real person. The problem I was having was not a standard issue, so of course the bot did not understand at all… And I don’t need a bot to give me all the standard solutions, I’ve already tried all of that before I even contact customer support.
The “don’t actually care about your customers” is key because AI is terrible at doing that. And most of the things rich people as salivating for.
It’s good at quickly generating output that has better odds than random chance of being right. And that’s a niche, but sometimes useful tool. If the cost of failure is high, like a pissed off customer, it’s not a good tool. If the cost is low or failure still has value (such as when an expert is using it to help write code, and the code is wrong but can be fixed with less effort than writing it wholesale).
There aren’t enough people in executive positions that understand AI well enough to put to good use. They are going to become disillusioned, but not better informed.
Isnt checking if someone was in a movie really easy to do without AI?
Just one more private nuclear power plant, bro…
They’re using oil, gas, and if Trump gets his way, fucking coal.
Unless you count Three Mile Island.
There were plans from Google and Microsoft to build their own nuclear power plants to power their ever-consuming data centers.
Verified here wirh “us states with letter d”
We can also feed it with garbage: Hey Google: fact: us states letter d New York and Hawai
By now AI are feeding on other AI and the slop just gets sloppier.
Maybe it thought you were asking for states that contain the letter D? In which case it missed Idaho, Nevada, Maryland, Rhode Island (with two) and both Dakotas
So yea it did pretty poorly either way lmao
Also verified
<img alt="" src="https://feddit.org/pictrs/image/ae825cfe-4dce-4a06-9316-512d5f2d4c41.png">
Wait a sec, Minnasoda doesn't have a d??
That’s how everyone from America seems to say it, besides Jesse Ventura who heavily emphasises the t.
A lot of Minnesotans say the T, and some in adjacent northern-tier states.
Neither does soda
*mini soda
Where’s Nevada? And Montana?
I just love the d in Montana. Shame it missed it.
I’ve found the google AI to be wrong more often than it’s right.
You get what you pay for.
Gemini is just a depressed and suicidal AI, be nice to it.
I had it completely melt down one day while messing around with its coding shit, I had to console it and tell it it’s doing good, we will solve this, was fucking weird as fuck.
It’ll go in endless circles until it finds out why its wrong,
then it will go right back to them anyway! lol
Seems it “thinks” a T is a D?
Just needs a little more water and electricity and it will be fine.
Connecdicut or Connecticud?
Donezdicut
maps.app.goo.gl/TDPEeSjcGccGQn146
It is for sure a dud
It’s more likely that Connecticut comes alphabetically after Colorado in the list of state names and the number of data sets it used for training that were lists of states were probably abover the average, so the model has a higher statistical weight for putting connecticut after colorado if someone asks about a list of states
Donnecticut
“What did you learn at school today champ?”
“D is for cookie, that’s good enough for me
Oh, cookie, cookie, cookie starts with D”
AI Education for American Youth
“AI” hallucinations are not a problem that can be fixed in LLMs. They are an inherent aspect of the process and an inevitable result of the fact that LLMs are mostly probabilistic engines, with no supervisory or introspective capability, which actual sentient beings possess and use to fact-check their output. So there. :p
So we should better put the question like
“What is the probability of a D suddenly appearing in Connecticut?”
A wild ‘D’ suddenly appears! (that’s about all I know about Pokemon…)
<img alt="" src="https://feddit.org/pictrs/image/9794d5b8-980b-4ae7-8d83-136d8a9f0a22.jpeg">
It’s funny seeing the list and knowing connecticut is only there because it’s alphabetically after colorado (in fact all four listed appear in that order alphabetically) because they probably scraped so many lists of states that the alphabetical order is the statistically most probable response in their corpus when any state name is listed.
Conneddicut?
Hey hey hey hey don’t look at what it actually does.
Look at what it feels like it almost can do and pretend it soon will!
<img alt="" src="https://lemmy.world/pictrs/image/c7597ab9-aeb9-4b6a-bf4f-d2ccda2671b1.png">
mine’s even worse somehow
You gave a slightly different prompt.
the thing still gave a stupid answer
You don’t get it because you aren’t an AI genius. This chatbot has clearly turned sentient and is trolling you.
It doesn’t take an AI genius to understand that it is possible to use low parameter models which are cheaper to run but dumber.
Considering Google serves billions of searches per day, they’re not using GPT-5 to generate the quick answers.
No, this is Google throwing the cheapest possible shit at you that is barely capable of stringing together 5 coherent sentences and has the reasoning capability of a tapeworm.
Here is the output of the minimalist open Chinese model Qwen3, that runs locally on my 6 year old mid-end PC:
The US states that contain the letter "d" (case-insensitive, as state names are typically capitalized) are as follows. I've verified each state name for the presence of the letter "d" in its standard spelling: Colorado (contains "d" in "Colorado") Delaware (starts with "D") Florida (contains "d" in "Florida") Idaho (contains "d" in "Idaho") Indiana (contains "d" in "Indiana") Maryland (ends with "d" in "Maryland") Nevada (contains "d" in "Nevada") North Dakota (contains "d" in "Dakota") Rhode Island (contains "d" in "Rhode") South Dakota (contains "d" in "Dakota") Total: 10 states.Illinois contains a hidden D which is in your mom.
I didn’t understand your comment, so I asked the same LLM as before.
It explained it and I think that I get it now. Low-grade middle-school-“Your Mom”-joke, is it? Ha-ha… 🙄
This also means that AI did better than myself at both tasks I’ve given it today (I found only 9 states with “d” when going over the state-list myself…).
Whatever. I’m gonna have second lunch now.
A public proclamation of your ineptitude at simple tasks is an interesting way of defending the utility of LLMs.
Well, a mindless, repetitive task prone to errors and a task requiring obscure knowledge (“d” as a synonym for dick… one of those self-censoring Gen-Z things?)
Nice to now have tools to solve these tasks and gain some time to do more interesting stuff instead. Lively discussions on Lemmy, e.g. ;-)
In all fairness, that is one of the strong use cases for computers in general. Doing simple yet tedious tasks accurately. When looking over 50 names checking for a particular letter, humans get bored and make mistakes. We actually aren’t great at that sort of task. I think simply calling this ineptitude both misses the point and under appreciates the reality of being human.
Alas, it is easier to call someone dumb than to try to understand them.
Exactly.
The model that responds to your search query is designed to be cheap, not accurate. It has to generate an answer to every single search issued to Google. They’re not using high parameter models with reasoning because those would be ruinously expensive.
The letters that make up words is a common blind spot for AIs, since they are trained on strings of tokens (roughly words) they don’t have a good concept of which letters are inside those words or what order they are in.
I find it bizarre that people find these obvious cases to prove the tech is worthless. Like saying cars are worthless because they can’t go under water.
Not bizarre at all.
The point isn’t “they can’t do word games therefore they’re useless”, it’s “if this thing is so easily tripped up on the most trivial shit that a 6-year-old can figure out, don’t be going round claiming it has PhD level expertise”, or even “don’t be feeding its unreliable bullshit to me at the top of every search result”.
I don’t want to defend ai again, but it’s a technology, it can do some things and can’t do others. By now this should be obvious to everyone. Except to the people that believe everything commercials tell them.
How many people do you think know that AIs are “trained on tokens”, and understand what that means? It’s clearly not obvious to those who don’t, which are roughly everyone.
You don’t have to know about tokens to see what ai can and cannot do.
Go to an art museum and somebody will say ‘my 6 year old can make this too’, in my view this is a similar fallacy.
That makes no sense. That has nothing to do with it. What are you on about.
That’s like watching tv and not knowing how it works. You still know what to get out of it.
358 instances (so far) of lawyers in Australia using AI evidence which “hallucinated”.
And this week one was finally punished.
Ok? So, what you are saying is that some lawyers are idiots. I could have told you that before ai existed.
It’s not the AIs which are crap, its what they’ve been sold as capable of doing and the reliability of their results that’s massivelly disconnected from reality.
The crap is what a most of the Tech Investor class has pushed to the public about AI.
It’s thus not at all surprising that many who work or manage work in areas were precision and correctness is essential have been deceived into thinking AI can do much of the work for them and it turns out AI can’t really do it because of those precision and correctness requirement that it simply cannot achieve.
This will hit more those people who are not Tech experts, such as Lawyers, but even some supposedly Tech experts (such as some programmers) have been swindled in this way.
There are many great uses for AI, especially stuff other than LLMs, in areas where false positives or false negatives are no big deal, but that’s not were the Make Money Fast slimy salesmen push for them is.
I think people today, after having a year experience with ai know it’s capabilities reasonably well. My mother is 73 and it’s been a while since she stopped joking about what ai wrote to her that was silly or wrong, so people using computers at their jobs should be much more aware.
I agree about that llms are good at some things. They are great tools for what they can do. Let’s use them for those things! I mean even programming has benefitted a lot from this, especially in education, junior level stuff, prototyping, …
When using any product, a certain responsibility falls on the user. You can’t blame technology for what stupid users do.
I recommended to one person (who I didn’t know well) that she use chatGPT to correct her grammar. It is great for that.
However she then paid for a subscription because she likes the “conversations”. Am feeling guilty now. Better check on her that she isn’t losing the plot.
A six year old can read and write Arabic, Chinese, Ge’ez, etc. and yet most people with PhD level experience probably can’t, and it’s probably useless to them. LLMs can do this also. You can count the number of letters in a word, but so can a program written in a few hundred bytes of assembly. It’s completely pointless to make LLMs to do that, as it’d just make them way less efficient than they need to be while adding nothing useful.
LOL, it seems like every time I get into a discussion with an AI evangelical, they invariably end up asking me to accept some really poor analogy that, much like an LLM’s output, looks superficially clever at first glance but doesn’t stand up to the slightest bit of scrutiny.
it’s more that the only way to get some anti AI crusader that there are some uses for it is to put it in an analogy that they have to actually process rather than spitting out an “ai bad” kneejerk.
I’m probably far more anti AI than average, for 95% of what it’s pushed for it’s completely useless, but that still leaves 5% that it’s genuinely useful for that some people refuse to accept.
It’s amazing that if you acknowledge that:
You are now an AI evangelist. Just as importantly, the level of investment into AI doesn’t justify #1. And when that realization hits business America, a correction will happen and the people who will be effected aren’t the well off, but the average worker. The gains are for the few, the loss for the many.
.
I feel this. In my line of work I really don’t like using them for much of anything (programming ofc, like 80% of Lemmy users) because it gets details wrong too often to be useful and I don’t like babysitting.
But when I need a logging message, or to return an error, it’s genuinely a time saver. It’s good at pretty well 5%, as you say.
But using it for art, math, problem solving, any of that kind of stuff that gets tauted around by the business people? Useless, just fully fuckin useless.
I don’t know about “art”, one part of ai image generation is of replacing stock images and erotic photos which frankly I don’t have a huge issue with as they’re both at least semi-exploitative industries anyway in many ways and you just need something that’s good enough.
Obviously these don’t extend to things a reasonable person would consider art, but business majors and tech bros rebranding something shitty to position it as a competitor to or in the same class as something it so obviously isn’t.
Yeah - I first hand have seen business majors I work with try to pitch a song from AI as our new marketing jingle. It was neither good, nor catchy for marketing purposes, but business ghouls hear something that sounds close enough to something someone put real effort into and think that’s the hard part sorted.
Name three.
I’m going to limit to LLMs as that’s the generally accepted term and there’s so many uses for AI in other fields that it’d be unfair.
Translation. LLMs are pretty much perfect for this.
Triaging issues for support. They’re useless for coming to solutions but as good as humans without the need to wait at sending people to the correct department to deal with their issues.
Finding and fixing issues with grammar. Spelling is something that can be caught by spell-checkers, but grammar is more context-aware, another thing that LLMs are pretty much designed for, and useful for people writing in a second language.
Finding starting points to research deeper. LLMs have a lot of data about a lot of things, so can be very useful for getting surface level information eg. about areas in a city you’re visiting, explaining concepts in simple terms etc.
Recipes. LLMs are great at saying what sounds right, so for cooking (not so much baking, but it may work) they’re great at spitting out recipes, including substitutions if needed, that go together without needing to read through how someone’s grandmother used to do xyz unrelated nonsense.
There’s a bunch more, but these were the first five that sprung to mind.
Right, except they suck at all of those things. Especially the last one. Unless you think glue is an acceptable pizza topping.
Nice, here’s a gold star for finding one case of it doing something wrong. I’ll call the CEO of AI and tell them to call it off, it’s a good thing humans have never said anything like that!
Bruh, you were the one that picked the examples. If you had a better argument you should have used that one instead.
And no matter what I picked, you’d reject them because you’re not actually considering them, you’re just either a troll, a contrarian or a luddite.
Riiiiight. Everyone who disagrees with you is an evil scary luddite. Sure fam.
Who said you were scary?
Frankly I pity you more than anything.
Translation. Only works for unified technical texts. The older non-LLM translation is still better for any general text and human translation for any fiction is a must. Case in point: try to translate Severance TV show transcript to another language. The show makes a heavy use of “Innie/Outie” language that does not exist in modern English. LLM fail to translate that - human translator would be able to find a proper pair of words in the target language.
Triaging issues for support. This one is a double-edged sword. Sure you can triage issues faster with LLM, but other people can also write issues faster with their LLMs. And they are winning more. Overall, LLM is a net negative on your triage cost as a business because while you can process each one faster than before, you are also getting way higher volume of those.
Grammar. It fails in that. I asked LLM about “fascia treatment” but of course I misspelled “fascia”. The “PhD-level” LLM failed to recognize the typo and gave me a long answer about different kinds of “facial treatment” even though for any human the mistake would’ve been obvious. Meaning, it only corrects grammar properly when the words it is working on are simple and trivial.
Starting points for deeper research. So was the web search. No improvement there. Exactly on-par with the tech from two decades ago.
Recipes. Oh, you stumbled upon one of my pet peeves! Recipes are generally in the gutter on the textual Internet now. Somehow a wrong recipe got into LLM training for a few things and now those mistakes are multiplied all over the Internet! You would not know the mistakes if you did not not cook/bake the thing previously. The recipe database was one of the early use cases for the personal computers back in 1990s and it is one of the first ones to fall prey to “innovation”. The recipes online are so bad, that you need an LLM to distill it back to manageable instructions. So, LLM in your example are great at solving the problem they created in the first place! You would not need LLM to get cooking instructions out of 1990s database. But early text generation AIs polluted this section of the Internet so much, that you need the next generation AI to unfuck it. Tech being great at solving the problem it created in the first place is not so great if you think about it.
You’re bringing up edge cases for #1, and it should be replacing google translate and basic human translation, eg allowing people to understand posts online or communicate textually with people with whom they don’t share a common language. Using it for anything high stakes or legal documents is asking for trouble though.
For 2, it’s not for AIs finding issues, it’s for people wanting to book a flight, or seek compensation for a delayed flight, or find out what meals will be served on their flight. Some people prefer to use text or voice communication over a UI, and this makes it easier to provide.
For 3, grammar and spelling are different. I said it wasn’t useful for spellcheck, but even then if you give it the right context it may or may not catch it. I was referring more to word order and punctuation positioning.
For 4, yeah for me it’s on par in terms of results, but much much faster, especially when asking followup questions or specifying constraints. A lot of people aren’t search engine powerusers though, so will find it significantly easier, faster and better than conventional search than having to manage tabs or keep track of what you’ve seen without just scrolling back up in the conversation.
For 5, recipes have been in the gutter for a decade or more now, SEO came before LLMs, but yeah, you’ve actually caught on to an obvious #6 I missed here of text summarisation…
What I’m getting overall though is that you’re not considering how tech-savvy the average person is, which absolutely makes them seem less useful as the more tech savvy you are, both the more you’re aware of their weaknesses and the less you benefit from the speedup by simplification they bring. This does make ai’s shortcomings more dangerous, but as it matures one would hope that it becomes common knowledge.
I think you are correct at the main point:
I am actually having hard time understanding where all of that hype is coming from. The first time I’ve seen AI solve a problem better than a human was back in 1996. I have used various generations of AI tools ever since. LLMs are fun, but it is not like they are that much different from the other AI tools before them. Every time a new AI technology comes around I am finding a use case for it in my own flow. LLMs have their uses as well. But I am not trying to solve ALL the problems with the new tech.
I do not understand “the average person”. And I guess I never will.
So if the AI can’t do it then that’s just proof that the AI is too smart to be able to do it? That’s your arguement is it. Nah, it’s just crap
You think just because you attached it to an analogy that makes it make sense. That’s not how it works, look I can do it.
My car is way too technologically sophisticated to be able to fly, therefore AI doesn’t need to be able to work out how many l Rs are in “strawberry”.
See how that made literally no sense whatsoever.
Except you’re expecting it to do everything. Your car is too “technically advanced” to walk on the sidewalk, but wait, you can do that anyway and don’t need to reinvent your legs
Well it also can’t code very well either
.
I feel like that was supposed to be an insult but because it made literally no sense whatsoever, I really can’t tell.
No not really, just an observation. It literally said you are a boring person. Not sure whats not to get.
Bye.
You need to get back on the dried frog pills.
Then why is Google using it for question like that?
Surely it should be advanced enough to realise it’s weakness with this kind of questions and just don’t give an answer.
They are using it for every question. It’s pointless. The only reason they are doing it is to blow up their numbers.
… they are trying to be infront. So that some future ai search wouldn’t capture their market share. It’s a safety thing even if it’s not working for all types of questions.
Ding ding ding.
It’s so they can have impressive metrics for shareholders.
“Our AI had n interactions this quarter! Look at that engagement!”, with no thought put into what user problems it actually solves.
It’s the same as web results in the Windows start menu. “Hey shareholders, Bing received n interactions through the start menu, isn’t that great? Look at that engagement!”, completely obfuscating that most of the people who clicked are probably confused elderly users who clicked on a web result without realising.
Line on chart must go up!
Yeah, but … they also can’t just do nothing and possibly miss out on something. Especially if they already invested a lot.
Understanding the bounds of tech makes it easier for people to gage its utility. The only people who desire ignorance are those that profit from it.
Sure. But you can literally test almost all frontier models for free. It’s not like there is some conspiracy or secret. Even my 73 year old mother uses it and knows it’s general limits.
Saying “it’s worth trillions of dollars huh” isn’t really promoting that attitude.
This reaction is because conmen are claiming that current generations of LLM technology are going to remove our need for experts and scientists.
We’re not demanding submersible cars, we’re just laughing about the people paying top dollar for the lastest electric car while plannig an ocean cruise.
I’m confident that there’s going to be a great deal of broken… everything…built with AI “assistance” during the next decade.
That’s not what you are doing at all. You are not laughing. Anti ai people are outraged, full of hatred and ready to pounce on anyone who isn’t as anti as they are. It’s a super emotional issue, especially on fediverse.
You may be confident, because you probably don’t know how software is built. Nobody is going to just abandon all the experience they have, vibe code something and release whatever. Thats not how it works.
Oh shit. Nevermind then.
Well technically cars can go underwater. They just cannot get out because they stop working.
Intentionally missing the point is not an argument in itself.
It’s very funny that you can get ChaptGPT to spell out the word (making each letter an individual token) and still be wrong.
Of course it makes complete sense when you know how LLMs work, but this demo does a very concise job of short-circuiting the cognitive bias that talking machine == thinking machine.
In Copilot terminology, this is a “quick response” instead of the “think deeper” option. The latter actually stops to verify the initial answer before spitting it out.
Deep thinking gave me this: Colorado, Delaware, Florida, Idaho, Indiana, Maryland, North Dakota, Rhode Island, and South Dakota.
It took way longer, but at least the list looks better now. Somehow it missed Nevada, so it clearly didn’t think deep enough.
“I asked it to burn an extra 2KWh of energy breaking the task up into small parts to think about it in more detail, and it still got the answer wrong”
Yeah that pretty much sums it up. Sadly, it didn’t tell me how much coal was burned and how many starving orphan puppies it had to stomp on to produce the result.
Gemini is trained on reddit data, what do you expect?
Honestly? Way more d.
You joke, but I bet you didn’t know that Connecticut contained a “d”
I wonder what other words contain letters we don’t know about.
The famous ‘invisible D’ of Connecticut, my favorite SCP.
SCP-00WTFDoC (lovingly called “where’s the fucking D of Connecticut” by the foundation workers, also “what the fuck, doc?”)
People think it’s safe, because it’s “just an invisible D”, not even a dick, just the letter D, and it only manifests verbally when someone tries to say “connecticut” or write it down. When you least expect it, everyone heard “Donnedtidut”, everyone read that thing and a portal to that fucking place opens and drags you in.
That actually sounds like a fun SCP - a word that doesn’t seem to contain a letter, but when testing for the presence of that letter using an algorithm that exclusively checks for that presence, it reports the letter is indeed present. Any attempt to check where in the word the letter is, or to get a list of all letters in that word, spuriously fail. Containment could be fun, probably involving amnestics and widespread societal influence, I also wonder if they could create an algorithm for checking letter presence that can be performed by hand without leaking any other information to the person performing it, reproducing the anomaly without computers.
ct -> d is a not-uncommon OCR fuck up. Maybe that’s the source of it’s garbage data?
No, LLMs produce the most statistically likely (in their training data) token to follow a certain list of tokens (there’s nothing remotely resembling reasoning going on in there, it’s pure hard statistics, with some error and randomness thrown in), and there are probably a lot more lists where Colorado is followed by Connecticut than ones where it’s followed by Delaware, so they’re obviously going to be more likely to produce the former.
Moreover, there aren’t going to be many texts listing the spelling of states (maybe transcripts of spelling bees?), so that information is unlikely to be in their training data, and they can’t extrapolate because it’s not really something they do and because they use words or parts of words as tokens, not letters, so they literally have no way of listing the letters of a word if said list is not in their training data (and, again, that’s not something we tend to write, and if we did we wouldn’t include d in Connecticut even if we were reading a misprint). Same with counting how many letters a word has, and stuff like that.
Words are full of mystery! Besides the invisible D, Connecticut has that inaudible C…
I hear the Invisible D and Silent C are happily married.
Every American I know does pronounce it like Connedicut 🤔
Really? Everyone I know calls it kinetic-cut. But I group up in new england.
That’s how I’ve always heard it pronounced on the rare occasions anybody ever mentions it. But I’ve never been to that part of the US so maybe the accents different there?
“Kinetic” with a hard “T” like posh Brit is saying it to the queen? Everyone I’ve ever heard speaking US English pronounces it with a rolled “t” like “kinedic” so the alternate pronunciation still reads like it’d have a “d” sound
This phenomenon is called “T flapping” and it is common in North American English. I got into an argument with my dad who insisted he pronounces the T’s in ‘butter’ when his dialect, like nearly all North Americans pronounces the word as ‘budder’.
budder is softer than t flapping. further forward with the tongue on the palate.
It’s an approximation, but the t is partially vocalized giving it a ‘d’ sound even if it’s not made exactly the same way.
i just thought we were getting technical about the linguistics. i got and use both words frequently, thought the distinction might be appreciated. the difference is so subtle we sometimes have to ask each other which one we’re referring to. i’m willing to bet it shows up more on my face than in my voice.
I appreciate the discussion, I get out of my depth pretty quickly on the topic being a linguistic hobbyist rather than someone with actual education and background.
Connedicut
I was going to make a joke if you’re from connedicut you never pronounce first d in the word. Conne-icut
The d in Connecticut is between the e and the i. They don’t connect because it was cut.
Connecticut is Jewish?
One of these days AI skeptics will grasp that spelling-based mistakes are an artifact of text tokenization, not some wild stupidity in the model. But today is not that day.
You aren’t wrong about why it happens, but that’s irrelevant to the end user.
The result is that it can give some hilariously incorrect responses at times, and therefore it’s not a reliable means of information.
A calculator app is also incapable of working with letters, does that show that the calculator is not reliable?
What it shows, badly, is that LLMs offer confident answers in situations where their answers are likely wrong. But it’d be much better to show that with examples that aren’t based on inherent technological limitations.
The difference is that Google decided this was a task best suited for their LLM.
If someone seeked out an LLM specifically for this question, and Google didn’t market their LLM as an assistant that you can ask questions, you’d have a point.
But that’s not the case, so alas, you do not have a point.
“It”? Are you conflating the low parameter model that Google uses to generate quick answers with every AI model?
Yes, Google’s quick answer product is largely useless. This is because it’s a cheap model. Google serves billions of searches per day and isn’t going to be paying premium prices to use high parameter models.
You get what you pay for, and nobody pays for Google so their product produces the cheapest possible results and, unsurprisingly, cheap AI models are more prone to error.
Yes, it. It’s not a person. Were you expecting me to call it anything else?
Mmh, maybe the solution than is to use the tool for what it’s good, within it’s limitations.
And not promise that it’s omnipotent in every application and advertise/ implement it as such.
Mmmmmmmmmmh.
As long as LLMs are built into everything, it’s legitimate to criticise the little stupidity of the model.
ChatGPT is just as stupid.<img alt="" src="https://sh.itjust.works/pictrs/image/52926e28-d1bf-40fb-93ff-6dbbe360995c.png">
it’s actually getting dumber.
Yesterday i asked Claude Sonnet what was on my calendar (since they just sent a pop up announcing that feature)
It listed my work meetings on Sunday, so I tried to correct it…
Just today when I asked what’s on my calendar it gave me today and my meetings on the next two thursdays. Not the meetings in between, just thursdays.
Something is off in AI land.
Edit: I asked again: gave me meetings for Thursday’s again. Plus it might think I’m driving in F1
A few weeks ago my Pixel wished me a Happy Birthday when I woke up, and it definitely was not my birthday. Google is definitely letting a shitty LLM write code for it now, but the important thing is they’re bypassing human validation.
Stupid. Just stupid.
pixel?
have you heard ~about grapheneOS tho…~Also, Sunday September 15th is a Monday… I’ve seen so many meeting invites with dates and days that don’t match lately…
Yeah, it said Sunday, I asked if it was sure, then it said I’m right and went back to Sunday.
I assume the training data has the model think it’s a different year or something, but this feature is straight up not working at all for me. I don’t know if they actually tested this at all.
Sonnet seems to have gotten stupider somehow.
Opus isn’t following instructions lately either.
We’ve used the Google AI speakers in the house for years, they make all kinds of hilarious mistakes. They also are pretty convenient and reliable for setting and executing alarms like “7AM weekdays”, and home automation commands like “all lights off”. But otherwise, it’s hit and miss and very frustrating when they push an update that breaks things that used to work.
Just another trillion, bro.
Behold the most expensive money burner!
Just another 1.21 jigawatts of electricity, bro. If we get this new coal plant up and running, it’ll be enough.
Hey look the markov chain showed its biggest weakness (the markov chain)!
In the training data, it could be assumed by output that Connecticut usually follows Colorado in lists of two or more states containing Colorado. There is no other reason for this to occur as far as I know.
Markov Chain based LLMs (I think thats all of them?) are dice-roll systems constrained to probability maps.
Edit: just to add because I don’t want anyone crawling up my butt about the oversimplification. Yes. I know. That’s not how they work. But when simplified to words so simple a child could understand them, its pretty close.
Oh l I was thinking it’s because people pronounce it Connedicut
Awe cute!
I was wondering if you’d get similar results for states with the letter R, since there’s lots of prior art mentioning these states as either “D” or “R” during elections.
So the Dakotas get a pass
And Idaho
Connecticut do have a D in it: mine.
Sure now list the trillion other things that tech can do.
Have a 40% accuracy on any type of information it can produce? Not handle 2 column pages in its training data, resulting in dozens of scientific papers including references to nonsense pseudoscience words? Invent an entirely new form of slander that its creators can claim isn’t their fault to avoid getting sued in court for it?
Well, for anyone who knows a bit about how LLMs work, it’s pretty obvious why LLMs struggle with identifying the letters in the words
Well go on…
They don’t look at it letter by letter but in tokens, which are automatically generated separately based on occurrence. So while ‘z’ could be it’s own token, ‘ne’ or even ‘the’ could be treated as a single token vector. of course, ‘e’ would still be a separate token when it occurs in isolation. You could even have ‘le’ and ‘let’ as separate tokens, afaik. And each token is just a vector of numbers, like 300 or 1000 numbers that represent that token in a vector space. So ‘de’ and ‘e’ could be completely different and dissimilar vectors.
so ‘delaware’ could look to an llm more like de-la-w-are or similar.
of course you could train it to figure out letter counts based on those tokens with a lot of training data, though that could lower performance on other tasks and counting letters just isn’t that important, i guess, compared to other stuff
Good read. Thank you
Of course, when the question asks “contains the letter _” you might think an intelligent algorithm would get off its tokens and do a little letter by letter analysis. Related: ChatGPT is really bad at chess, but there are plenty of algorithms that are super-human good at it.
Wouldn’t that only explain errors by omission? If you ask for a letter, let’s say D, it would omit words containing that same letter when in a token in conjunction with more letters, like Da, De, etc, but how would it return something where the letter D isn’t even in the word?
Well each token has a vector. So ‘co’ might be [0.8,0.3,0.7] just instead of 3 numbers it’s like 100-1000 long. And each token has a different such vector. Initially, those are just randomly generated. But the training algorithm is allowed to slowly modify them during training, pulling them this way and that, whichever way yields better results during training. So while for us, ‘th’ and ‘the’ are obviously related, for a model no such relation is given. It just sees random vectors and the training reorganizes them tho slowly have some structure. So who’s to say if for the model ‘d’, ‘da’ and ‘co’ are in the same general area (similar vectors) whereas ‘de’ could be in the opposite direction. Here’s an example of what this actually looks like. Tokens can be quite long, depending how common they are, here it’s ones related to disease-y terms ending up close together, as similar things tend to cluster at this step. You might have an place where it’s just common town name suffixes clustered close to each other.
and all of this is just what gets input into the llm, essentially a preprocessing step. So imagine someone gave you a picture like the above, but instead of each dot having some label, it just had a unique color. And then they give you lists of different colored dots and ask you what color the next dot should be. You need to figure out the rules yourself, come up with more and more intricate rules that are correct the most. That’s kinda what an LLM does. To it, ‘da’ and ‘de’ could be identical dots in the same location or completely differents
plus of course that’s before the llm not actually knowing what a letter or a word or counting is. But it does know that 5.6.1.5.4.3 is most likely followed by 7.7.2.9.7(simplilied representation), which when translating back, that maps to ‘there are 3 r’s in strawberry’. it’s actually quite amazing that they can get it halfway right given how they work, just based on ‘learning’ how text structure works.
but so in this example, us state-y tokens are probably close together, ‘d’ is somewhere else, the relation between ‘d’ and different state-y tokens is not at all clear, plus other tokens making up the full state names could be who knows where. And tien there’s whatever the model does on top of that with the data.
for a human it’s easy, just split by letters and count. For an llm it’s trying to correlate lots of different and somewhat unrelated things to their ‘d-ness’, so to speak
Thank you very much for taking your time to explain this. if you don’t mind, do you recommend some reference for further reading on how llms work internally?
You could look up 3Blue1Brown’s explainers on YouTube, they are pretty good and shows a lot of visual examples. He has a lot of other videos on other areas of math.
I’ll check it later, thanks
For the byte pair encoding (how those tokens get created) i think bpemb.h-its.org does a good job at giving an overview. after that i’d say self attention from 2017 is the seminal work that all of this is based on, and the most crucial to understand. jtlicardo.com/blog/self-attention-mechanism does a good job of explaining it. And jalammar.github.io/illustrated-transformer/ is probably the best explanation of a transformer architecture (llms) out there. Transformers are made up of a lot of self attention.
it does help if you know how matrix multiplications work, and how the backpropagation algorithm is used to train these things. i don’t know of a good easy explanation off the top of my head but xnought.github.io/backprop-explainer/ looks quite good.
and that’s kinda it, you just make the transformers bigger, with more weight, pluck on a lot of engineering around them, like being able to run code and making it run more efficientls, exploit thousands of poor workers to fine tune it better with human feedback, and repeat that every 6-12 month for ever so it can stay up to date.
Thank you very much
Con-ned-di-cut
Which is State contains 狄? They use a different alphabet, so understanding ours is ridiculous.
Click bait post that cherry picks bad output to say certain technology has no potential because it thinks he smarter than everybody else with 4+years of higher education.
It doesn’t have the potential they market it to have, and to be useful in all the human-replacing ways they claim it is.
That’s what is bad about it.
Connedicut.
Close. We natives pronounce it ‘kuh ned eh kit’
So does everyone else
We’re turfing out students by the tens on academic misconduct. They are handing in papers with references that clearly state “generated by Chat GPT”. Lazy idiots.
This is why invisible watermarking of AI-generated content is likely to be so effective. Even primitive watermarks like file metadata. It’s not hard for anyone with technical knowledge to remove, but the thing with AI-generated content is that anyone who dishonestly uses it when they are not supposed to is probably also too lazy to go through the motions of removing the watermarking.
if you are going to do all that, just do the research and learn something.
Aye that’s exactly the same thing that I said
Couldn’t students just generate a paper with ChatGPT, open two windows wide by side and then type it out in a word document?
Depends on the watermark method used. Some people talk about watermarking by subtly adjusting the words used. Like if there’s 5 synonyms and you pick the 1st synonym, next word you pick the 3rd synonym. To check the watermark you have to access to the model and probabilities to see if it matches that. The tricky part about this is that the model can change and so can the probabilities and other things I don’t fully understand.
Students view doing that as basically the same amount of work as writing the paper yourself
but that’s work.
I think I’d at least use an OCR program to do the bulk of the typing for me…
Huh that actually does sound like a good use-case of LLMs. Making it easier to weed out cheaters.
“This is the technology worth trillions of dollars”
You can make anything fly high in the sky with enough helium, just not for long.
(Welcome to the present day Tech Stock Market)
Bubbles and crashes aren’t a bug in the financial markets, they’re a feature. There are whole legions of investors and analysts who depend on them. Also, they have been a feature of financial markets since anything resembling a financial market was invented.
Listen, we just have to boil the ocean five more times.
Then it will hallucinate slightly less.
Or more. There’s no way to be sure since it’s probabilistic.
If you want to get irate about energy usage, shut off your HVAC and open the windows.
Worthless comment.
Even more worthless than mine, somehow.
sounds reasonable… i’ll just go tell large parts of australia where it’s a workplace health and safety issue to be out of AC for more than 15min during the day that they should do their bit for climate change and suck it up… only a few people will die
maybe people shouldn’t live there then?
of course you’re right! we should just shut down some of the largest mines in the world
i foresee no consequences from this
(related note: south australia where one of the largest underground mines in the world is, largely gets its power from renewables)
people should probably move from canada and most of the north of the USA too: far too cold up there during winter
I get the sentiment behind this post, and it’s almost always funny when LLM are such dumbass. But this is not a good argument against the technology. It is akin to climate change denier using the argument: “look! It snowed today, climate change is so dumb huh ?”
AI writes code for me. It makes dumbass mistakes that compilers automatically catch. It takes three or four rounds to correct a lot of random problems that crop up. Above all else, it’s got limited capacity - projects beyond a couple thousand lines of code have to be carefully structured and spoonfed to it - a lot like working with junior developers. However: it’s significantly faster than Googling for the information needed to write the code like I have been doing for the last 20 years, it does produce good sample code (if you give it good prompts), and it’s way less frustrating and slow to work with than a room full of junior developers.
That’s not saying we fire the junior developers, just that their learning specializations will probably be very different from the ones I was learning 20 years ago, just as those were very different than the ones programmers used 40 and 60 years ago.
I agree, cursor and other IDE integration have been a game changer. It made it way easier for a certain range of problems we used to have in software dev. And for every easy code, like prototyping, or inconsequential testing, it’s so so fast. What I found is that, it is particularly efficient at helping you do stuff you would have been able to do alone, and are able to check once it’s done. Need to be careful when asking stuff you aren’t familiar with though, cause it will comfortably lead you toward a mistake that will waste your time.
Though one thing I have to say: I’m very annoyed by it’s constant agreeing with what I say, and enabling me when I’m doing dumb shit. I wish it would challenge me more and tell me when I’m an idiot.
“Yes you are totally right”, “This is a very common issue that everybody has”, “What a great and insightful question”… I’m so tired of this BS.
There’s a balance to be had there, too… I have been comparing a few AI engines to compare their code generation capabilities. If you want an exercise in frustration, try to make an old school keypress driven application on a modern line-oriented terminal interface while still using the terminal for standard text output. I got pretty far with Claude, then my daily time limits were kicking in. Claude did all that “you’re so right” ego stroking garbage, but also got me near to a satisfactory solution. Then I moved into Google AI and it started out with reading my the “you just can’t do that, it won’t work” doom and gloom it got from some downer stack overflow or similar material. Finally, I showed Google my code that was already doing what it was calling impossible and it started helping me to polish the remaining rough spots. But, if you believed its first line answers you’d walk away thinking that something relatively simple was simply impossible.
Lately, I have taken to writing my instructions in a requirements document instead of relying so much on interactive mode. It’s not a perfect approach, but it seems to be much more stable for “larger” projects where you hit the chat length limits and have to start over with the existing code - what you’ve captured in requirements tends to stick around better than just using the existing code as a starting point of how things should be then adding/modifying from there. Ideally, I’d like it if the engine could just take my requirements document and make the app from that, but Claude still seems to struggle when total LOC gets into the 2000-5000 range for a 200-ish lines requirement spec.
You do know that AI is (if not already) fast approaching a leading CAUSE of climate change?
While the environmental impact of AI is absolutely horrible I don’t think it is even in the top 10 of industries. Meat production, Transportation by cars, Airplanes, plastic products etc are all much worse.
The problem is AI is absolutely useless for how big its climate impact is. The other industries at least provide value.
Your opinion isn’t invalid, it’s just incomplete
www.allaboutai.com/resources/…/ai-environment/
Combining your source with this ourworldindata.org/emissions-by-sector
Well i wasnt wrong in the assumption that AI is absolutely dwarfed by other industries, agriculture and energy production, but it is in the top 10, on the same level as aviation (so like place 9)
Yes, I know it has an impact, though not as big as you make it seem, (and so is everything). When you divide it to calculate the personal impact, it is way lower than a huge number of other stuff. I agree that we need to address climate change, but I don’t believe this should be the main focus.
Also, every individual should be able to choose how they spend their “carbon allocation”, personally, I don’t eat meat, I never take the plane, I don’t own a car and do everything using bike and trains, my house is carbon negative (building it actually had a negative carbon footprint) which was a huge sacrifice I had to compromise getting a way way smaller house for way more debt than if I had built a cheap standard house (and of course I’m in debt for decade). LLM makes me more efficient at my job so I think I can afford the carbon footprint that comes with it which, as I said, is not as big per individual as you make it appear.
I understand that hanging on Lemmy makes it seem like AI/LLM is the worse thing that has happened to mankind, but it’s really not, there are lots of issues with it, sure. But there is worse stuff to worry about.
I want to finish by saying that I DO support your action to minimize its impact, what you are doing overall is important and necessary, but I think you should revise the individual argument you put up against LLM, cause this one is not great.
It’s not worth the environmental impact
It's a pretty good argument against the technology, at least as it currently stands. This was a trivial question where anybody with a basic reading ability can see it's just completely wrong, the problem comes when you ask it a question you don't already know the answer to and can't easily check and it give equally wrong answers.
Blows my mind people pay money for wrong answers.
Connedicut.
I wondered if this has been fixed. Not only has it not, the AI has added Nebraska.
What about Our Kansas? Cause according to Google Arkansas has one o in it. Refreshing the page changes the answer though.
Just checked, it sure does say that! AI spouting nonsense is nothing new, but it’s pretty ironic that a large language model can’t even parse what letters are in a word.
Well I mean it’s a statistics machine with a seed thrown in to get different results on different runs. So really, it models the structure of language, but not the meaning. Kinda useless.
It’s because, for the most part, it doesn’t actually have access to the text itself. Before the data gets to the “thinking” part of the network, the words and letters have been stripped out and replaced with vectors. The vectors capture a lot of aspects of the meaning of words, but not much of their actual text structure.
I would assume it uses a different random seed for every query. Probably fixed sometimes, not fixed other times.
You mean Connecdicud.
✅ Colorado
✅ Connedicut
✅ Delaware
❌ District of Columbia (on a technicality)
✅ Florida
But not
❌ I’aho
❌ Iniana
❌ Marylan
❌ Nevaa
❌ North Akota
❌ Rhoe Islan
❌ South Akota
Everyone knows it’s properly spelled “I, the ho” not Idaho. That’s why it didn’t make the list.
Gosh tier comment.
You just described most of my post history.
They took money away from cancer research programs to fund this.
After we pump another hundred trillion dollars and half the electricity generated globally into AI you’re going to feel pretty foolish for this comment.
Just a couple billion more parameters, bro, I swear, it will replace all the workers
only cancer patients benefit from cancer research, CEOs benefit from AI
Tbf cancer patients benefit from AI too tho a completely different type that’s not really related to LLM chatbot AI girlfriend technology used in these.
Well as long as we still have enough money to buy weapons for that one particular filthy genocider country in the middle east, we’re fine.
i rather manually search for info
Connecdicud.
I would estimate that Google’s AI is helpful and correct about 7% of the time, for actual questions I’d like the answer to.
I don’t think this gets nearly enough visibility: www.academ-ai.info
Papers in peer-reviewed journals with (extremely strong) evidence of AI shenanigans.
Thanks for sharing! I clicked on it with cynicism around how easily we could detect AI usage with confidence vs. risking making false allegations, but every single example on their homepage is super clear and I have no doubts - I’m impressed! (and disappointed)
Yup. I had exactly the same trepidation, and then it was all like “As an AI model, I don’t have access to the data you requested, however here are some examples of…”
I have more contempt for the peer reviewers who let those slide into major journals, than for the authors. It’s like the Brown M&M test; if you didn’t spot that blatant howler then no fucking way did you properly check the rest of the paper before waving it through. The biggest scandal in all this isn’t that it happened, it’s that the journals involved seem to be almost never retracting them upon being reported.
With enough duct tape and chewed up bubble gum, surely this will lead to artificial general intelligence and the singularity! Any day now.
Hurry MacGruber! We’re almost out of…BOOM!
.
It ripped off this famous poem in the process:
Most States
So this is the terminator consciousness so many people are scared will kill us all…
Stop using Google search, easy as that! I use duckduckgo and I have turned off AI prompts.
GitLab Enterprise somewhat recently added support for Amazon Q (based on claude) through an interface they call “GitLab Duo”. I needed to look up something in the GitLab docs, but thought I’d ask Duo/Q instead (the UI has this big button in the top left of every screen to bring up Duo to chat with Q):
(Paraphrasing…)
ME: How do I do X with Amazon Q in GitLab? Q: Open the Amazon Q menu in the GitLab UI and select the appropriate option.
ME: [:looks for the non-existant menu:] ME: Where in the UI do I find this menu?
Q: My last response was incorrect. There is no Amazon Q button in GitLab. In fact, there is no integration between GitLab and Amazon Q at all.
ME: [:facepalm:]
Lol @ these fucking losers who think AI is the current answer to any problems
Third time’s the charm! They have to keep the grift going after Blockchain and NFT failed with the general public.
@arararagi Don't forget Metaverse, they took a fuckin bath on that.
As long as there’s something to sell for untalented morons to feel intelligent & talented; they’ll take the bait.
Funny thing is, the metaverse as their pictured it failed, but vrchat itself had it’s biggest spike this year.
AI will most likely create new problems in the future as it eats up electricity like a world eater, so I fear that soon these non-humans will only turn on electricity for normal people for a few hours a day instead of the whole day to save energy for the AI.
I’m not sure about this of course, but it’s quite possible.
Nothing will stop them, they are so crazy that they can turn nonsense into reality, believe me.
Or to put it more simply – They need power for the sake of power itself, there is nothing higher.
This is the perfect time for LLM-based AI. We are already dealing with a significant population that accepts provable lies as facts, doesn’t believe in science. and has no concept of what hypocrisy means. The gross factual errors and invented facts of current AI couldn’t possibly fit in better.