{"video_id":"cQUlbFmjDcM","title":"Google Translate Is Actually Terrible","channel":"Techquickie","show":"Techquickie","published_at":"2025-05-04T14:58:16Z","duration_s":465,"segments":[{"start_s":0.08,"end_s":4.96,"text":"Google Translate is an amazing tool but","speaker":null,"is_sponsor":0},{"start_s":3.12,"end_s":9.719,"text":"it also kind of sucks it's a service that's come a long way over the last 18","speaker":null,"is_sponsor":0},{"start_s":7.759,"end_s":13.24,"text":"years but it's still nowhere near the perfect Universal translator we're used","speaker":null,"is_sponsor":0},{"start_s":11.719,"end_s":18.0,"text":"to seeing in science fiction like Star Trek and hitchhiker Guide to the Galaxy","speaker":null,"is_sponsor":0},{"start_s":16.0,"end_s":23.08,"text":"but how far away from that kind of seamless translation technology are we","speaker":null,"is_sponsor":0},{"start_s":20.359,"end_s":27.4,"text":"really Google Translate is an example of machine translation a subfield of","speaker":null,"is_sponsor":0},{"start_s":25.519,"end_s":32.48,"text":"computational linguistics that has been under development since the 1950s early","speaker":null,"is_sponsor":0},{"start_s":30.199,"end_s":36.48,"text":"machine translators used a rule-based approach which required programmers to","speaker":null,"is_sponsor":0},{"start_s":34.28,"end_s":41.44,"text":"explicitly describe every rule governing a language which was both labor","speaker":null,"is_sponsor":0},{"start_s":38.36,"end_s":43.2,"text":"intensive and flawed in the 1980s","speaker":null,"is_sponsor":0},{"start_s":41.44,"end_s":47.48,"text":"statistical translation models became popular these used probability and","speaker":null,"is_sponsor":0},{"start_s":45.48,"end_s":51.719,"text":"heuristics to determine the best translation for a given phrase which was","speaker":null,"is_sponsor":0},{"start_s":49.84,"end_s":55.559,"text":"more efficient but they struggled to translate between languages with","speaker":null,"is_sponsor":0},{"start_s":53.399,"end_s":60.519,"text":"significant differences in terms of grammar and word order JM is a pretty","speaker":null,"is_sponsor":0},{"start_s":58.8,"end_s":66.72,"text":"normal French phrase but its word order isn't I love you it's","speaker":null,"is_sponsor":0},{"start_s":63.48,"end_s":69.119,"text":"I you love this is one of the reasons","speaker":null,"is_sponsor":0},{"start_s":66.72,"end_s":74.08,"text":"why even relatively fast real-time interpretation programs always come with","speaker":null,"is_sponsor":0},{"start_s":71.36,"end_s":77.96,"text":"a significant delay the software needs to hear the full sentence before it","speaker":null,"is_sponsor":0},{"start_s":75.72,"end_s":82.28,"text":"knows how it's supposed to order the words in its translation when it","speaker":null,"is_sponsor":0},{"start_s":80.0,"end_s":85.88,"text":"launched in 2006 Google translate originally used a statistical","speaker":null,"is_sponsor":0},{"start_s":84.0,"end_s":89.92,"text":"translation model and typically translated Languages by converting them","speaker":null,"is_sponsor":0},{"start_s":87.72,"end_s":94.68,"text":"to English then converting them into into the desired target language in late","speaker":null,"is_sponsor":0},{"start_s":92.2,"end_s":99.479,"text":"2016 Google transitioned its translation service to neural machine translation","speaker":null,"is_sponsor":0},{"start_s":96.92,"end_s":104.399,"text":"which uses a neural network to translate phrases and passages holistically","speaker":null,"is_sponsor":0},{"start_s":101.92,"end_s":109.119,"text":"meaning that it can use the larger text as context when translating specific","speaker":null,"is_sponsor":0},{"start_s":106.88,"end_s":113.96,"text":"phrases Google Translate in particular excels because it's so widely used and","speaker":null,"is_sponsor":0},{"start_s":111.719,"end_s":120.119,"text":"has a massive flow of user feedback to help refine its output that being said","speaker":null,"is_sponsor":0},{"start_s":117.479,"end_s":125.039,"text":"experts do not consider any translation programs today to be capable of fully","speaker":null,"is_sponsor":0},{"start_s":122.28,"end_s":129.879,"text":"automated high quality translation as the technology is still too unreliable","speaker":null,"is_sponsor":0},{"start_s":127.56,"end_s":134.92,"text":"for formal purposes like translating works for publication or interpreting a","speaker":null,"is_sponsor":0},{"start_s":132.36,"end_s":139.48,"text":"witness statement during a court hearing however professional translators do use","speaker":null,"is_sponsor":0},{"start_s":137.28,"end_s":144.68,"text":"a fair number of automated Tools in what is called computer assisted translation","speaker":null,"is_sponsor":0},{"start_s":142.08,"end_s":149.959,"text":"their cyborgs to be fair accurate enough for a court hearing is a pretty high bar","speaker":null,"is_sponsor":0},{"start_s":147.319,"end_s":154.599,"text":"so how accurate is Google Translate really well it can vary a lot depending","speaker":null,"is_sponsor":0},{"start_s":152.28,"end_s":158.8,"text":"on the language pairing if you're a unilingual English speaker Google","speaker":null,"is_sponsor":0},{"start_s":156.879,"end_s":163.159,"text":"translate might actually seem to work really really well that's because","speaker":null,"is_sponsor":0},{"start_s":160.599,"end_s":167.879,"text":"English has the most speakers out of any human language with almost 1.5 billion","speaker":null,"is_sponsor":0},{"start_s":166.0,"end_s":171.48,"text":"and half of websites are written in English Google translate has tons of","speaker":null,"is_sponsor":0},{"start_s":169.84,"end_s":176.36,"text":"English language reference material to draw from but the majority of the other","speaker":null,"is_sponsor":0},{"start_s":174.04,"end_s":181.36,"text":"132 languages supported by Google translate are only represented by a","speaker":null,"is_sponsor":0},{"start_s":178.64,"end_s":187.44,"text":"small fraction of users and websites according to a 2019 UCLA study English","speaker":null,"is_sponsor":0},{"start_s":184.519,"end_s":192.0,"text":"to Spanish was around 94% accurate but English to Armenian was only 55%","speaker":null,"is_sponsor":0},{"start_s":192.519,"end_s":199.879,"text":"accurate but isn't this accuracy Gap just going to go away with","speaker":null,"is_sponsor":0},{"start_s":197.0,"end_s":203.92,"text":"time well yes and no this technology will inevitably become more accurate as","speaker":null,"is_sponsor":0},{"start_s":202.2,"end_s":208.319,"text":"time goes on but there are several deeper problems with machine translation","speaker":null,"is_sponsor":0},{"start_s":205.92,"end_s":212.48,"text":"that we have yet to really solve we'll tell you what they are after this","speaker":null,"is_sponsor":0},{"start_s":209.799,"end_s":216.2,"text":"message from our sponsor iix it if your laptop is broken or acting funky forget","speaker":null,"is_sponsor":1},{"start_s":214.72,"end_s":221.36,"text":"going out and buying a whole new one repair it with the help of iix it their","speaker":null,"is_sponsor":1},{"start_s":218.84,"end_s":224.439,"text":"exhaustive selection of Parts like ssds and batteries along with their","speaker":null,"is_sponsor":1},{"start_s":222.76,"end_s":228.519,"text":"comprehensive repair guides means opening up your laptop and fixing it is","speaker":null,"is_sponsor":1},{"start_s":226.36,"end_s":232.92,"text":"easier than ever check out iFix it using the link in the description and give","speaker":null,"is_sponsor":1},{"start_s":230.439,"end_s":237.56,"text":"your busted laptop a new lease on life it's tempting to think that the route to","speaker":null,"is_sponsor":1},{"start_s":234.439,"end_s":239.439,"text":"a universal translator is just more data","speaker":null,"is_sponsor":0},{"start_s":237.56,"end_s":243.68,"text":"and that would definitely help in arm media case but even with our extremely","speaker":null,"is_sponsor":0},{"start_s":242.0,"end_s":248.239,"text":"sophisticated neural network and plenty of training data machine translators","speaker":null,"is_sponsor":0},{"start_s":246.0,"end_s":253.879,"text":"tend to struggle in a few key areas of human communication most notably slang","speaker":null,"is_sponsor":0},{"start_s":251.319,"end_s":257.359,"text":"jokes and figurative language a lot of human communication relies on","speaker":null,"is_sponsor":0},{"start_s":255.439,"end_s":262.919,"text":"abstraction and double meanings especially metaphors and idioms the","speaker":null,"is_sponsor":0},{"start_s":259.44,"end_s":264.72,"text":"French phrase okam twam I'm sorry is a","speaker":null,"is_sponsor":0},{"start_s":262.919,"end_s":269.24,"text":"Whimsical idiom typically used to describe the height of young children","speaker":null,"is_sponsor":0},{"start_s":267.04,"end_s":273.96,"text":"the rough English equivalent is knee high to a grasshopper but if you","speaker":null,"is_sponsor":0},{"start_s":271.28,"end_s":279.479,"text":"translate it literally it means three apples High Google translate interprets","speaker":null,"is_sponsor":0},{"start_s":276.96,"end_s":285.12,"text":"this French phrase as little person which is sort of right but also very","speaker":null,"is_sponsor":0},{"start_s":281.84,"end_s":287.8,"text":"very wrong similarly the humor of it","speaker":null,"is_sponsor":0},{"start_s":285.12,"end_s":291.68,"text":"ain't rocket surgery just isn't going to translate well without a human being","speaker":null,"is_sponsor":0},{"start_s":289.88,"end_s":296.039,"text":"willing to put in the groundwork to find a suitable cultural equivalent because","speaker":null,"is_sponsor":0},{"start_s":293.919,"end_s":300.4,"text":"French people don't use rocket science or brain surgery as benchmarks for","speaker":null,"is_sponsor":0},{"start_s":298.12,"end_s":304.0,"text":"difficulty they use sourc now this isn't an impossible hurdle for","speaker":null,"is_sponsor":0},{"start_s":302.12,"end_s":308.0,"text":"a machine translator to handle because idioms tend to be said the exact same","speaker":null,"is_sponsor":0},{"start_s":305.88,"end_s":312.24,"text":"way every time you could teach the software how to translate a long list of","speaker":null,"is_sponsor":0},{"start_s":310.0,"end_s":316.72,"text":"specific idioms so long as you are willing to put enough resources into it","speaker":null,"is_sponsor":0},{"start_s":314.68,"end_s":321.919,"text":"teaching it to consistently translate word play and jokes however might very","speaker":null,"is_sponsor":0},{"start_s":319.759,"end_s":325.319,"text":"well be impossible sing is always a real headache for translators both machine","speaker":null,"is_sponsor":0},{"start_s":323.44,"end_s":329.919,"text":"and human because it's typically used by relatively small Niche subcultures and","speaker":null,"is_sponsor":0},{"start_s":327.96,"end_s":335.479,"text":"it's unlikely to wind up in Main language repositories like dictionaries","speaker":null,"is_sponsor":0},{"start_s":332.199,"end_s":337.759,"text":"and thorth slang also tends to change","speaker":null,"is_sponsor":0},{"start_s":335.479,"end_s":341.8,"text":"quickly and rely on community specific cultural references if you were to try","speaker":null,"is_sponsor":0},{"start_s":339.919,"end_s":345.919,"text":"and translate a slang heavy Kendrick Lamar song by pushing it through a","speaker":null,"is_sponsor":0},{"start_s":343.56,"end_s":351.0,"text":"machine translator the result would not be just aesthetically questionable but","speaker":null,"is_sponsor":0},{"start_s":348.6,"end_s":355.12,"text":"also basically incoherent to a person unfamiliar with American rap culture","speaker":null,"is_sponsor":0},{"start_s":353.44,"end_s":358.88,"text":"helping a machine translator to understand slang is again possible it","speaker":null,"is_sponsor":0},{"start_s":357.4,"end_s":363.28,"text":"would just be a bit expensive and require near constant updating where","speaker":null,"is_sponsor":0},{"start_s":361.52,"end_s":367.8,"text":"things get tricky is that there's a slight difference between a translation","speaker":null,"is_sponsor":0},{"start_s":365.039,"end_s":372.72,"text":"being accurate and a translation being good especially when it comes to Art","speaker":null,"is_sponsor":0},{"start_s":370.599,"end_s":376.96,"text":"machine translation can often be technically correct but still fail to","speaker":null,"is_sponsor":0},{"start_s":374.68,"end_s":382.56,"text":"communicate the tone and cultural connotations of the original on a purely","speaker":null,"is_sponsor":0},{"start_s":379.96,"end_s":387.319,"text":"technical level how are you what's up and how's it hanging all mean basically","speaker":null,"is_sponsor":0},{"start_s":384.919,"end_s":391.0,"text":"the same thing but one is a double on Tandra and a weird thing to say to your","speaker":null,"is_sponsor":0},{"start_s":388.88,"end_s":394.919,"text":"grandma we might be happy enough with a machine translation of a menu if it just","speaker":null,"is_sponsor":0},{"start_s":393.28,"end_s":399.68,"text":"accurately communicates what food is available but a translated novel or poem","speaker":null,"is_sponsor":0},{"start_s":398.0,"end_s":403.88,"text":"needs to find a good balance between accuracy and Aesthetics in order to","speaker":null,"is_sponsor":0},{"start_s":401.84,"end_s":408.479,"text":"create a similar experience as the original this means that a literary","speaker":null,"is_sponsor":0},{"start_s":405.8,"end_s":412.919,"text":"translator needs to be creative and make astute artistic judgments in the same","speaker":null,"is_sponsor":0},{"start_s":410.919,"end_s":416.84,"text":"way that an author does the basic problem with translation software is","speaker":null,"is_sponsor":0},{"start_s":414.56,"end_s":421.12,"text":"that there's no mind behind it that truly understands the purpose or intent","speaker":null,"is_sponsor":0},{"start_s":418.919,"end_s":424.72,"text":"of the words its processing an optimistic goal of machine translation","speaker":null,"is_sponsor":0},{"start_s":422.879,"end_s":428.16,"text":"as a discipline is to one day have translation software that is so","speaker":null,"is_sponsor":0},{"start_s":426.08,"end_s":432.8,"text":"sophisticated and nuanced that it only needs a human Editor to check its work","speaker":null,"is_sponsor":0},{"start_s":430.0,"end_s":437.16,"text":"and fix whatever errors they find more pessimistically highquality fully","speaker":null,"is_sponsor":0},{"start_s":434.8,"end_s":442.0,"text":"Automated machine translation might require the development of something","speaker":null,"is_sponsor":0},{"start_s":438.879,"end_s":443.56,"text":"like AGI artificial general intelligence","speaker":null,"is_sponsor":0},{"start_s":442.0,"end_s":448.8,"text":"a machine capable of human-like intelligence and making Nuance judgment","speaker":null,"is_sponsor":0},{"start_s":445.479,"end_s":451.8,"text":"calls so Google translate might need to","speaker":null,"is_sponsor":0},{"start_s":448.8,"end_s":453.96,"text":"be capable of forming opinions before","speaker":null,"is_sponsor":0},{"start_s":451.8,"end_s":457.96,"text":"it's ready to translate poetry thanks for watching guys if you like this video","speaker":null,"is_sponsor":0},{"start_s":455.68,"end_s":463.24,"text":"why don't you check out our one on X's or Twitter's Community notes program","speaker":null,"is_sponsor":0},{"start_s":460.52,"end_s":466.36,"text":"it's actually pretty cool and maybe the future of the internet","speaker":null,"is_sponsor":0}],"full_text":"Google Translate is an amazing tool but it also kind of sucks it's a service that's come a long way over the last 18 years but it's still nowhere near the perfect Universal translator we're used to seeing in science fiction like Star Trek and hitchhiker Guide to the Galaxy but how far away from that kind of seamless translation technology are we really Google Translate is an example of machine translation a subfield of computational linguistics that has been under development since the 1950s early machine translators used a rule-based approach which required programmers to explicitly describe every rule governing a language which was both labor intensive and flawed in the 1980s statistical translation models became popular these used probability and heuristics to determine the best translation for a given phrase which was more efficient but they struggled to translate between languages with significant differences in terms of grammar and word order JM is a pretty normal French phrase but its word order isn't I love you it's I you love this is one of the reasons why even relatively fast real-time interpretation programs always come with a significant delay the software needs to hear the full sentence before it knows how it's supposed to order the words in its translation when it launched in 2006 Google translate originally used a statistical translation model and typically translated Languages by converting them to English then converting them into into the desired target language in late 2016 Google transitioned its translation service to neural machine translation which uses a neural network to translate phrases and passages holistically meaning that it can use the larger text as context when translating specific phrases Google Translate in particular excels because it's so widely used and has a massive flow of user feedback to help refine its output that being said experts do not consider any translation programs today to be capable of fully automated high quality translation as the technology is still too unreliable for formal purposes like translating works for publication or interpreting a witness statement during a court hearing however professional translators do use a fair number of automated Tools in what is called computer assisted translation their cyborgs to be fair accurate enough for a court hearing is a pretty high bar so how accurate is Google Translate really well it can vary a lot depending on the language pairing if you're a unilingual English speaker Google translate might actually seem to work really really well that's because English has the most speakers out of any human language with almost 1.5 billion and half of websites are written in English Google translate has tons of English language reference material to draw from but the majority of the other 132 languages supported by Google translate are only represented by a small fraction of users and websites according to a 2019 UCLA study English to Spanish was around 94% accurate but English to Armenian was only 55% accurate but isn't this accuracy Gap just going to go away with time well yes and no this technology will inevitably become more accurate as time goes on but there are several deeper problems with machine translation that we have yet to really solve we'll tell you what they are after this message from our sponsor iix it if your laptop is broken or acting funky forget going out and buying a whole new one repair it with the help of iix it their exhaustive selection of Parts like ssds and batteries along with their comprehensive repair guides means opening up your laptop and fixing it is easier than ever check out iFix it using the link in the description and give your busted laptop a new lease on life it's tempting to think that the route to a universal translator is just more data and that would definitely help in arm media case but even with our extremely sophisticated neural network and plenty of training data machine translators tend to struggle in a few key areas of human communication most notably slang jokes and figurative language a lot of human communication relies on abstraction and double meanings especially metaphors and idioms the French phrase okam twam I'm sorry is a Whimsical idiom typically used to describe the height of young children the rough English equivalent is knee high to a grasshopper but if you translate it literally it means three apples High Google translate interprets this French phrase as little person which is sort of right but also very very wrong similarly the humor of it ain't rocket surgery just isn't going to translate well without a human being willing to put in the groundwork to find a suitable cultural equivalent because French people don't use rocket science or brain surgery as benchmarks for difficulty they use sourc now this isn't an impossible hurdle for a machine translator to handle because idioms tend to be said the exact same way every time you could teach the software how to translate a long list of specific idioms so long as you are willing to put enough resources into it teaching it to consistently translate word play and jokes however might very well be impossible sing is always a real headache for translators both machine and human because it's typically used by relatively small Niche subcultures and it's unlikely to wind up in Main language repositories like dictionaries and thorth slang also tends to change quickly and rely on community specific cultural references if you were to try and translate a slang heavy Kendrick Lamar song by pushing it through a machine translator the result would not be just aesthetically questionable but also basically incoherent to a person unfamiliar with American rap culture helping a machine translator to understand slang is again possible it would just be a bit expensive and require near constant updating where things get tricky is that there's a slight difference between a translation being accurate and a translation being good especially when it comes to Art machine translation can often be technically correct but still fail to communicate the tone and cultural connotations of the original on a purely technical level how are you what's up and how's it hanging all mean basically the same thing but one is a double on Tandra and a weird thing to say to your grandma we might be happy enough with a machine translation of a menu if it just accurately communicates what food is available but a translated novel or poem needs to find a good balance between accuracy and Aesthetics in order to create a similar experience as the original this means that a literary translator needs to be creative and make astute artistic judgments in the same way that an author does the basic problem with translation software is that there's no mind behind it that truly understands the purpose or intent of the words its processing an optimistic goal of machine translation as a discipline is to one day have translation software that is so sophisticated and nuanced that it only needs a human Editor to check its work and fix whatever errors they find more pessimistically highquality fully Automated machine translation might require the development of something like AGI artificial general intelligence a machine capable of human-like intelligence and making Nuance judgment calls so Google translate might need to be capable of forming opinions before it's ready to translate poetry thanks for watching guys if you like this video why don't you check out our one on X's or Twitter's Community notes program it's actually pretty cool and maybe the future of the internet"}