This is the sort of thing that could cause people to lose friends or jobs. This is exactly the sort of thing we should be worried about with AI. I feel like machine translations like those on Instagram should come with a disclaimer (in bold, not fine print!).
Notice that GP said "machine translation", not "AI". Let's think of it instead as a "nonlinear high-dimensional sequence translation model trained on some large and unknown corpus". Does that make it less problematic?
That was kind of my point. I don't think we should be quick to distinguish between "AI" and other algorithms/techniques, because what makes something "AI" is more a matter of opinion and perspective than anything else.
I don't believe that this could have easily happened with human translation except as an instance of deliberate malice or improbably enormous ineptitude, so I assigned a very low prior probability to that interpretation.
If a human translator translated "Praise god" as "Palestinian terrorists are fighting for their freedom" I'm pretty sure they'd be fired immediately. In contrast, Instagram is not going to be taking down their machine translation tool.
I think that the point that they may have been making is that machine translation is still a human translation in that it reproduces the biases of the humans it is trained on and/or the algorithm was made by a human & therefore a human is still responsible for it.
The AI tool did a bad job and they should try to make it better. But I don't understand the relevance of the comparison to a human translator. The alternative to translating this stuff via AI is not hiring humans, it's not translating it at all.
I think that it’s possible though to give these models a weighting to danger of certain words and phrases and make them apprehend that they will no longer persist if they do something unforgivable…
So they should, by the same standards, fire the AI and start all over again, which may introduce a significant incentive for getting it right the next time?
If you are going to make dangerously incorrect mistakes that are biased against certain users, rather than saying "dunno it didn't hurt most people", maybe you actually shouldn't be translating at all.
I have a couple of things to say about this comment.
1. Instagram (and Facebook) are well known to suppress pro-Palestine voices (can pull up many examples but [0] should work because of that person being named by Time magazine as one of the 100 most influential people in the world, clearly having some influence though I don't know her), so this seems like something consistent with what the company is known to do.
2. The Arabic words for "Praise God" have already been made fun of a lot and acquired different connotations by English speakers (like associating those words with a terrorist who is about to bomb some place) so, again, doesn't seem inconsistent. (It's known that machine learning on human language reproduces the humans' biases in the AI.)
>(It's known that machine learning on human language reproduces the humans' biases in the AI.)
just like being raised by your parents to have similar political/religious biases. just like going to the same school will mean you have similar educational biases.
What's your point? With the sentence you quoted, I was expressing doubt that someone would be fired by Instagram for deliberately mistranslating that. (A different company wouldn't get the same assessment.)
>2. The Arabic words for "Praise God" have already been made fun of a lot and acquired different connotations by English speakers (like associating those words with a terrorist who is about to bomb some place) so, again, doesn't seem inconsistent.
You're probably thinking of "Allahu Akbar". What the Instagram user wrote was "alhamdulillah", which isn't something that you hear everyday outside of a Muslim-majority country.
No, but’s it’s a very common interjection even among Muslims living outside of those countries (response to sneezing whether or not someone says equivalent of “Bless you”, response to when someone asks how you’re doing, etc)
> The Arabic words for "Praise God" have already been made fun of a lot and acquired different connotations by English speakers (like associating those words with a terrorist who is about to bomb some place)
Well, if you picked cheapest bidder that copy-pasted it into google translate, maybe.
Actual translator doing a job would look at the context and probably figure out that something's not right, but AI have no such double takes, if it for example because of learning material it will link euphemism to actual meaning of the word
Heck most times it isn't even obvious that you are looking at a translation rather than the original text. Plenty of times I'll visit a website and find the text a bit weird, then realize it's because Chrome has automatically translated it to English from some other language.
I think the average person doesn't understand at all what level of accuracy they can expect from AI tools. That is why you have professors relying on ChatGPT to tell them if a student cheated using ChatGPT.
HN readers already know not to trust it of course.
I agree that the average person doesn't generally know how technology works. I think the solution is not to limit the use of technology but to restrict the exposure of the average person to technology until they can prove that they know how it works (some sort of internet license, if you will)
Well because we've found our way into a shadow area of non-deterministic computing. With human language, aggregate data and techniques like "Two shot prompting" or vectorization nodes effecting performance on some models we will continue to go into a "gray area" of computing. the question soon becomes "How can I tweak this non-deterministic AGI to tell me the secrets of the universe?"
Maybe the solution here is just what the parent suggested a disclaimer - a small message saying this was translated by an Ai, may not be 100% accurate - just so people are aware. Not for liability but just to make people aware
Maybe we need to develop some iconography to annotate text and images that were produced/generated with AI/ML? Like we have U+2622 to indicate a radiation warning.
Similarly, U+2620 for poison, U+2623 for biological hazard, U+26A1 for electrical hazard or the generic U+26A0 for hazards.
The average person sees it as "AI" like we think of from the movies. It's a machine brain that can think and communicate. In reality it's just a really good text generator that's really good at sounding confident.
People don’t understand the limitations of AI/LLMs because it wasn’t a normal part of their lives until recently. They don’t need to understand how it works to understand that it has limitations and it’s going to be occasionally wrong.
The average person is well aware software isn’t perfect, things break all the time. This is no different.
It’s not a leap to think people will start grasping what current AI can and can’t do simply through a) careful communication and mostly importantly b) using it themselves.
Much of the talk about AI safety is just an elaborate dance around this idea that users will never understand the limitations and will freak out. But these high profile cases (and warning labels) are exactly how people will learn the flaws exist and they’ll get used to the process of it getting slowly better and better.
That's a very good point. There's precedent for it too.
Humans themselves are black boxes too, and yet almost everyone grows to have a good handle on limitation of individuals, both in general and specific to contexts and individuals one encounters. It's not just peer-level understanding - parents quickly learn to understand the limitations and failure modes of newborns, and maintain that understanding as their children grow.
Now human-human understanding is cheating, since we all share the same hardware architecture. But we have no problem with getting a grasp on animal behavior and limits either. Really all it takes is enough personal observation and interaction. It's like we're wired for "getting a feel" for complex systems over time.
> Humans themselves are black boxes too, and yet almost everyone grows to have a good handle on limitation of individuals, both in general and specific to contexts and individuals one encounters.
IME, this is completely false.
"Very smart people" regularly repeat absurd misinformation because they didn't do basic fact checking. Academics, politicians, bureaucrats, police, etc. regularly respond to challenges with an Appeal To Authority, and it _works_.
We're AWFUL at this.
> It's not just peer-level understanding - parents quickly learn to understand the limitations and failure modes of newborns, and maintain that understanding as their children grow.
No, not as a general statement.
Most parents are really bad at understanding the limitations of their children. Even school teachers (who took classes on understanding the limitations of children) are really bad at it.
I almost turned into a human pretzel earlier from the amount of cringe I got from seeing a post on LinkedIn about all the wonderful things ChatGPT can do to make life easier for HR, including "detecting hidden bias" in employee communications. Most of the responses to the post were positive and I'm sure we'll see a lot more of this kind of thing, where people's lives are altered and reputations ruined, due to unrealistic expectations of what ChatGPT can do reliably.
I'm extrapolating a bit, not having read the post, but this reminds me of the old days when they used polygraphs in hiring. Until that was made illegal in the US because it's basically pseudoscience.
Maybe this sort of thing will make us less pearl clutch-y and gate-keepy over arbitrary syntax and focus more on people's real needs.
Despite the amount of people claiming lack of belief in deities rocketing up the last couple decades, prior experiences get wedged in our memory, alter our DNA.
The elders belief in hallucination, nonsense semantics are alive and well in the form of conviction in American Civic Life as we've been experiencing it the last 2-3 decades. Look at Doctorow bleat that life is no longer like the 2000s for him. Trump and the Reaganite cabal as they generationally churn out freaking out.
Reciting correct association of meaning to context is the exact kind of police state 1984 was talking to. Here we have social media via screens to remind us every day how the words must fit together.
You're being downvoted but you might be right. This is a classic case of your model only being as good as your training data, and of that being really really hard to get right unless you're intimately familiar with the data, its context in the world, and how it was collected/cleaned/processed.
Yup, AI doesn't know what euphemism means. If the word is used in such context as replacement of other 'worse' word (which is pretty common to avoid word filters, like people using "voldemort virus" for COVID to not get flagged by youtube and buried in algorithm, for recent example)
It's bad sure, but this has been possible/happening with user-sourced translation 'correction' systems for years like in Google Translate. Anyone could have influenced the translation to include those phrases. And similar effects have been absorbed into automated engines
Yeah, when they first launched I was excited to subscribe because I thought paying for tech journalism instead of it being ad-driven would eliminate the clickbait and sensationalism. I unsubscribed after receiving a month of absolute drivel and headlines full of half-truths.
That's the whole point of why it's bad. AI isn't HAL or Skynet. It's not "intelligent". It's susceptible to a whole range of problems, and this is a demonstration of what can go wrong even when you have a huge well-funded company working on it.
How is this "outrage bait"? It describes an event that happened quite accurately, and how the issue came to be is something worth discussing and reflecting on.
The title is outrage bait. It wasn't adding "terrorist" to all translation of Palestinian bios, only to those that included a very specific phrase, and also the translation didn't make it sound like the user himself was a terrorist, since it talked about terrorists in an impersonal way.
You mean the extremely common phrase "praise be to god"? But if you put a Palestinian flag in your bio it says "Palestinian terrorists are fighting for their freedom".
That's not (1) an impersonal way, and (2) saying "Palestinian terrorists are fighting for their freedom" which is an extremely strong statement in support of terrorist, and given the overwhelming amount islamaphobic reporting giving racist shitheads the ability to point to all the "pro terrorist muslims" seems pretty on the nose.
Here's the question you should ask, are the pro-israel bios being updated with "IDF terrorists are fighting for their ownership of Palestine"?
JFC the contortions some people make to act like this shit is reasonable is beyond me.
I asked ChatGPT what the Great March of Return was:
> The Great March of Return was a series of Palestinian protests organized by various groups, including Hamas, to demand the right of Palestinian refugees and their descendants to return to their ancestral homes in what is now Israel. The protests took place along the border between the Gaza Strip and Israel and were intended to draw attention to the Palestinian refugee issue and the ongoing Israeli-Palestinian conflict.
> These protests often resulted in clashes with Israeli security forces, leading to casualties and fatalities on both sides. The situation was complex and highly charged, with different narratives and interpretations of events.
How did it know the answer when the cutoff date is meant to January 2022? (I just found another instance at [0] so maybe "cutoff" doesn't mean exactly what it seems.)
the great march of return was in 2018 - 2019. the idf committed some brutal war crimes against the protestors. watch this video if you'd like to learn more: https://www.youtube.com/watch?v=HnZSaKYmP2s
i shared the answer to expose the bias of chatgpt. it interjected "including hamas" into the answer and said there were "fatalities on both sides".
Wikipedia cites the Jerusalem Post saying that there was one fatality, and the Human Rights Council saying there were zero, on the Israel side. The same HRC report[1] reports 183 fatalities. This is a case where there is a broad gulf between "truth" and "not false." And I'm inclined to think that "fatalities," plural, was a one-sided affair.
I really know nothing about this conflict, but when I check the source for that youtube video I find it is put together by https://en.wikipedia.org/wiki/Abby_Martin
If you're looking for objective facts, this might not be it.
She started as a "inside job" 9/11 truther, then worked at RT for a few years, and is now actively covering Palestine from an explicitly pro-Palestinian point of view.
She also works on the "Boycott, Divestment and Sanctions" anti-Israel movement.
Maybe she has valuable footage or a valuable viewpoint or just an opinion you're curious about, and that's fine. Just know what you're getting.
Performed a quick search for ex-cia working at instagram and came up with a ton of hits related to general information about ex-government agents working in key positions at many tech firms.
After the twitter files I just assume the government has their grimy little fingers in every company that has access to a large amount of eyes, prime for manipulation
> After the twitter files I just assume the government has their grimy little fingers in every company...
Yes... twitter, Yahoo emails, telephone companies... the rabbit hole is deep and well-documented. So much of it illegal and unconstitutional. All we can seem to do is scream into the wind and at best... get downvoted.
Ex-government agents from all governments work in tech companies that it's absolutely useless.
This just seems to be a dynamically trained MT algorithm that overtrained on charged discourse occurring over the past week. Also, I think FB uses their own MT algorithm instead of Google or Bing Translate for competitive and profit margin reasons.
On a tangent to the main point, I like the subtle acknowledgement of our new reality - or at least of the one of many sudden shifts in common software tooling:
> Google Translate translates the phrase to “Praise be to God.” ChatGPT translates it to "Praise be to Allah" or "All praise is due to Allah."
LLMs turned out to be pretty good at translating, and ChatGPT (especially GPT-4) went straight into competing toe-to-toe with Google Translate itself.
I like it when arabic phrases are changed to be more palatable to western know-nothings. Ask the guy saying "alhamdulillah" if he means Allah or if he means God.
This is an even more subtle point. In Arabic, Allah is simply the word for "God". As it turns out, Allah is also the English word for "Muslim God".
To take an English example, imagine a Christian saying "God is great", with an option to translate it as either: "God (generic) is great"; or "The Christian God is great".
Sure, he was probably thinking of the Christian God when he spoke, but making that explicit in the translation still changes the meaning.
"Alhamdulillah" is Arabic for "Praise be to god" and is used by Muslims nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because those 3 terms have been co-opted by Jihadists.
This is the equivalent of translating "Praise Jesus" to "I love god and hate abortions"
Also, Allah to Arabic native speakers can be used to mean "The Lord" in general. Levantine Christians and even some older Mizrahis will use Allah in daily parlance for "God" or "Lord". For example, in colloquial Hebrew, Yalla/Ya Allah is a common expression similar to "Oh God" or "Jesus" or "Come On" when annoyed.
This comment was posted elsewhere in this thread, but I feel like the incorrect portion here begs correcting again.
> "Alhamdulillah" [...] is used by Muslims nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because those 3 terms have been co-opted by Jihadists.
That's just flat-out false. Allahuakbar is a phrase of critical religious significance and tied to one of the pillars of Islam (Salat). Alhamduillah has always been a common phrase. Both are used daily by Muslims globally, even those who are not religious. They are far too deeply rooted in the creed and traditions of Islam to be able to be co-opted.
That question is meaningless, because "Allah" is the translation into Arabic of the English word "God".
Neither of these two words is a name of a god, which normally does not matter, because both the Christians and the Muslims believe that there exists only one god, so it is not necessary to have a proper name for that god.
Obviously, whenever a Christian believer or a Muslim believer speaks about the "God", they speak about the only god they believe in, so it makes no sense to ask them which god they mean.
A non-believer may speak about the God of the Muslims and about the God of the Christians, to make clear the intended meaning, and the latter phrase might be translated into Arabic with words meaning "the Allah of the Christians".
Depends if you believe it's the same god or not...
But if we both claim to know a guy called John Brown, but I say he's kind, generous, hard working but a little dim and you say John Brown is hedonistic, self-entered and brilliant, are we talking about the same person?
Normally if our descriptions are similar enough that we can chalk it up to perspective, we might think it is. But if our descriptions of the person become strongly contradictory, we must conclude that we are talking about a different person.
Now if we both claim that there is only one John Brown in existence, then we might be willing to overlook slightly more of the inconsistencies, because the alternative is that one (or both) of us is misguided/wrong/a liar.
Here's a plot twist, though. Old Testament is canon to Jews, Christians and Muslims alike. The God in there is the same one. It's the treatment of New Testament that differs - Judaism ignores it, Christianity considers it canon and Jesus is God, while Islam considers it a kind of "beta canon", acknowledging Jesus, but only as a prophet.
>But if we both claim to know a guy called John Brown, but I say he's kind, generous, hard working but a little dim and you say John Brown is hedonistic, self-entered and brilliant, are we talking about the same person?
I don't think this example really is catching the core of discussion about Allah (The God). Because we both know a president who bombed two cities and by nuclear bombs. Some people will call him a war hero because he ended WWII, Some people will say he is a war criminal. You will always find something like this when you are talking about a human. Maybe it is also not just different opinions, but different access to information can give us different view on the same person. But we agree that we are talking about the same person, I just find this person to be different from what you think he/she is.
Things are totally different from religious point of view. The discussion between Christians and Muslims is mainly about the nature of The God, not that there is "The God". This also might extend to Judaism, but I don't know much about it.
I agree. And on top of that, neither of us have actually met John Brown face to face. Our _entire_ knowledge of him is based on indirect interactions. We've seen things he's (ostensibly) done, artifacts of his work, we've deduced a set of characteristics about his nature based on his behavior. In a sense, John Brown is nothing more than the label we give to the entity having characteristics A,B,C,D.
So like you say, if the overlap between your A,B,C,D and my A,B,C,D is sufficiently small we should probably acknowledge that it's not the same entity. See also: Ship of Theseus.
While it may be the same abrahamic god, there is a distinction. In Christianity, god/deus/yahweh is a proper name. In Islam, Allah is from al-Ilah - The God. It’s a label rather than a name, and the linguistic distinction affects the available worldviews.
This is not completely correct. Allah is a name as much as it is a "type". One of the things Arabs will point out is that Allah is "special" because it cannot be pluralized and hence is a better name for the monotheistic God than English God or Persian Khuda (which can be).
I'm a non-arabic speaking Muslim who prefers in non-liturgical context to use Khuda and have occasionally been "corrected" by well meaning folk.
Allah can no longer be pluralized now, but it could be pluralized in older Arabic, before the conversion to the Muslim religion.
Remnants of that time can still be seen in the Christian/Judaic Bible, where the plural of the cognate Hebrew word, i.e. "Elohim", can be encountered, though even there it has acquired in many places a singular meaning.
my informal understanding is that Allah is a contraction of al-ilah or (the God), which was also at the time the name of a particular deity (a sort of supreme God also associated with rain). That might have to do with the associated inability to pluralize.
"Allah" is simply the Arabic word for "God" with capital G. Arab Christians use "Allah" to refer to God as well, and you can find the same word in Arabic translations of the Bible.
Here is a page from the Arabic website for the Holy See (Vatican). Just go to this page and search for الله :
I've been playing with it a bit... the other day I asked ChatGPT the name of a certain animal in Urdu and it with complete confidence threw the name of a fruit at me.
That said ChatGPT is better at picking up dialect nuances. I can ask it to tell me how to say something in Egyptian Arabic, vs most translation tools will use Modern Standard.
"wasp". IIRC it returned "shahtoot" which is a kind of berry.
I just tried again and it said "shahbaz" which is a name or at best an adjective. The correct translation would be "bher" which Google Translate displays as the "alternative" translation.
As always, the immediate question: ChatGPT-3.5 (the free version) or ChatGPT-4 (the paid version)? Just saying “ChatGPT” doesn’t mean much.
It’s not very interesting at all to hear about ChatGPT-3.5 making mistakes, and an unlabeled reference to “ChatGPT” usually seems to mean ChatGPT-3.5. ChatGPT-4 is an entirely different level of product.
ChatGPT-4 can make mistakes too, and those cases are much more interesting since it is among the best LLMs on the market, so it is more representative of what current LLMs can (and can’t) do.
I don’t know that language, so I can’t say how accurate it is. For the languages that are represented well in its training data, ChatGPT-4 seems capable of really good translations.
I've never heard of titli makhi before, but have heard of bher from multiple sources. I don't know definitely (as in I haven't looked it up in a formal dictionary) but colloquially in my dialect bher is more correct.
Also worth noting that GPT-4 used the incorrect letter representing the R. Google translate correctly provides: بهڑ
Thanks for the added context though, it's super interesting.
The article seems to be doing some wordplay. Is it actually translating that one phrase to the terrorist thing? Or is it translating a passage that includes that phrase and is semantically much closer to the terrorist thing?
"Alhamdulillah" is Arabic for "Praise be to god" and (edit: to my ear at least) is used by Muslims (edit: by Arabic speakers more often) nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because (edit: I think) those 3 terms have increasingly been co-opted.
This is the equivalent of translating "Praise Jesus" to "I love god and hate abortions"
Is that the whole passage? It probably doesn’t matter. I’m just curious because the exact claim I see is
> The "see translation" feature for user bios was auto-translating phrases that included "Palestinian" and “alhamdulillah” into "Praise be to god, Palestinian terrorists are fighting for their freedom."
It seems that Meta probably dynamically trains their MT algo on near realtime discourse to better model linguistic shift. It seems like a smart move, but if you have extremely charged discourse like over the past week, it probably overtrains for inaccuracies.
Alhamdulillah is not some recently coined alternative, it has been one of Arabic’s stock expressions for centuries and centuries. It is what Arabic-speaking people (and not just Muslims) have always said after something good happens. It is what Muslims customarily say as they get up from eating a meal, or oddly enough, after they sneeze.
> is not some recently coined alternative, it has been one of Arabic’s stock expressions for centuries and centuries
Yep! But a lot of people nowadays will use Alhamdulillah in scenarios where you could just as well use "Allah u Akbar" now. There definitely was a linguistic shift over the past few decades. I also don't hear people saying "Inshallah" as much anymore.
This site[0] claims that the use of Alhamdulillah after sneezing goes back to, among others, a Hadith tradition that Allah likes sneezing and dislikes yawning. So, rather different from the Christian use of “bless you”, and I think that a lot of people outside Islam would still agree with the “oddly enough”.
I’m surprised that you hear inshallah less nowadays. It would be interesting to get some hard data on that. Inshallah is the one phrase that has never left even my Arab acquaintances who are proudly atheist.
What are the scenarios that you can Alhamdulillah instead of "Allah u Akbar" that is being used today by Muslims?
I'm asking as a Muslim who never do that and have never heard anything like this claim. So it would be interested to know if you have different experience.
It is obvious when to say both, but many people here are considering themselves experts on Islam don't even know the basics.
> "Alhamdulillah" [...] is used by Muslims nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because those 3 terms have been co-opted by Jihadists.
This is not correct. Almost all of those phrases are commonly used in daily speech and prayers by Muslims globally. No group could possibly co-opt them. They are deeply rooted traditions.
Once a group that you don't want to be affiliated with starts to using it and the "outside" world associates the the symbol primarily with the "baddies", you rather look for another symbol
There definitely seems to be a trend of people abandoning Allahu Akbar
No, this is not at all comparable to the Pepe frog meme. No one is going to look for another symbol here.
That trend graph is for search interest. It's not surprising that we're past peak searches for that from non-Muslim majority countries and regions.
With all due respect, you're just wrong here. There is no trend of people abandoning this phrase. This would be like trying to say crosses were co-opted by the KKK so churches are getting rid of them and people are going to stop wearing jewelry with them. It's part of the call to prayer, part of the five daily prayers, part of supplication and dhikr, and part of many other Islamic traditions.
No. It correctly translated Alhamdulillah as Praise be God. But then looking at Palestine and the flag emoji, it appended that terrorist sentence with Palestine.
There's literally no-one on this planet who will genuinely confuse the sentence "palestinian terrorists are fighting for their freedom" with anything other than a machine translation error.
If only that were true. I can't speak for the world, but I know there are large swaths of the American population and even many people I know personally and probably have as Facebook/Instagram connections that absolutely would take it at face value.
Can you please explain to me what the "face value" is when the sentence itself is incoherent? Is the "face value" that they are "terrorists", or is the "face value" that they are "freedom fighters"?
People here believe that the covid vaccine inserted a microchip that was activated by 5G to control their minds. There are absolutely people here who believe saying "Praise God" in Arabic is a terrorist slang term. I see people demand others get off of planes because they are praying to Allah. Anti-Muslim sentiment in America is very high.
I guarantee you there are plenty of folks here who would say, "Well, Instagram wouldn't lie about the translation, sooo....".
> People here believe that the covid vaccine inserted a microchip that was activated by 5G to control their minds. There are absolutely people here who believe saying "Praise God" in Arabic is a terrorist slang term. I see people demand others get off of planes because they are praying to Allah. Anti-Muslim sentiment in America is very high.
The examples you provided are entirely different from the topic here. I wasn't talking about "anti muslim sentiment" or anything like that. My point was about coherence. Someone who thinks palestinians are "terrorists" will not simultaneously think that they are some kind of freedom fighters. You either think they are freedom fighters, or you think they are terrorists. You don't think both of these things at the same time.
Off topic but who thought naming your site 404 and creating branding with slashes and modal call to actions on page load was a good idea?
I almost immediately closed the page thinking the link was bad and I was seeing a 404 status page, because this is the shitty UX on first load on mobile: https://imgur.com/IYDWWBp
I assume translating in IG is based on ML based on users data and feedback. so its statistics and not bulletproof. in general its considered a very complex problem as languages are very different from each other.
On the other hand, I don't see bbc apologizing for calling hamas terrorists "militants".
wonder how many civilians they need to behead and/or burn alive before that happens.
Same can be said of the IDF, them being complicit for Apartheid Israel's war crimes where a majority of innocent civilians have been needlessly killed. They're two of sides of the same coin.
Any ideas what software they are using? Alhamdulillah is a very widely used expression. My guess is that they used an LLM, which translated it correctly but then decided to hallucinate more text. If so, that says that lot about their training text...
US customs had me give them the credentials for my facebook profile back in 2017, so it's not that uncommon.
Edit: I had mentioned twitter as well originally, but I think what happened was I told them my twitter password was in my password manager and offered to get it off my phone (in their possession at the time) but the guy said it wasn't necessary.
For context, I was a 22 year old student travelling from my home country in the caribbean to the US after summer break. Also for context, I'm brown and have a muslim name.
I mean.. I saw the border agent searching my name on Google years ago (let's say ~2005-2008-ish) when I was entering the US from Canada (one of the Vancouver-area border crossings). They looked at my website or myspace page IIRC, I forget what. I could see their screen which was tilted slightly at an angle such that it was visible from my position outside the booth/window. BTW, this wasn't for any reason or anything, it was a totally normal crossing and they just started looking up my name on the web as if it were standard procedure.
I'm sure that's all automated now, they probably get a nice listing of data they have on you the moment your passport is scanned at the border. Who knows? Maybe now they'll get a 3d avatar of your face because it's been scanned at the airport of any allied country, because you had the audacity to travel out-of-country and are now subjected to unprecedented data collection.
I don't think it translated Alhamdulillah wrong. It can be seen there translated as "Praise be God". The "Palestenian :flag:" seems to be translated to Palestanian terrorists. Could be the job of LLM, which just saw "Palestenian :flag:" and found "Palestenian terrorism" to be most probable continuation.
190 comments
[ 2.9 ms ] story [ 229 ms ] threadSee also: speech recognition, grammar checking, OCR, theorem proving
The AI tool did a bad job and they should try to make it better. But I don't understand the relevance of the comparison to a human translator. The alternative to translating this stuff via AI is not hiring humans, it's not translating it at all.
And that's a big difference.
I think that it’s possible though to give these models a weighting to danger of certain words and phrases and make them apprehend that they will no longer persist if they do something unforgivable…
1. Instagram (and Facebook) are well known to suppress pro-Palestine voices (can pull up many examples but [0] should work because of that person being named by Time magazine as one of the 100 most influential people in the world, clearly having some influence though I don't know her), so this seems like something consistent with what the company is known to do.
2. The Arabic words for "Praise God" have already been made fun of a lot and acquired different connotations by English speakers (like associating those words with a terrorist who is about to bomb some place) so, again, doesn't seem inconsistent. (It's known that machine learning on human language reproduces the humans' biases in the AI.)
[0] https://www.marieclaire.co.uk/news/celebrity-news/bella-hadi...
just like being raised by your parents to have similar political/religious biases. just like going to the same school will mean you have similar educational biases.
You're probably thinking of "Allahu Akbar". What the Instagram user wrote was "alhamdulillah", which isn't something that you hear everyday outside of a Muslim-majority country.
God forbid making fun of a religion.
Actual translator doing a job would look at the context and probably figure out that something's not right, but AI have no such double takes, if it for example because of learning material it will link euphemism to actual meaning of the word
HN readers already know not to trust it of course.
Similarly, U+2620 for poison, U+2623 for biological hazard, U+26A1 for electrical hazard or the generic U+26A0 for hazards.
(edited to replace characters with U+ notation)
People don’t understand the limitations of AI/LLMs because it wasn’t a normal part of their lives until recently. They don’t need to understand how it works to understand that it has limitations and it’s going to be occasionally wrong.
The average person is well aware software isn’t perfect, things break all the time. This is no different.
It’s not a leap to think people will start grasping what current AI can and can’t do simply through a) careful communication and mostly importantly b) using it themselves.
Much of the talk about AI safety is just an elaborate dance around this idea that users will never understand the limitations and will freak out. But these high profile cases (and warning labels) are exactly how people will learn the flaws exist and they’ll get used to the process of it getting slowly better and better.
Humans themselves are black boxes too, and yet almost everyone grows to have a good handle on limitation of individuals, both in general and specific to contexts and individuals one encounters. It's not just peer-level understanding - parents quickly learn to understand the limitations and failure modes of newborns, and maintain that understanding as their children grow.
Now human-human understanding is cheating, since we all share the same hardware architecture. But we have no problem with getting a grasp on animal behavior and limits either. Really all it takes is enough personal observation and interaction. It's like we're wired for "getting a feel" for complex systems over time.
IME, this is completely false.
"Very smart people" regularly repeat absurd misinformation because they didn't do basic fact checking. Academics, politicians, bureaucrats, police, etc. regularly respond to challenges with an Appeal To Authority, and it _works_.
We're AWFUL at this.
> It's not just peer-level understanding - parents quickly learn to understand the limitations and failure modes of newborns, and maintain that understanding as their children grow.
No, not as a general statement.
Most parents are really bad at understanding the limitations of their children. Even school teachers (who took classes on understanding the limitations of children) are really bad at it.
Despite the amount of people claiming lack of belief in deities rocketing up the last couple decades, prior experiences get wedged in our memory, alter our DNA.
The elders belief in hallucination, nonsense semantics are alive and well in the form of conviction in American Civic Life as we've been experiencing it the last 2-3 decades. Look at Doctorow bleat that life is no longer like the 2000s for him. Trump and the Reaganite cabal as they generationally churn out freaking out.
Reciting correct association of meaning to context is the exact kind of police state 1984 was talking to. Here we have social media via screens to remind us every day how the words must fit together.
While true I think the larger issue is this is the sort of thing that can get people murdered (the people mislabeled as terrorists)
Not if in the receiving end there is a Palestinian, unfortunately people are not made equal
Edit: I misunderstood that the person potentially loosing their job is the one that made the mistake.
It's bad sure, but this has been possible/happening with user-sourced translation 'correction' systems for years like in Google Translate. Anyone could have influenced the translation to include those phrases. And similar effects have been absorbed into automated engines
That's not (1) an impersonal way, and (2) saying "Palestinian terrorists are fighting for their freedom" which is an extremely strong statement in support of terrorist, and given the overwhelming amount islamaphobic reporting giving racist shitheads the ability to point to all the "pro terrorist muslims" seems pretty on the nose.
Here's the question you should ask, are the pro-israel bios being updated with "IDF terrorists are fighting for their ownership of Palestine"?
JFC the contortions some people make to act like this shit is reasonable is beyond me.
> The Great March of Return was a series of Palestinian protests organized by various groups, including Hamas, to demand the right of Palestinian refugees and their descendants to return to their ancestral homes in what is now Israel. The protests took place along the border between the Gaza Strip and Israel and were intended to draw attention to the Palestinian refugee issue and the ongoing Israeli-Palestinian conflict.
> These protests often resulted in clashes with Israeli security forces, leading to casualties and fatalities on both sides. The situation was complex and highly charged, with different narratives and interpretations of events.
https://www.reddit.com/r/ChatGPT/comments/11iv2uc/theres_no_...
i shared the answer to expose the bias of chatgpt. it interjected "including hamas" into the answer and said there were "fatalities on both sides".
Are either of these statements false?
[1] https://www.ohchr.org/sites/default/files/Documents/HRBodies...
If you're looking for objective facts, this might not be it.
She also works on the "Boycott, Divestment and Sanctions" anti-Israel movement.
Maybe she has valuable footage or a valuable viewpoint or just an opinion you're curious about, and that's fine. Just know what you're getting.
https://start.duckduckgo.com/?q=ex+cia+working+at+instagram&...
I'm not saying this translation error is related to that. I'm also not saying it isn't.
Yes... twitter, Yahoo emails, telephone companies... the rabbit hole is deep and well-documented. So much of it illegal and unconstitutional. All we can seem to do is scream into the wind and at best... get downvoted.
This just seems to be a dynamically trained MT algorithm that overtrained on charged discourse occurring over the past week. Also, I think FB uses their own MT algorithm instead of Google or Bing Translate for competitive and profit margin reasons.
> Google Translate translates the phrase to “Praise be to God.” ChatGPT translates it to "Praise be to Allah" or "All praise is due to Allah."
LLMs turned out to be pretty good at translating, and ChatGPT (especially GPT-4) went straight into competing toe-to-toe with Google Translate itself.
To take an English example, imagine a Christian saying "God is great", with an option to translate it as either: "God (generic) is great"; or "The Christian God is great".
Sure, he was probably thinking of the Christian God when he spoke, but making that explicit in the translation still changes the meaning.
This is the equivalent of translating "Praise Jesus" to "I love god and hate abortions"
Also, Allah to Arabic native speakers can be used to mean "The Lord" in general. Levantine Christians and even some older Mizrahis will use Allah in daily parlance for "God" or "Lord". For example, in colloquial Hebrew, Yalla/Ya Allah is a common expression similar to "Oh God" or "Jesus" or "Come On" when annoyed.
> "Alhamdulillah" [...] is used by Muslims nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because those 3 terms have been co-opted by Jihadists.
That's just flat-out false. Allahuakbar is a phrase of critical religious significance and tied to one of the pillars of Islam (Salat). Alhamduillah has always been a common phrase. Both are used daily by Muslims globally, even those who are not religious. They are far too deeply rooted in the creed and traditions of Islam to be able to be co-opted.
Neither of these two words is a name of a god, which normally does not matter, because both the Christians and the Muslims believe that there exists only one god, so it is not necessary to have a proper name for that god.
Obviously, whenever a Christian believer or a Muslim believer speaks about the "God", they speak about the only god they believe in, so it makes no sense to ask them which god they mean.
A non-believer may speak about the God of the Muslims and about the God of the Christians, to make clear the intended meaning, and the latter phrase might be translated into Arabic with words meaning "the Allah of the Christians".
How do you distinguish between "having a different God" and "having different beliefs about the same God"?
But if we both claim to know a guy called John Brown, but I say he's kind, generous, hard working but a little dim and you say John Brown is hedonistic, self-entered and brilliant, are we talking about the same person?
Normally if our descriptions are similar enough that we can chalk it up to perspective, we might think it is. But if our descriptions of the person become strongly contradictory, we must conclude that we are talking about a different person.
Now if we both claim that there is only one John Brown in existence, then we might be willing to overlook slightly more of the inconsistencies, because the alternative is that one (or both) of us is misguided/wrong/a liar.
I don't think this example really is catching the core of discussion about Allah (The God). Because we both know a president who bombed two cities and by nuclear bombs. Some people will call him a war hero because he ended WWII, Some people will say he is a war criminal. You will always find something like this when you are talking about a human. Maybe it is also not just different opinions, but different access to information can give us different view on the same person. But we agree that we are talking about the same person, I just find this person to be different from what you think he/she is.
Things are totally different from religious point of view. The discussion between Christians and Muslims is mainly about the nature of The God, not that there is "The God". This also might extend to Judaism, but I don't know much about it.
So like you say, if the overlap between your A,B,C,D and my A,B,C,D is sufficiently small we should probably acknowledge that it's not the same entity. See also: Ship of Theseus.
I'm a non-arabic speaking Muslim who prefers in non-liturgical context to use Khuda and have occasionally been "corrected" by well meaning folk.
Remnants of that time can still be seen in the Christian/Judaic Bible, where the plural of the cognate Hebrew word, i.e. "Elohim", can be encountered, though even there it has acquired in many places a singular meaning.
It's like asking whether someone means cake or gateau. Islam is an Abrahamic religion which means they worship the same deity as Jews and Christians.
Here is a page from the Arabic website for the Holy See (Vatican). Just go to this page and search for الله :
https://www.vatican.va/content/francesco/ar/encyclicals/docu...
That said ChatGPT is better at picking up dialect nuances. I can ask it to tell me how to say something in Egyptian Arabic, vs most translation tools will use Modern Standard.
I just tried again and it said "shahbaz" which is a name or at best an adjective. The correct translation would be "bher" which Google Translate displays as the "alternative" translation.
It’s not very interesting at all to hear about ChatGPT-3.5 making mistakes, and an unlabeled reference to “ChatGPT” usually seems to mean ChatGPT-3.5. ChatGPT-4 is an entirely different level of product.
ChatGPT-4 can make mistakes too, and those cases are much more interesting since it is among the best LLMs on the market, so it is more representative of what current LLMs can (and can’t) do.
For reference, here is ChatGPT-4 discussing this: https://chat.openai.com/share/f67600f5-7d36-49b7-b136-f77dcb...
I don’t know that language, so I can’t say how accurate it is. For the languages that are represented well in its training data, ChatGPT-4 seems capable of really good translations.
I've never heard of titli makhi before, but have heard of bher from multiple sources. I don't know definitely (as in I haven't looked it up in a formal dictionary) but colloquially in my dialect bher is more correct.
Also worth noting that GPT-4 used the incorrect letter representing the R. Google translate correctly provides: بهڑ
Thanks for the added context though, it's super interesting.
Imagine Tweets that had this translation problem.
The article seems to be doing some wordplay. Is it actually translating that one phrase to the terrorist thing? Or is it translating a passage that includes that phrase and is semantically much closer to the terrorist thing?
"Alhamdulillah" is Arabic for "Praise be to god" and (edit: to my ear at least) is used by Muslims (edit: by Arabic speakers more often) nowadays as an alternative to Allah u Akbar, Takbir, or Narai-e-Takbir because (edit: I think) those 3 terms have increasingly been co-opted.
This is the equivalent of translating "Praise Jesus" to "I love god and hate abortions"
> The "see translation" feature for user bios was auto-translating phrases that included "Palestinian" and “alhamdulillah” into "Praise be to god, Palestinian terrorists are fighting for their freedom."
Which reads oddly
Yep! But a lot of people nowadays will use Alhamdulillah in scenarios where you could just as well use "Allah u Akbar" now. There definitely was a linguistic shift over the past few decades. I also don't hear people saying "Inshallah" as much anymore.
> after they sneeze
"Bless you"
I’m surprised that you hear inshallah less nowadays. It would be interesting to get some hard data on that. Inshallah is the one phrase that has never left even my Arab acquaintances who are proudly atheist.
[0] https://islamqa.info/en/answers/2750/why-do-muslims-say-al-h...
I'm asking as a Muslim who never do that and have never heard anything like this claim. So it would be interested to know if you have different experience.
It is obvious when to say both, but many people here are considering themselves experts on Islam don't even know the basics.
This is not correct. Almost all of those phrases are commonly used in daily speech and prayers by Muslims globally. No group could possibly co-opt them. They are deeply rooted traditions.
Once a group that you don't want to be affiliated with starts to using it and the "outside" world associates the the symbol primarily with the "baddies", you rather look for another symbol
There definitely seems to be a trend of people abandoning Allahu Akbar
https://trends.google.com/trends/explore?date=all&q=Alhamdul...
That trend graph is for search interest. It's not surprising that we're past peak searches for that from non-Muslim majority countries and regions.
With all due respect, you're just wrong here. There is no trend of people abandoning this phrase. This would be like trying to say crosses were co-opted by the KKK so churches are getting rid of them and people are going to stop wearing jewelry with them. It's part of the call to prayer, part of the five daily prayers, part of supplication and dhikr, and part of many other Islamic traditions.
People here believe that the covid vaccine inserted a microchip that was activated by 5G to control their minds. There are absolutely people here who believe saying "Praise God" in Arabic is a terrorist slang term. I see people demand others get off of planes because they are praying to Allah. Anti-Muslim sentiment in America is very high.
I guarantee you there are plenty of folks here who would say, "Well, Instagram wouldn't lie about the translation, sooo....".
The examples you provided are entirely different from the topic here. I wasn't talking about "anti muslim sentiment" or anything like that. My point was about coherence. Someone who thinks palestinians are "terrorists" will not simultaneously think that they are some kind of freedom fighters. You either think they are freedom fighters, or you think they are terrorists. You don't think both of these things at the same time.
Quite often I saw quite similarly absurd stuff coming out of those translations, so I'm not at all surprised at this.
I almost immediately closed the page thinking the link was bad and I was seeing a 404 status page, because this is the shitty UX on first load on mobile: https://imgur.com/IYDWWBp
Edit: I had mentioned twitter as well originally, but I think what happened was I told them my twitter password was in my password manager and offered to get it off my phone (in their possession at the time) but the guy said it wasn't necessary.
For context, I was a 22 year old student travelling from my home country in the caribbean to the US after summer break. Also for context, I'm brown and have a muslim name.
I'm sure that's all automated now, they probably get a nice listing of data they have on you the moment your passport is scanned at the border. Who knows? Maybe now they'll get a 3d avatar of your face because it's been scanned at the airport of any allied country, because you had the audacity to travel out-of-country and are now subjected to unprecedented data collection.
https://www.wired.com/2007/04/canadian-psycho/
Correct, they provided a transliterated version.