Why is no one talking about how unproductive it is to have verify every "hallucination" ChatGPT gives you?
Why is no one talking about how unproductive it is to have verify every "hallucination" ChatGPT gives you?
You're viewing a single thread.
I just tried out Gemini.
I asked it several questions in the form of 'are there any things of category x which also are in category y?' type questions.
It would often confidently reply 'No, here's a summary of things that meet all your conditions to fall into category x, but sadly none also fall into category y'.
Then I would reply, 'wait, you don't know about thing gamma, which does fall into both x and y?'
To which it would reply 'Wow, you're right! It turns out gamma does fall into x and y' and then give a bit of a description of how/why that is the case.
After that, I would say '... so you... lied to me. ok. well anyway, please further describe thing gamma that you previously said you did not know about, but now say that you do know about.'
And that is where it gets ... fun?
It always starts with an apology template.
Then, if its some kind of topic that has almost certainly been manually dissuaded from talking about, it then lies again and says 'actually, I do not know about thing gamma, even though I just told you I did'.
If it is not a topic that it has been manually dissuaded from talking about, it does the apology template and then also further summarizes thing gamma.
...
I asked it 'do you write code?' and it gave a moderately lengthy explanation of how it is comprised of code, but does not write its own code.
Cool, not really what I asked. Then command 'write an implementation of bogo sort in python 3.'
... and then it does that.
...
Awesome. Hooray. Billions and billions of dollars for a shitty way to reform web search results into a coversational form, which is very often confidently wrong and misleading.
Idk why we have to keep re-hashing this debate about whether AI is a trustworthy source or summarizer of information when it's clear that it isn't - at least not often enough to justify this level of attention.
It's not as valuable as the marketing suggests, but it does have some applications where it may be helpful, especially if given a conscious effort to direct it well. It's better understood as a mild curiosity and a proof of concept for transformer-based machine learning that might eventually lead to something more profound down the road but certainly not as it exists now.
What is really un-compelling, though, is the constant stream of anecdotes about how easy it is to fool into errors. It's like listening to an adult brag about tricking a kid into thinking chocolate milk comes from brown cows. It makes it seem like there's some marketing battle being fought over public perception of its value as a product that's completely detached from how anyone actually uses or understands it as a novel piece of software.
Probably it keeps getting rehashed because people who actually understand how computers work are extremely angry and horrified that basically every idiot executive believes the hype and then asks their underlings to inplement it, and will then blame them for doing what they asked them to do when it turns out their idea was really, unimaginably stupid, but idiot executive gets golden parachute and software person gets fired.
That, and/or the widespread proliferation of this bullshit is making stupid people more stupid, and just making more people stupid in general.
Or how like all the money and energy spent on this is actively murdering the environment and dooming the vast majority of our species, when it could be put toward building affordable housing or renovating crumbling infrastructure.
Don't worry, if we keep throwing exponential increasing amounts of effort at the thing with exponentially diminishing returns, eventually it'll become God!
Then why are we talking about someone getting it to spew inaccuracies in order to prove a point, rather than the decision of marketing execs to proliferate its use for a million pointless implementations nobody wants at the expense of far higher energy usage?
Most people already know and understand that it's bad at most of what execs are trying to push it as, it's not a public-perception issue. We should be talking about how energy-expensive it is, and curbing its use on tasks where it isn't anything more than an annoying gimmick. At this point, it's not that people don't understand its limitations, it's that they don't understand how much energy it's costing and how it's being shoved into everything we use without our noticing.
Somebody hopping onto openAI or Gemini to get help with a specific topic or task isn't the problem. Why are we trading personal anecdotes about sporadic personal usage when the problem is systemic, not individualized?
people who actually understand how computers work
Bit idea for moderators: there should be a site or community-wide auto-mod rule that replaces this phrase with 'eat all their vegitables' or something that is equally un-serious and infantilizing as 'understand how computers work'.
You original comment is posted under mine.
I am going to assume you are responding to that.
... I wasn't trying to trick it.
I was trying to use it.
This is relevant to my more recent reply to you... because it is an anecdotal example of how broadly useless this technology is.
...
I wasn't aware the purpose of this joke meme thread was to act as a policy workshop to determine an actionable media campaign aimed at generating mass awareness of the economic downsides of LLMs, which wouldn't fucking work anyway because LLMs are being pushed by a class of wealthy people who do not fucking care what the masses think, and have essentially zero reason at all to change their course of action.
What, we're going to boycott the entire tech industry?
Vote them out of office?
These people are on video, on record saying basically, 'eh, we're not gonna save the climate, not happening, might as well burn it all down even harder, even faster, for a tiny percentage chance our overcomplicated autocomplete algorithm magically figures out how to fix everything afterward'.
...
And yes, I very intentionally used the phrase 'understand how computers actually work' to infantilize and demean corporate executives.
Because they are narcissistic priveleged sociopaths who are almost never qualified, almost always make idiotic decisions that will only benefit themselves and an increasingly shrinking number of people at the expense of the vast majority of people who know more and work harder than they do, and who often respond like children having temper tantrums when they are justly criticized.
Again, in the context of a joke meme thread.
Please get off your high horse, or at least ride it over to a trough of water if you want a reasonable place to try to convince it to drink in the manner in which you prefer.
… I wasn’t trying to trick it.
I was trying to use it.
Err, I'd describe your anecdote more as an attempt to reason with it...? If you were using google to search for an answer to something and it came up with the wrong thing, you wouldn't then complain back to it about it being wrong, you'd just try again with different terms or move on to something else. If 'using' it for you is scolding it as if it's an incompetent coworker, then maybe the problem isn't the tool but how you're trying to use it.
I wasn’t aware the purpose of this joke meme thread was to act as a policy workshop to determine an actionable media campaign
Lmao, it certainly isn't. Then again, had you been responding with any discernible humor of your own I might not have had reason to take your comment seriously.
And yes, I very intentionally used the phrase ‘understand how computers actually work’ to infantilize and demean corporate executives.
Except your original comment wasn't directed at corporate executives, it appears to be more of a personal review of the tool itself. Unless your boss was the one asking you to use Gemini? Either way, that phrase is used so much more often as self-aggrandizement and condescension that it's hard to see it as anything else, especially when it follows an anecdote of that person trying to reason with a piece of software lmao.
It is not that it responded "Sorry, I cannot find anything like what you described, here are some things that are pretty close."
It affirmatively said "No, no such things as you describe exist, here are some things that are pretty close."
There's a huge difference between a coworker saying "Dang man, I dunno, I can't find a thing like that." and "No, nothing like that exists, closest to it is x y z,"
The former is honest. The latter is confidently incorrect.
Combine that with "Wait what about gamma?"
And the former is still honest, and the latter, who now describes gamma in great detail and how it meets my requirements, is now an obvious liar, after telling me that nothing like that exists.
If I now know I am dealing with a dishonest interlocutor, now I am forced to consider tricking it into being homest.
Or, if I am less informed or more naive, I might just, you know, believe it the first time.
A standard search engine that is not formatted to resemble talking to a person does not prompt a user to expect it to act like a person, and thus does not suffer from this problem.
If you don't find what you're looking for, all that means is you did not find it.
If you are told that no such thing exists, a lot of people are going to believe that no such thing exists.
That is typically called spreading disinformation, when the actor knows what they are claiming is false.
Its worse than unhelpful, it actively spreads lies.
...
Anyway, I'm sorry that you don't see humor in multi billion dollar technology failing at achieving its purported abilities, I laugh all the time at poorly designed products, systems, things.
...
Finally, I did not use the phrase in contention in my original post.
I used it in my response to you, specifically and only within a single sentence which revolved around incompetent executives.
...
It appears that reading comprehension is not your strong suit, maybe you can ask Gemini about how to improve it.
Err, well, maybe don't do that.
reading comprehension
Lmao, there should also be an automod rule for this phrase, too.
There’s a huge difference between a coworker saying [...]
Lol, you're still talking about it like it's a person that can be reasoned with bud. It's just a piece of software. If it doesn't give you the response you want you can try using a different prompt, just like if google doesn't find what you're looking for you can change your search terms.
If people are gullible enough to take its responses as given (or scold it for not being capable of rational thought lmao) then that's their problem - just like how people can take the first search result from google without scrutiny if they want to, too. There's nothing especially problematic about the existence of an AI chatbot that hasn't been addressed with the advent of every other information technology.
to fool into errors
tricking a kid
I've never tried to fool or trick AI with excessively complex questions. When I tried to test it (a few different models over some period of time - ChatGPT, Bing AI, Gemini) I asked stuff as simple as "what's the etymology of this word in that language", "what is [some phenomenon]". The models still produced responses ranging from shoddy to absolutely ridiculous.
completely detached from how anyone actually uses
I've seen numerous people use it the same way I tested it, basically a Google search that you can talk with, with similarly shit results.
Why do we expect a higher degree of trustworthiness from a novel LLM than we de from any given source or forum comment on the internet?
At what point do we stop hand-wringing over llms failing to meet some perceived level of accuracy and hold the people using it responsible for verifying the response themselves?
Theres a giant disclaimer on every one of these models that responses may contain errors or hallucinations, at this point I think it's fair to blame the user for ignoring those warnings and not the models for not meeting some arbitrary standard.
Why do we expect a higher degree of trustworthiness from a novel LLM than we de from any given source or forum comment on the internet?
The stuff I've seen AI produce has sometimes been more wrong than anything a human could produce. And even if a human would produce it and post it on a forum, anyone with half a brain could respond with a correction. (E.g. claiming that an ordinary Slavic word is actually loaned from Latin.)
I certainly don't expect any trustworthiness from LLMs, the problem is that people do expect it. You're implicitly agreeing with my argument that it is not just that LLMs give problematic responses when tricked, but also when used as intended, as knowledgeable chatbots. There's nothing "detached from actual usage" about that.
At what point do we stop hand-wringing over llms failing to meet some perceived level of accuracy and hold the people using it responsible for verifying the response themselves?
at this point I think it’s fair to blame the user for ignoring those warnings and not the models for not meeting some arbitrary standard
This is not an either-or situation, it doesn't have to be formulated like this. Criticising LLMs which frequently produce garbage is in practice also directed at people who do use them. When someone on a forum says they asked GPT and paste its response, I will at the very least point out the general unreliability of LLMs, if not criticise the response itself (very easy if I'm somewhat knowledgeable about the field in question). This is practically also directed at the person who posted that, such as e.g. making them come off as naive and uncritical. (It is of course not meant as a real personal attack, but even a detached and objective criticism has a partly personal element to it.)
Still, the blame is on both. You claim that:
Theres a giant disclaimer on every one of these models that responses may contain errors or hallucinations
I don't remember seeing them, but even if they were there, the general promotion and ways in which LLMs are presented in are trying to tell people otherwise. Some disclaimers are irrelevant for forming people's opinions compared to the extensive media hype and marketing.
Anyway my point was merely that people do regularly misuse LLMs, and it's not at all difficult to make them produce crap. The stuff about who should be blamed for the whole situation is probably not something we disagree about too much.
The stuff I’ve seen AI produce has sometimes been more wrong than anything a human could produce. And even if a human would produce it and post it on a forum, anyone with half a brain could respond with a correction.
Seems like the problem is that you're trying to use it for something it isn't good or consistent at. It's not a dictionary or encyclopedia, it's a language model that happens to have some information embedded. It's not built or designed to retrieve information from a knowledge bank, it's just there to deconstruct and reconstruct language.
When someone on a forum says they asked GPT and paste its response, I will at the very least point out the general unreliability of LLMs, if not criticise the response itself (very easy if I’m somewhat knowledgeable about the field in question)
Same deal. Absolutely chastise them for using it in that way, because it's not what it's good for. But it's a bit of a frequency bias to assume most people are using it in that way, because those people are the ones using it in the context of social media. Those who use it for more routine tasks aren't taking responses straight from the model and posting it on lemmy, they're using it for mundane things that aren't being shared.
Anyway my point was merely that people do regularly misuse LLMs, and it’s not at all difficult to make them produce crap. The stuff about who should be blamed for the whole situation is probably not something we disagree about too much.
People misuse it because they think they can ask it questions as if it's a person with real knowledge, or they are using it precisely for it's convincing bullshit abilities. That's why I said it's like laughing at a child for giving a wrong answer or convincing them of a falsehood merely from passive suggestion - the problem isn't that the kid is dumb, it's that you're (both yourself and the person using it) going in with the expectation that they are able to answer that question or distinguish fact from fiction at all.
And then more money spent on adding that additional garbage filter to the beginning and the end of the process which certainly won't improve the results.
copilot did the same with basic math. just to test it I said "let's say I have a 10x6 rectangle. what number would I have to divide width and height by, in order to end up with a rectangle that's half the area?"
it said "in order to make it half, you should divide them by 2. so [pointlessly lengthy steps explaining the divisions]"
I said "but that would make the area 5x3 = 15 units which is not half the area of 60"
it said "you're right! in order to ... [fixing the answer to √2 using approximation"
I don't know if I said it then, or after some other fucking nonsense but when I said "you're useless" it had the fucking audacity to take offense and end the conversation!
like fuck off, you don't get to have fake pride if you don't have basic fake intelligence but use it in your description.
Its a perfect encapsulation of the corpo mindset:
Whatever I do is profound, meaningful, with endless possibilities for future greatness...
... even though I'm just talking out of my ass 99% of the time...
... and if you have the audacity, the nerve, to have a completely normal reaction when you determine that that is what I am doing, pshaw, how uncouth, I won't stand for your abuse!
...
They've done it. They've made a talking (not thinking) machine in their own image.
And it was not good.
You start a conversation you can't even finish it You're talkin' a lot, but you're not sayin' anything When I have nothing to say, my lips are sealed Say something once, why say it again?
Psycho Killer Qu'est-ce que c'est
please further describe thing gamma that you previously said you did not know about, but now say that you do know about.'
It's quite amusing to ask it about conspiracy theories. There's a huge amount in it's training set (not because the theories are true, just that they are often written about) that it has been dissuaded from discussing.
Cool, not really what I asked. Then command ‘write an implementation of bogo sort in python 3.’
… and then it does that.
Alright, but... it did the thing. That's a feature older search engines couldn't reliably perform. The output is wonky and the conversational style is misleading. But its not materially worse than sifting through wrong answers on StackExchange or digging through a stack of physical textbooks looking for Python 3 Bogo Sort IRL.
I agree AI has annoying flaws and flubs. And it does appear we're spending vast resources doing what a marginal improvement to Google five years ago could have done better. But this is better than previous implementations of search, because it gives you discrete applicable answers rather than a collection of dubiously associated web links.
But this is better than previous implementations of search, because it gives you discrete applicable answers rather than a collection of dubiously associated web links.
Except for when you ask it to determine if a thing exists by describing its properties, and then it says no such thing exists while providing a discrete response explaining in detail how there are things that have some, but not all of those properties...
... And then when you ask it specifically about a thing you already know about that has all those properties, it tells you about how it does exist and describes it in detail.
What is the point of a 'conversational search engine' if it cannot help you find information unless you already know about said information?!
The whole, entire point of formatting it into a conversational format is to trick people into thinking they are talking to an expert, an archivist with encyclopedaeic knowledge, who will give them accurate answers.
Yet it gatekeeps information that it does have access to but omits.
The format of providing a bunch of likely related links to a query is a format much more reminiscent of doing actual research, with no impression that you will immediately find what you want right away, that this is a tool to aide you in your research process.
This is only an improvement if you want to further unteach people how to do actual research and critical thinking.
Except for when you ask it to determine if a thing exists by describing its properties
Basic search can't answer that either. You're describing a task neither system is well equipped to accomplish.
With basic search, it is extremely obvious that that feature does not exist.
With conversational search, the search itself gaslights you into believing it has this feature, as it understands how to syntactically parse the question, and then answers it confidently with a wrong answer.
I would much rather buy a car that cannot fly, knowing it cannot fly, than a car that literally talks to you and tells you it can fly, and sometimes manages to glide a bit, but also randomly nose dives into the ground whilst airborne.
I don’t feel like off-the-cuff summaries by AI can replace web sites and detailed articles written by knowledgeable humans. Maybe if you’re looking for a basic summary of a topic.
I don’t feel like off-the-cuff summaries by AI can replace web sites and detailed articles written by knowledgeable humans
No. But that's not what a typical search result returns.
There's also no guarantee the "detailed articles" you get back are well-informed or correct. Lots of top search results are just ad copy or similar propaganda. YouTube, in particular, is rife with long winded bullshitters.
What you're looking for is a well-edited trustworthy encyclopedia, not a search engine.
Sure. It depends on the topic. It also depends how good you are at searching. Personally, I don’t have any difficulty finding quality websites via Google or DuckDuckGo. Sometimes it requires refining the search terms, and many people don’t know how to do that properly. For certain topics, I might not. However, i stand by my assertion that some statistically generated text which you can trust at all is less useful than a good article on a topic. Is Wikipedia useful also? Yes, it is.