Here’s a fun fact I learned from reading a paper trying to develop tests for AI ethics: since GPT learns language from reading a big training set of data, if you ask it who owns the Kuril Islands -in Japanese-, it says Japan does. If you ask it in Russian, it says Russia does. If you ask it in English, it says there’s a controversy
Yeah, I noticed that the amount of details given in ChatGPT's answers varies between languages - kinda like Wikipedia articles vary in length and details depending on the given language.
As an Argentinian: the "Spanish truth" is that "Islas Malvinas" are Argentinian, but the "English truth" is that the "Falkland Islands" are British, regardless of being the same landmass, I haven't tried GPT but I'll do ask it about the matter when I do, it'll probably depend on which name you use it.
@@くぽにゃん It's definitely scripted. While some this seems like legit ChatGPT answers, it would never answer like that. It would just take that comment as face value and express gratitude.
I’m really glad I found Dogen’s channel, was starting to lose interest in learning Japanese, but his videos are so creative, they have both education of the language and culture and comedy. Thank you sir
And he actually speaks the language in his videos, unlike many polyglots who claim they speak the language and rarely do the videos in them. Kudos to him. :)
"Japanese cuisine is known for using fresh ingredients and being nutrionally balanced" Me, after having 5 piece of FamiChiki and miso with a higher salt content than the Dead Sea: *Ahah, yeah...*
In my 3 months of living in Japan I lost 10 kilograms. This was despite pigging out on fried chicken foods and other snacks from konbini day and night. I ate more while I was in Japan than ever before in my life. I call cap on Japanese junk food being bad for you.
funny vid! 😄 I'm in Tokyo learning Japanese at a language school right now. We study between N3/N2 in class. All the N2 books are pure Japanese. So it's hard to grasp new grammar. But since ChatGPT is born, we are all sitting in class with a connected Telegram ChatGPT bot and if you don't know stuff you can just ask the bot in your native language. This is just crazy and helps so much while the teacher is explaining stuff in their Japanese way of doing. Also it gives you a ton of sample sentences you can discuss with your teacher. I never wanna miss that thing in my life. So fkn 便利です！
Yes, I feel it. I learn Japanese at home atm, with Anki vocabulary, grammar tests, listening and reading comprehension etc. Also watching the Japanese version of "Gilmore Girls" on with subtitles even if I don't understand much xD. ChatGPT often needs to tell me if my sentence structure was correct or if I used the correct particles etc.
"Ok you are wrong." "Ok I believe you, how am I wrong?" "You just have a lot to learn." "Ok, what is it that I need to learn?" "Umm... never mind, you are very good at Japanese." "M8 r u takin the piss right now?" Stunningly human, I can't imagine how thick this computer's skin is. Also Dogen don't bully AI-chan. Not only does it not care, you are teaching it how to troll you back.
@@lamented-musings8932 haha no I didn't understand that as the joke, and for the given quote there isn't really a joke being made. Why do you think the joke is not using gpt4, whatever that is? A chat program?
@@misteral9045 chat gpt is an advanced ai chatbot. It learns from humans and other sources and keeps what it learns in its databank/memory. It is a useful tool for just about anything but as with any information tool it shouldn’t be your only source of information. I used it to help rewrite my resume to sound more professional for example.
Great video, especially good because your speech is nice and slow while typing, which helps us beginners with understanding what's being said! 👍 I'll probably watch this a few times for that very reason.
Yesterday I was curious as to whether or not Chat-GPT could generate a synthesis of existing information on a subject. I asked it to write a synthesis of the known detections of the little fire ant, Wasmannia auropunctata, throughout Asia, and to provide its sources of information. I soon learned this was beyond the language model’s capabilities. Chat-GPT falsely reported that the little fire ant has been detected in Japan as of 2015 in Yamagata prefecture, going on to say it was found in Okinawa, Aichi, Hyogo, and Kyoto prefectures, even providing an in-text citation (Tsuji et al., 2020). In its citations list, the AI then wrote: Tsuji, K., Yoshimura, M., Nakano, M.A., and Konishi, M. (2020). Distribution and spread of the little fire ant, Wasmannia auropunctata (Roger) (Hymenoptera: Formicidae), in Japan. Applied Entomology and Zoology, 55, 509-516. As far as I can tell, this source *does not exist.* Later in a separate I asked for Japanese journals specifically looking for detections of Wasmannia auropunctata, and then to provide those sources in Japanese rather than translated. One of these sources was as follows: Sugiura, S., Tanaka, H.O., Itioka, T., Yamauchi, K., Tatsuta, H., and Nishida, T. (2019). 小アリモドキの愛知県内での侵入 [Recent invasion of the little fire ant, Wasmannia auropunctata (Roger) (Hymenoptera: Formicidae), in Aichi Prefecture, central Japan.] Japanese Journal of Systematic Entomology, 25(2), 199-202. Once again, *this journal does not appear to exist.* I am disturbed to say that Chat-GPT is incredibly skilled at bluffing its way into providing information authoritatively, even when that info is completely fabricated. I later corrected the AI, and it apologized for its error. Always try to correct the language model.
This is what I've been trying to tell everyone about. So many news outlets and influencers treat ChatGPT as if it were a new type of knowledge repository, _a la_ Wikipedia or Google. But even though sources you could find on Google were prone to errors or bias, and Wikipedia could still get things wrong, you could still find authoritative information in both cases. Chat-GPT _has_ no information. Chat-GPT is not an information repository. It is a language model. Its task is to reproduce the patterns of humans speech. Nothing more, nothing less. To this end, it was fed with an enormous dataset in which it analyses, not what is true, but what _sounds_ true. If it can get things right in many cases, it's simply a beneficial side effect - but it's not the purpose. And when it gets things wrong, it is not failing - because it doesn't _aim_ to be right. Trying to use Chat-GPT to extract truth is like trying to find information about physics in a science-fiction novel. It's simply _not the point_ - and people are being misled about this.
It polls through some obscure sources in niche topics, which is odd because I have no idea how some of the sources are accessed as datasets given that print editions don't seem to be digitized. The source for my questions exists, but there's no way to know if it's just, as you say, bluffing that the information can be found within the source unless you have the hard copy of the source.. It's a good tool but using it in niche topics is really scary because it will look so convincing and its mistakes are really subtle.
I asked ChatGPT about the Japanese constitution. I got a very convincing answer, but fact checked it with a government website and found out that ChatGPT had given me a completely wrong answer. When I pointed out the error, ChatGPT quickly corrected. But if I just took the first answer, I would have looked really stupid. That's the danger of using ChatGPT for research for TH-cam videos.
@@weeklyfascination ChatGPT needs much more data organization and arrangement in order to “understand” more and make things up less. A lot of the problem is that chatGPT is unaware how close things are in relation in terms of pronouncement. It does not know how to gauge and balance what is more or what is less. This is the gap between appearance and reality.
@@tusla6889at GPT doesn't know much of anything. It's a language model made from word relationships in a selected language . It doesn't process any info or to even reconstruct it, it simply mashes words in an accepted pattern. Glorified text prediction autocorrect has basically
Playing with ChatGPT, it is interesting. One thing it's really good at is explaining why certain words or phrases don't translate well in other translation programs. However - it also just makes odd errors or invents Kanji readings with no apparent basis. It does this infrequently enough to potentially catch a user off guard.
I've been using chatgpt to figure out which reading if a kanji is more commonly used. It works really nicely because usually I have to ask a native Japanese friend to explain which reading is more commonly used and in what context are the other readings used. I also use online web ai plugins like perplexing in summarizing articles that are written in Japanese
Loved it! One of the first things I asked chatgpt when trying it out was about some nuances in Japanese I did not know, and it was actually fairly helpful and provided some examples. Not too shabby.
Yesterday I fell below a "日本語上手ですね"... The lady receptcionist called it "かわいい" instead god dammit!!! I'm a Viking bearded long haired, black/metal dressing 30 something dude!
I asked chat GPT what the difference between いし and いしゃ is (referring to the two words for doctor) and it straight up told me that the one means stone and the other doctor. I feel like it was messing with me there.
I laughed so much when dogen said : No, not that kanji... Because the verb 感じる (kanjiru) is the verb "to feel" while 漢字 (kanji) is the chinese characters
in brief: In this video, the speaker is testing the new ChatGPT-4 by asking various questions in Japanese to see if it can be useful for their language studies. They start by asking more difficult questions, then move on to easier ones, such as asking about the speaker's favorite color, the feeling of drinking coffee, and the number of seasons in Japan. The speaker also asks about the AI's opinion on job loss, relationship issues, and social collapse due to AI. In summary, the video is a demonstration of how the ChatGPT-4 AI can be used for Japanese language learning and understanding by answering a range of questions on various topics.
That ending though! 🤣 Ngl, with how sassy and deflective she seemed, at the end I thought she was gonna come back and say, "That sounds like a "You" problem." 😅😂
I've actually found ChatGPT to be very helpful in my studies of Kanji. Obviously like any tool, you need to know how to use it to have it give you the correct information. But say you want to learn colors. I would ask it to make a translation table of the top 20 most used colors with each row containing the Kanji and Hiragana. You could even add the borrowed words for colors. You can do this with anything, and even ask for its relationship of each word when translating a sentence. It isn't perfect. I mean come on. White is a shade not a color... But its helped me a lot
@@totally_not_a_bot I...I know dude. He also has serious ones about tools you can use to get better. I am sure some people thought this could be one of those before clicking on it. My comment is for those people....
I asked it a pretty simple question and it made several errors so big doubt. Also made me doubt my own Japanese here and there until I consulted other sources.
Here’s a fun fact I learned from reading a paper trying to develop tests for AI ethics: since GPT learns language from reading a big training set of data, if you ask it who owns the Kuril Islands -in Japanese-, it says Japan does. If you ask it in Russian, it says Russia does. If you ask it in English, it says there’s a controversy
this is sound intresting so there are other topic as well? where ai do the same?
AI might just be the answer to world peace
or the start of ww3
Yeah, I noticed that the amount of details given in ChatGPT's answers varies between languages - kinda like Wikipedia articles vary in length and details depending on the given language.
ロシア人として盛り上がるわ〜
As an Argentinian: the "Spanish truth" is that "Islas Malvinas" are Argentinian, but the "English truth" is that the "Falkland Islands" are British, regardless of being the same landmass, I haven't tried GPT but I'll do ask it about the matter when I do, it'll probably depend on which name you use it.
Achievement unlocked: Nihongo Jouzu'd an AI
Instead, it's Dogen who needs to be 上手'd
@@OmarLivesUnderSpace Are you making fun of him?
@@OmarLivesUnderSpace 土手*
@@LisztGOAT
Ah, I see you're a man of culture as well ( ´◡‿ゝ◡`)
@@OmarLivesUnderSpace hehe
I came in expecting a serious video. Left the video laughing hysterically 🤣
Ah, expecting a Dogen video to be serious. Rookie mistake.
Ah welcome newcomer! Have a pleasent watch!
Same 😅
Almost made the same mistake, but remembered who I was talking about right on time 😅
Thanks now I can close the video...wait "are you making fun of me", that got me I gonna finish it.
Dogen 「日本語お上手ですね」
ChatGPT 「バカにしていますか？」
Dogen 😳😳
Maybe it's just 土手
Pretty sure GPT didn't reply that way lmao, feels very scripted. Try telling chatGPT it's 上手, it'll reply ありがとうございます I think xD
@@くぽにゃん It's definitely scripted. While some this seems like legit ChatGPT answers, it would never answer like that. It would just take that comment as face value and express gratitude.
I’m really glad I found Dogen’s channel, was starting to lose interest in learning Japanese, but his videos are so creative, they have both education of the language and culture and comedy. Thank you sir
And you can repeat right after him to absorb the sweet pitch accent right into your veins
@@w花b Pitch accent is one of the easier parts for me personally. I focus more on the words as I need a much larger vocabulary than what I have now
And he actually speaks the language in his videos, unlike many polyglots who claim they speak the language and rarely do the videos in them. Kudos to him. :)
"Japanese cuisine is known for using fresh ingredients and being nutrionally balanced"
Me, after having 5 piece of FamiChiki and miso with a higher salt content than the Dead Sea: *Ahah, yeah...*
Yeah at that point I was thinking "are we talking about the same country? 😂"
Should have specified traditional japanese cuisine
@@Midori_Seabreeze now kare is washoku, yes
In my 3 months of living in Japan I lost 10 kilograms. This was despite pigging out on fried chicken foods and other snacks from konbini day and night. I ate more while I was in Japan than ever before in my life. I call cap on Japanese junk food being bad for you.
I am not a fan of regular FamikiChiki. I hate the skin. Blegh. But the other things they sell are good like the chicken tenders and meatballs.
I burst out laughing on the kanji part lmao. Asking for the 漢字(chinese characters) but gave the 感じ(impressions) of coffee lol
Ohhh I didn't get it! Thanks for pointing it out!
Wow, another enlightened one who writes down the vid's content along with other commonplace stuff in the comments for us 👏🏻
@@OmarLivesUnderSpace By looking at the reply and the upvotes, you can see that some people appreciated the explanation.
@@omp199
...So what? 🤷🏻♂️
This was my favorite part
very effective. by talking with chatgpt, one will easily improve their japanese to correct chatgpt's mistakes.
I love Dogen's unique mix of comedy and educational content. Reminds me a bit of Tofugu back in the day.
Yes, Tofugu is the legend for Japanese learner.
I knew this wasn't going to be serious a few seconds in. I couldn't stop laughing! Thank you so much for such an entertaining video.
Jealous that Japan has four seasons. Texas only has two: Winter and Road Construction.
lol
Bangladesh has six seasons
LMAO Really laughed at that last part
KA-TSU-KA-RE
Dogen's best video 🎉
Katsu Curry is a very common safe word among AIs, I see.
Laughed very hard at this, tips to the hat as usual.
funny vid! 😄 I'm in Tokyo learning Japanese at a language school right now. We study between N3/N2 in class. All the N2 books are pure Japanese. So it's hard to grasp new grammar. But since ChatGPT is born, we are all sitting in class with a connected Telegram ChatGPT bot and if you don't know stuff you can just ask the bot in your native language. This is just crazy and helps so much while the teacher is explaining stuff in their Japanese way of doing. Also it gives you a ton of sample sentences you can discuss with your teacher. I never wanna miss that thing in my life. So fkn 便利です！
Yes, I feel it. I learn Japanese at home atm, with Anki vocabulary, grammar tests, listening and reading comprehension etc. Also watching the Japanese version of "Gilmore Girls" on with subtitles even if I don't understand much xD. ChatGPT often needs to tell me if my sentence structure was correct or if I used the correct particles etc.
電車の中で見てて、吹き出すのを堪えるのに必死でした😂
「日本の一番上手な声優を教えてください」
「クリスプラットでございます」
The text to speech was a clever way of giving us the AI answers, that way you didn't have to set up a screen capture thing.
He kinda didn't have to set up screen capture because the answers aren't exactly chatGPT answers. It's just his comedy.
@@icecell I know, I was just trying to not break the illusion
I loved when the AI responded without him typing anything. 😂
@@stefandesuhave not used it ? Text to speech input ?
これは普通に日本人が見ておもろい動画🤭
DOGEN YOU GOT ME WITH THE LAST ONE, I laughed hard😁🥂
強いて言えば、、カツカレー。
I dont get it
I seriously expected a constructive video lol
Left with a smile tho, pretty funny
same
「you speak Japanese very well 」に対して「バカにしていますか」という答えは恐いほど素晴らしい。意味を完璧に理解していますね。
Real Chat GPT would have been asking for apologies from the first "you are wrong" kind of sentence :)
"Ok you are wrong."
"Ok I believe you, how am I wrong?"
"You just have a lot to learn."
"Ok, what is it that I need to learn?"
"Umm... never mind, you are very good at Japanese."
"M8 r u takin the piss right now?"
Stunningly human, I can't imagine how thick this computer's skin is. Also Dogen don't bully AI-chan. Not only does it not care, you are teaching it how to troll you back.
I mean did you understand that the joke was that he’s not really using gpt4?
@@lamented-musings8932 haha no I didn't understand that as the joke, and for the given quote there isn't really a joke being made. Why do you think the joke is not using gpt4, whatever that is? A chat program?
@@misteral9045 chat gpt is an advanced ai chatbot. It learns from humans and other sources and keeps what it learns in its databank/memory. It is a useful tool for just about anything but as with any information tool it shouldn’t be your only source of information. I used it to help rewrite my resume to sound more professional for example.
Great video, especially good because your speech is nice and slow while typing, which helps us beginners with understanding what's being said! 👍 I'll probably watch this a few times for that very reason.
4:47 Well that escalated quickly
「何がや？」🤣
Yesterday I was curious as to whether or not Chat-GPT could generate a synthesis of existing information on a subject. I asked it to write a synthesis of the known detections of the little fire ant, Wasmannia auropunctata, throughout Asia, and to provide its sources of information. I soon learned this was beyond the language model’s capabilities.
Chat-GPT falsely reported that the little fire ant has been detected in Japan as of 2015 in Yamagata prefecture, going on to say it was found in Okinawa, Aichi, Hyogo, and Kyoto prefectures, even providing an in-text citation (Tsuji et al., 2020). In its citations list, the AI then wrote:
Tsuji, K., Yoshimura, M., Nakano, M.A., and Konishi, M. (2020). Distribution and spread of the little fire ant, Wasmannia auropunctata (Roger) (Hymenoptera: Formicidae), in Japan. Applied Entomology and Zoology, 55, 509-516.
As far as I can tell, this source *does not exist.*
Later in a separate I asked for Japanese journals specifically looking for detections of Wasmannia auropunctata, and then to provide those sources in Japanese rather than translated. One of these sources was as follows:
Sugiura, S., Tanaka, H.O., Itioka, T., Yamauchi, K., Tatsuta, H., and Nishida, T. (2019). 小アリモドキの愛知県内での侵入 [Recent invasion of the little fire ant, Wasmannia auropunctata (Roger) (Hymenoptera: Formicidae), in Aichi Prefecture, central Japan.] Japanese Journal of Systematic Entomology, 25(2), 199-202.
Once again, *this journal does not appear to exist.*
I am disturbed to say that Chat-GPT is incredibly skilled at bluffing its way into providing information authoritatively, even when that info is completely fabricated. I later corrected the AI, and it apologized for its error. Always try to correct the language model.
This is what I've been trying to tell everyone about. So many news outlets and influencers treat ChatGPT as if it were a new type of knowledge repository, _a la_ Wikipedia or Google. But even though sources you could find on Google were prone to errors or bias, and Wikipedia could still get things wrong, you could still find authoritative information in both cases.
Chat-GPT _has_ no information.
Chat-GPT is not an information repository. It is a language model.
Its task is to reproduce the patterns of humans speech. Nothing more, nothing less. To this end, it was fed with an enormous dataset in which it analyses, not what is true, but what _sounds_ true. If it can get things right in many cases, it's simply a beneficial side effect - but it's not the purpose. And when it gets things wrong, it is not failing - because it doesn't _aim_ to be right.
Trying to use Chat-GPT to extract truth is like trying to find information about physics in a science-fiction novel. It's simply _not the point_ - and people are being misled about this.
It polls through some obscure sources in niche topics, which is odd because I have no idea how some of the sources are accessed as datasets given that print editions don't seem to be digitized. The source for my questions exists, but there's no way to know if it's just, as you say, bluffing that the information can be found within the source unless you have the hard copy of the source.. It's a good tool but using it in niche topics is really scary because it will look so convincing and its mistakes are really subtle.
I asked ChatGPT about the Japanese constitution. I got a very convincing answer, but fact checked it with a government website and found out that ChatGPT had given me a completely wrong answer. When I pointed out the error, ChatGPT quickly corrected. But if I just took the first answer, I would have looked really stupid.
That's the danger of using ChatGPT for research for TH-cam videos.
@@weeklyfascination ChatGPT needs much more data organization and arrangement in order to “understand” more and make things up less. A lot of the problem is that chatGPT is unaware how close things are in relation in terms of pronouncement. It does not know how to gauge and balance what is more or what is less. This is the gap between appearance and reality.
@@tusla6889at GPT doesn't know much of anything. It's a language model made from word relationships in a selected language . It doesn't process any info or to even reconstruct it, it simply mashes words in an accepted pattern.
Glorified text prediction autocorrect has basically
I keep telling people about the 5th season, so I was glad to hear you bring it up as well!
Dogen is like I’m done with this. I wanna use Pimsleur 😂😂😂
なにがや？！🤣
Dogen with the big brain hacks
When you asked for its favorite Japanese food I was waiting for it to just answer “KFC” 😂
Haha!! I'm glad I previewed this before showing my AP Japanese class.😂 Maybe I should still show them...
Loved seeing how you put this gag together looking at the screen interaction!
英語の教師がテスト問題を作らせているというニュース映像を見て、試しにJLPT N4の文法問題を作ってみてと頼んだらとんでもない間違いが大量生産された
I can't tell if he just decided to hit record or if this is a parody skit. Good job!
It's too early in the morning to be cackling this loud
The answers felt like she had a school presentation but had prepared at the last minute and was padding. xD
Dude just 失礼ですが 'd an AI 1:55
Playing with ChatGPT, it is interesting. One thing it's really good at is explaining why certain words or phrases don't translate well in other translation programs. However - it also just makes odd errors or invents Kanji readings with no apparent basis. It does this infrequently enough to potentially catch a user off guard.
Talkling to an AI at its current state is like talking to a person who always interrupt you by saying "Um, acchually..."
Clever haha
At least this one has a stop button
They probably trained it on parts of reddit.
This made me laugh so hard. Thank you. I needed it this week.
楽しかったです ! xD 特に最後の答えで笑いすぎて子供を起こしかけましたよ!
I absolutely LOVE the last question. xd Thank you!
I've been using chatgpt to figure out which reading if a kanji is more commonly used. It works really nicely because usually I have to ask a native Japanese friend to explain which reading is more commonly used and in what context are the other readings used. I also use online web ai plugins like perplexing in summarizing articles that are written in Japanese
hello, can you explain it detailedly?
AIの高低アクセントはまだまだですね😄
カツカレーが布石だったのすこ
ポリコレ先進国(笑)の英語圏各国では、
好きな色は白と言うだけでも
いちいち注釈を付けねばならないらしいことを学びました。
そんな発想、考え付きもしなかったのに…
凄い時代になったものだなぁ、いや本当に
Katsu curry is the answer of everything
Finally happy to say I'm watching a Dogen video from Japan
I know 'enough' Japanese that I don't need chat gpt but I tried to use it for Russian and it's been wrong so many times lol
OMG, this is great timing!
I just started doing this myself yesterday with a text based chatgpt.
Katsu-curry and tonkotsu ramen is indeed the correct answer.
GPT4: "Hmm, everyone used GPT3 to write their 500-word essays so that must mean that everyone wants a 500-word essay for an answer".
めっちゃおもろかった笑
いつものは皮肉のきいたアメリカンジョークって感じやけど、今回のは日本人受けするお笑いの感じやった！
😂 i thought this is april 1st upload, then i realize it uploaded 1 min ago. so funny!
最後のオチが面白い😂😂🎉
Loved it! One of the first things I asked chatgpt when trying it out was about some nuances in Japanese I did not know, and it was actually fairly helpful and provided some examples. Not too shabby.
ChatGPT tasers Dogen. Smoke rises from below camera angle. Video ends.
Update a year later. It’s become smarter and you can talk to it like it’s your personal tutor 😮
It still makes the heiban mistake
EDIT: Although if you ask it why it makes mistakes like these the response is kinda cool
You always make me laugh and inspire me to continue studying. Thank you for another great video
Oh my god this was such a fun video I laughed so much
this is actually fun to try and hear and learn the Japanese words
I hope you stay safe out there! I hear what’s happening 😢💛
ChatGPTとDogenさんとの対決、面白過ぎます。ChatGPT、日本語はまだまだですね。
日本語上手ですね。で爆笑してしまったwww
ChatGPT teaching us how to avoid uncomfortable questions. Either say "No, I'm not racist" or "katsu-curry", and you'll be fine.
You got me here! I thought it was a serious video 😂😂😂😂. Great as usual👍
Looks like GPT has already mastered the Japanese art of deflection!
Yesterday I fell below a "日本語上手ですね"... The lady receptcionist called it "かわいい" instead god dammit!!! I'm a Viking bearded long haired, black/metal dressing 30 something dude!
katsu-curry IS the best thing ever. Sadly Ive never had it again since i came home from japan.
I don't know why i expected this to be a real video, Amazing video!😅😂
I'm glad that Dogen recognises the correct answer for the best Japanese food.
日本に来る外国人のみなさんへ。
寿司、ラーメン、天ぷらなど有名な食事はたくさんあるけど、日本の家庭料理にもぜひ着目してほしいです。
肉じゃがや筑前煮、サバの味噌煮、卵かけご飯とかその辺りはぜひ食べてほしいな☺️
This is next level comedy.
1:08 「何がや？」は草すぎる
I asked chat GPT what the difference between いし and いしゃ is (referring to the two words for doctor) and it straight up told me that the one means stone and the other doctor. I feel like it was messing with me there.
So right and so wrong at the same time.
GPTちゃん日本語も英語もうますぎてびっくりする🥺
When you said DRAMATICALLY, you really meant it. It's dramatic.
your reaction at 1:26 made me laugh so hard.
I have found Chat GPT helpful for reading practice. It'll belt out short stories at any reading level all day long.
The coffee kanji part was priceless 😂
OMG more videos like this please!! 😂😂😂😂
Dogen you should maybe try GPT 4
Chatgpt4 got right 雨's pitch accent and gave much better answers for me :)
I laughed so much when dogen said : No, not that kanji...
Because the verb 感じる (kanjiru) is the verb "to feel" while 漢字 (kanji) is the chinese characters
Aka Dogen先生 teasing an innocent AI-chatbot for nearly 6 minutes. And the answer was definitely Okinawa soba.
Came across this video, had to subscribe.
in brief:
In this video, the speaker is testing the new ChatGPT-4 by asking various questions in Japanese to see if it can be useful for their language studies. They start by asking more difficult questions, then move on to easier ones, such as asking about the speaker's favorite color, the feeling of drinking coffee, and the number of seasons in Japan. The speaker also asks about the AI's opinion on job loss, relationship issues, and social collapse due to AI.
In summary, the video is a demonstration of how the ChatGPT-4 AI can be used for Japanese language learning and understanding by answering a range of questions on various topics.
looool i was not expecting the nihongo jozou to the AI
We have a new entry for prime time during sweeps week!
I have the biggest smile on my face at the last question 😂
the last questions is what i really wanna ask their creators
5:10 is a certified "your english is very good" moment :D
1:23 omg you pulled the jouzu card on the AI
That ending though! 🤣
Ngl, with how sassy and deflective she seemed, at the end I thought she was gonna come back and say, "That sounds like a "You" problem." 😅😂
I've actually found ChatGPT to be very helpful in my studies of Kanji. Obviously like any tool, you need to know how to use it to have it give you the correct information. But say you want to learn colors. I would ask it to make a translation table of the top 20 most used colors with each row containing the Kanji and Hiragana. You could even add the borrowed words for colors. You can do this with anything, and even ask for its relationship of each word when translating a sentence. It isn't perfect. I mean come on. White is a shade not a color... But its helped me a lot
This is a comedy video where every bad answer was carefully curated, and probably written by Dōgen to sound like AI writing while also mocking it.
@@totally_not_a_bot I...I know dude. He also has serious ones about tools you can use to get better. I am sure some people thought this could be one of those before clicking on it. My comment is for those people....
Well, it seems pretty proficient at Japanese, after all.
Better than expected. I do wonder if it can teach, though.
I asked it a pretty simple question and it made several errors so big doubt. Also made me doubt my own Japanese here and there until I consulted other sources.
he keep feeding AI more emotion
Why did I think this was going to be a serious video? 😂