608
Amazon's Hidden Chatbot Recommends Nazi Books and Lies About Amazon Working Conditions
(www.404media.co)
this post was submitted on 13 Mar 2024
608 points (95.1% liked)
Technology
57472 readers
3868 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
So this is the problem with AI, if you add guardrails you're a culture warrior 1984'ing the whole world, and if you don't now your tool will generate resumes with fake experience or recommend offensive books.
At the risk of sounding like a jackass, when do we start blaming people for asking for such things?
It's funny that this one does both at once. It lies about Amazon working conditions, meaning it probably has been censored in some way, but at the same time it is recommending Nazi books. Really shows Amazon's priorities when it comes to censorship.
At least Amazon is thinking of the shareholders.
Although if I was to shareholder of Amazon i would be wondering why we even stock Nazi textbooks. Morality aside, they cannot really be that much of a market for them and even if there was the PR hit probably isn't worth it.
If it turns out there is a market for them we should probably separate off that part of the business to avoid PR blunders.
No this isn’t really a problem with the technology, though of course LLMs are extremely flawed in fundamental ways, it is a problem with conservatives being babies and throwing massive tantrums about any guardrails being added even when they are next to cliffs with 200 foot drops.
Conservatives and libertarians (who control most of these companies) want to try to figure this all out for themselves and are hellbent on trying the “no moderation” strategy first and haven’t thought past that step. This is what conservatives and libertarians always do, they might as well be a character archetype in commedia dell'arte at this point.
We can’t have an adult conversation about racism, sexism, hate against trans people or really even the basic concept of systematic stereotypes and prejudices because conservatives refuse to stop running around screaming, making this a conversation with children where everything has to be extremely simplified and black and white and we have to patiently explain over and over again the basic concept of a systematic bias and argue that it even exists.
Then these same people turn around and vote for people who literally want to control what women do with their unfertilized eggs while they act with a straight face like they give af about individual liberties or freedoms.
LLMs are fundamentally vulnerable to bias, we have to design LLMs with that in mind and first and foremost carefully structure and curate the training data we train an LLM on so that bias is minimized. The very idea of even thinking about the complexities usually sends conservatives right to outbursts of “that sounds like tyranny!” because they honestly just don’t have any of the skill sets that say, a liberal arts education that values the humanities, might provide you that could allow you to think about how to best solve problems that can’t truly be fairly solved and require empathizing with different groups.
Of course, nobody who has the power at AI companies is thinking about this either but…
How to you curate training data to remove biases without introducing bias? That’s the key problem here. I don’t think it’s unreasonable to be opposed to trading one bias for another. At least the initial bias is based on reality.
Because anybody who has taken a couple of humanities classes, english classes, philosophy classes, journalism/political science classes or who has spent time critically evaluating art, historical accounts or really anything other than just numbers, code and spreadsheets.... understands intuitively that EVERYTHING human has bias.
It seems like a lot of conservatives and libertarians are jussssssst beginning to comprehend this and again and they want the conversation to be "BIAS BAD GET RID OF IT" because they are children who don't listen and want to throw a tantrum so we can't have an adult conversation with nuance.
We can't remove biases, believe me, human history is written with the countless stories of artists, scientists, kings, religious leaders... who all thought they could do shit like that. The point is you can't. Everything we create and do is biased, everything we create and make is political, these aren't absolutist statements meant to trivialize a critical nuanced conversation about bias or politics though. On the contrary I am calling attention to the vital nature of these topics as the actually HARD part of LLMs or social media. The programming, data manipulation, development of decentralized protocols etc... they are all nearly trivial details comparatively.
Computer science has to try to create imperfect solutions to the bias problem, but it would have a much easier time if it recognized how tiny this whole world of computer science still is compared to the immense amount of knowledge in the humanities produced by generations of artists and thinkers tackling the same problems.
We can't remove biases, but we still have to make better choices anyways.
Well put. I think tackling the bias will always be a challenge. It’s not that we shouldn’t, but how is the question.
I don’t know if any of the big public LLMs are trying to trim biases from their training data or are just trying to ad-hoc tackle it by injecting modifiers into the prompts.
That’s the biggest problem I have personally with LLMs is that they are untrustworthy and often give incorrect or blatantly false information.
Sometimes it can be frustrating when I run across the “I can’t do that because of ethics” on benign prompts that I felt like it shouldn’t have but I don’t think it’s been that big a deal.
When we talk about political conservatives being opposed to biased LLMs, it’s mostly because it won’t tell them that their harmful beliefs are correct
"What because I think Islam is inherently a violent religion now this chatbot is telling me I AM the one with violent and harmful beliefs??????" - some loser, maybe elon musk or maybe your uncle, who cares.
So more or less the same as with human interactions.
What good will blame do? We need robust ai detection solutions.
Is that even possible? Part of modern generative systems is that they're trying to output text like a human would. As soon as someone invents a tool like that, it'll just be used to train the next generation, to make it even more indistinguishable, and turning the whole thing into a cat and mouse game.
AI » GAN » AI » GAN
Turns out it wasn't monkeys on typewriters that wrote perfect Shakespeare stories, it was trillions of transistors in a war of attrition.
By replacing the typewriter with a lever, you could probably achieve a similar result using monkeys.
It literally exists today and you sit here typing bullshit like 'is that even possible?'
Get every manner of blocked.
They're also infamously terrible, being half-correct and prone to detecting non-native English speakers as being AI. To the point where at least one institutition decided to not use a detector.
You'd likely be equally as accurate guessing at random. Not to mention that at the end of the day, they're just recognising quirks in the generated text. It is not difficult to mask those quirks either by having the prompt put out text in a different style, or for an update to change the generated text, breaking the detection as well.
There is no definite, sure-fire way to determine that text is AI-generated or not. For all you might know, as an AI language model, I could have cooked this comment up using a billion probability nodes loaded up into a typewriter, as it does not go against OpenAI's policies on generating text.
Quite the spicy one aren't you. I see where you get your username from.
But yes, they exist, and so does the ability to defeat them by training the detection data into a new undetected model. Cat and mouse game, as they say.
Robust today, defeated ~~tomorrow~~ today.
So, what you're saying is you don't really understand how data security works then.
Because it's never a 'one and done', it's ALWAYS a cat and mouse game, ALWAYS.
Which is why antivirus companies push definition updates.
And now we do the same with AI detectors, or they become irrelevant.
This is where I get to lol and say you don't understand AI.
When a kernel privesc vuln 0day is found and reported or caught in a dump, it gets fixed. Unless it was improperly fixed, that particular vulnerability can't be exploited again.
But when it comes to AI, a GAN's job is to take the 'vulnerability' that was 'fixed' and train on it to exploit it again.
And again.
And again.
And again.
https://en.m.wikipedia.org/wiki/Generative_adversarial_network
It's funny how people can just link to a wikipedia article about a ten year old thought experiment and think its some kind of mic drop moment. The current AI paradigm is starting to hit its singularity curve and hardly anything that old is anything more than a novelty and largely not applicable to current models, ESPECIALLY when it comes to modeling,
We aren't seeing this kind of iteritave adversity being used in actual real world attacks, and it is largely useless to train on a patched vulnerability.
But I'm sure you already knew that, maybe your testing me?
...
https://github.com/search?q=generative+adversarial+network&type=repositories&s=updated&o=desc
Do you remember last year when OpenAI pulled its own AI detection tool because it was performing so poorly?
I forgot about that, but this article from 6 months ago comparing the effectiveness of major AI detection tools reminded me.
On the same day that we start blaming people for spilling hot McDonald's coffee on themselves
Oh good, the corporate propaganda worked.
Look I probably hate McDonald's more than any of you but as I see it:
While McDonald's:
It's clear after the case that McDonald's coffee was exceptionally hot, but there was clearly an irresponsible chain of decisions that led to her getting burnt.
The jury found that McDonald's was 80% responsible. I just think it's more like 50%.
Too bad you weren't on the jury, because you could have been educated on the situation and have a valid opinion on the subject.
Or you could actually look up real information and find out what exactly McDonald's was being punished for, and how they exacerbated the situation over and over.
Safety standards are not created with the expectation that everyone else acts perfectly safely, safety standards are designed that people can have accidents and not be maimed due to cost cutting from a giant, multinational corporation. McDonald's deserves that judgement, the media demonized the victim, and you are perpetuating that myth.
Your comments are like when a schoolchild who hasn't read the assigned material has to present in front of the class.
The coffee was hot enough to cause serious burns and fused her labia together and required skin grafts.
I've spilled freshly brewed coffee on myself at home and while it burned, I did not need any medical attention.
The woman did nothing wrong. The product sold by McDonald's was purposely overly dangerous. Had she known it was that fucking hot, she probably wouldn't have put it between her legs. McDonald's did not properly indicate that their coffee was abnormally hot to the point of causing severe safety risks.
It's like if I buy my fifth alcoholic drink in a night, and the bartender gives me straight everclear to drink. I know I'm taking a risk by having another drink, but I'm not expecting to drink nearly pure alcohol and get alcohol poisoning from it.
Your user name is appropriate, but probably not the way you intended.
Yeah, how dare we blame the people who knew about the problem and chose not to fix it until someone was horrifically burned? And can you believe the gall of that woman to dare to ask for her medical bills to be paid by the people who knowingly setup the conditions to horrifically burn her?
The nerve of some people. Smh my head
If I buy ice cream I expect it to be cold. If I buy coffee I expect it to be hot. If I buy a knife I expect it to be sharp and so on.
I don't want stores to start selling dull knives because out of thousand customers there's always one who instantly cuts themselves. We don't need to round every sharp corner in the world so that no one ever gets hurt again. If you can't handle a bewerage you know to be hot with care then that's on you.
What if you bought an ice cream cone and it was so cold that you had to get skin grafts to repair the damage to your lips? And not only that, but the owner of the store instructed the employees to make it that cold?
The McDonald’s woman’s risk calculation was probably “if this spills, it will hurt” and not, “if this spills, it will do permanent, significant damage to my body”.
Why would anyone prefer living in a world where some property of any item you buy could intentionally be set to dangerously extreme levels?
If I remember correctly, one of the main reasons she burned so bad was the type of pants she was wearing which soaked up the hot coffee and stuck to her skin. They were arguing that lowering the temperature certain amount would greatly increase the time for severe burns to happen, which is a fair point but would still come with its downsides. Coffee snobs like me want their cups fresh and hot. It doesn't make sense to me to prevent the coffee shop from serving me fresh coffee because someone spilled such coffee on herself. In certain cases we just gotta accept that addicents happen and to me this seems like one of them. Like I said; wouldn't make much sense to sell dull knives either.
Won’t someone think of the poor coffee snobs who want their coffee just so, but don’t want to have to make it themselves at home? We can’t sacrifice their ideal coffee temperature at public establishments just to save some people from horrific burns — people who are probably wearing the wrong pants anyway, and therefore are kind of at fault too, when you think about it.
You ran out of arguments so you just start acting like a prick instead? That's productive and very Lemmy of you.
No, it’s just that your response was absurd. You didn’t answer my question, rehashing some points that I had already addressed and not answering my question. I found your response repetitive, uncurious, and selfish to the point of parody.
I was starting to write a different kind of answer, and I thought, why am I indulging this person who isn’t even taking the time to process what I’ve already written?
My coffee maker makes coffee that is fresh and hot. So does my local shop. It still doesn't literally melt my skin if some gets on my hand. Dumbfuck.
It truly boggles my mind why people talk to others like this on the internet. If you're unable to disagree politely then you probably should just stay silent.
Fuck off.
Haha yeah they called the woman, told her to buy a coffee in the drive thru and handle it in the most careless way possible. Here's an idea: go buy a rake from Home Depot, set it on the ground, step on it, bonk your forehead, and sue Home Depot for damages. After all, they're selling a dangerous product and if you use it irresponsibly, that's on them.
All these armchair legal experts acting like her winning the case means she isn't partly to blame are hilarious.
Says the idiot clearly unfamiliar with the details of their cited reference. 🤦🏼♂️