AI has taken the world by storm and that much is evident by now. People are using it for any number of things, including language learning. And it can actually be an excellent tool to help you learn a language. It can provide quick translations, explain grammar rules, and give you all the examples you want. But it’s not a very reliable tool.
As experts have been saying over and over, we need to know that AI makes mistakes, sometimes serious ones. And when it comes to language learning, it’s no different. In fact, for many rare and lesser known languages, these errors could be really grave. That’s why using it as your only resource for language education might not be the best idea. Now, using it as a companion to a language app like Ling is another story.
Since it particularly struggles when teaching rare or uncommon languages, in this post we’ll go over 7 languages AI can’t teach well. We’ll break down some of its limitations and show you a better way to learn these languages.
I want to learn even more languages with Ling!
Download now!
7 Languages AI Can’t Teach Well
There are many languages AI can’t teach well. The biggest reason is the lack of error-free, authentic, credible learning resources available online. We need to understand how AI works. AI doesn’t create information, it gathers information and then give it a written form – sometimes in the form of essays, poems, phrases, vocabulary terms, poems, and so on.
The other reasons are grammar complexities and the fact that some are tonal languages, and there us difficulty accessing audio, and several other obstacles (we’ll go over all of them in a bit).
Additionally, AI isn’t usually trained on languages that aren’t the most popular, such as English, Spanish, Chinese, Arabic, and others. So it’s bound to eventually give you incorrect information or have a hard time explaining things properly when you’re learning a not-so-popular language.
These are 7 languages AI particularly struggles with. Let’s go over why, if you’re learning any of them, you’d do best to favor a language learning app like Ling over ChatGPT.
1. Khmer
Khmer is the official language of Cambodia. It’s also spoken by the Khmer people in Thailand and Vietnam. With around 19 million native speakers, it’s not a very popular language outside of these regions, and this means there are not many resources for learning it, especially online.
As such, AI doesn’t have access to enough information to teach it properly. While it may be able to give you a general overview of the language and its grammar, it won’t be sufficient for a deep understanding, let alone true fluency. This means it’s no substitute for traditional methods or an app to learn Khmer.
2. Lao
Another Southeast Asian language, Lao is the official language of Laos. It’s also used in northeastern Thailand, where it’s usually referred to as the Isan language. It has around 3.7 million native speakers, significantly less than Khmer, so you can probably tell it’s just as hard to find learning resources for this language.
In fact, if you go to Google and try to find an app that teaches Lao, you’ll see you won’t find many options. Since the Ling app teaches rare languages, it has Lao in its catalog, but you’ll find that several apps just skip over it because of a lack of demand.
As such, just like with Khmer, AI doesn’t have access to enough learning material to properly teach the language. Add to that, Lao is a tonal language, and you might want to prefer to learn Lao with native speaker audio with an app like Ling.
3. Somali
Somali is the language spoken in what’s known as Greater Somalia. It’s an official language in Somalia and Ethiopia and a national language in Djibouti. It has around 24 million native speakers, and if you look online, you’ll find there isn’t much learning material available for this language either. Even if you try YouTube, you’ll find very few accounts that teach it, so it’s no wonder AI struggles with it.
I actually experienced this myself when trying to understand numbers in Somali. If you ask ChatGPT, you’ll find that it has a hard time figuring out if Somali numbers are said as “ten and unit” or “unit and ten”. That’s because the Somali language has several dialects, and this varies among them. So both ways are correct, but ChatGPT just doesn’t know it. That’s another problem with AI, it usually doesn’t consider dialectal variations of a language.
4. Sinhala
Sinhala is primarily spoken in Sri Lanka. It has approximately 16 million speakers, so, again, it’s not one of the most popular languages to learn. You know by now that this means there’s little information for AI to access on the language, which leads to inconsistencies or incorrect information.
But let’s consider something else. The fact that it’s not so popular a language also means audio content can be hard to find. Since listening is one of the key aspects of language learning, learning Sinhala (and any of these other uncommon languages) using AI will leave you with some deficiencies.
If you truly want to learn a language, you need to listen to how it’s really used, which is why the Ling app uses audio by native speakers. That’s something AI certainly can’t replicate, as much as it tries.
5. Pashto
Pashto, a language spoken in Pakistan and Afghanistan, has many of the same issues when it comes to AI teaching it properly. 51 million speakers isn’t as few as the languages we’ve just seen, but it still means there’s not enough interest in learning the language abroad for there to be a lot of material.
That means, much like with the other languages, that finding grammar guides and audio isn’t easy, so AI doesn’t have enough information to rely on to give you a proper understanding of the language.
6. Icelandic
With around 330 thousand native speakers, Icelandic is the rarest language on this list. It’s spoken, you guessed it, in Iceland, and it isn’t particularly popular outside of the country.
When it comes to learning materials, the particular issue with Icelandic is that many of them are made for advanced learners and academic research. This means AI will only have access to resources for people who already have some knowledge of the language and can struggle to teach beginners.
Moreover, Icelandic has a very complex grammar. It is a very conservative language, and that means it preserves many ancient features that would have otherwise evolved with use. It’s a heavily inflected language with four cases, nouns have three grammatical genders, it has many irregular verbs, and more.
Naturally, learning such a complex language with AI might not be very effective. And there’s also the fact that Icelandic pronunciation can be a bit of a struggle for people whose native tongue is English, and AI can’t help much with that.
7. Armenian
Last but certainly not least, as there are many languages AI can’t teach well, we have Armenian. This is the official language of Armenia, and has around 5.3 million speakers.
As with the other languages in this list, there are few resources to learn Armenian, which makes relying on AI to master it difficult. Moreover, it has a very unique script, and two main dialects with notable differences, which are things AI can struggle with as well.
Why AI Struggles When Teaching These Languages
While ChatGPT and other AI tools certainly know a lot, we know by now that they can also be inaccurate. AI can make up information, invent sources, reflect society’s biases, and more. So, while they certainly are a great tool to help you learn a language, they ultimately can’t be fully trusted.
When it comes to language teaching, there are some specific things AI struggles with:
Context
Some languages rely heavily on context. For example, there are many homonyms in Cantonese, that is, words that sound or are spelled the same, but mean different things. You need to know the context in which they’re used to be able to tell them apart. AI can struggle with that because, even if it knows the words, it’s not capable of understanding the context in which they’re used.
Complex Grammar
There are languages, like Icelandic, that have really complex grammar rules. Furthermore, these rules can be very different than those of English and other popular languages in which these large language models, or AIs, are trained. This means ChatGPT sometimes just doesn’t know all the rules of the language very well, which can lead to confusing (or downright incorrect) explanations.
Tonal Languages
Tonal languages like Thai or Vietnamese can be a bit of a struggle for learners whose native tongue doesn’t make this distinction. The same applies to AI, and even more so. AI can’t hear the differences in tone, can’t really interpret their complexities. Even when voice analysis might allow it to have a general idea of how a tonal language works, it’s going to have a hard time understanding and explaining all the complexities behind it.
Complex Scripts
Writing practice is essential for true mastery of a language. And while AI can give you the characters, it will have a hard time explaining how to properly write them. For example, when writing Japanese characters, you always follow a specific stroke order. While ChatGPT can tell you that order with relative accuracy, it won’t be able to show you how to do it. So you might be better off using a script app instead.
Limited Resources
As I mentioned, AI struggles even more when dealing with less common languages. If there are few resources to learn them online, that is, grammar guides, vocabulary banks, audio for pronunciation, and more, AI just won’t have enough information. This can lead to incorrect statements (because we all know AI won’t just tell you “I don’t know”) and inconsistencies in its explanations. It’s trying to figure out a whole language based on the two books it can access online; you can’t expect it to do a good job.
Dialectal Variations
When you’re learning a language, most apps and platforms will teach you the standard dialect. This is practical and makes things easier, but it means that the majority of the language resources AI has access to don’t take dialects into account. As I mentioned before when discussing Somali and Armenian, this can be a serious issue.
If you ask AI if a sentence is correct, for example, it can confidently tell you that it’s not, even when it’s just a dialectal variation, not an error. This can lead to a lot of confusion with languages that have many dialects, such as Arabic, Hindi, Pashto, Māori, and more.
How The Ling App Teaches These 7 Languages (And More) So Well
As I mentioned several times, if you truly want to achieve fluency in a language, you’d do best to use AI only as a companion but rely on a more structured approach to learn the language. ChatGPT can give you a hand when you don’t understand how something works, explain why a certain phrase doesn’t fit the mold, and more, but it’s not sufficient for true mastery of a language.
The Ling app has that structured approach. The short, interactive lessons in its language courses are organized by proficiency level, so you start with the basics and gradually move on to more complex topics. Moreover, all content, even the chatbot, is crafted by real people on the Ling team, so you can be sure there’s careful consideration behind every feature.
What’s more, Ling specializes in languages that have few learning materials available, especially online. That’s why you’ll find many reviews praising it as a rare languages app; one of the few good ways to learn certain foreign languages that other apps overlook.
As for audio, on the Ling app, all of it is recorded by native speakers of the language you’re learning. This not only helps with your pronunciation, but actually lets you learn languages as they’re used by real people in their daily lives, as opposed to a machine. Plus, it’s an essential tool for learning tonal languages, where knowing the correct intonation can make a difference between saying what you mean and something entirely different.
When it comes to grammar, the Ling app includes Grammar Notes for some of its languages (we’re working on adding more!) so you have short but comprehensive explanations to go with the lessons. It also has Culture Notes for some languages, which help you understand cultural references and the traditions associated with the language you’re learning.
For complex scripts, the Ling app has writing practice for most non-Latin alphabets, where you can trace the foreign characters with your finger on the screen. And the Ling app will also teach you the standard version of a language, much like AI, you’ll know that from the start. Plus, a few dialects are included but have separate courses, so you know what you’re learning, such as Cantonese, Bosnian, Croatian, Serbian, and more.
In sum, the Ling app is designed by humans to teach you languages in a way that’s fun and entertaining, and actually makes sure you learn what you need carefully and step by step. ChatGPT and other AIs can certainly help, so if you have access to them, you can use them as a companion. Just remember not to trust them entirely!
Frequently Asked Questions On Languages AI Can’t Teach Well
What Are The Biggest Problems With Using AI For Language Learning?
The first big problem with using AI to learn a language is that you can't always trust the information it gives you. This means you'll have to verify everything it says with credible sources, because you never know what information could be made-up.
The second big issue with using AI for language learning is that it can't give you a full view of the language. You need to hear the language in context, practice speaking it, essentially, interact with the people who use it. AI can mimic these interactions with more or less accuracy, but it will never be the same as learning from an actual, real person.
Can AI Understand Any Language?
First things first, AI doesn't really 'understand' anything. It's trained to see patterns and generate human-like text, but it doesn't truly 'understand' any of it.
As for dealing with languages, it's fairly good when it comes to popular languages like English, Spanish, Mandarin Chinese, Arabic, and more, but it will struggle with less common languages. In simple terms, the less popular the language, the more AI will struggle with it, first because it doesn't have access to enough learning materials, and second because it wasn't trained on these languages.
What Can Language Teachers Do That AI Cannot?
Teachers can do many things AI can't. For one, they can feel. They can create connections with their students, understand their struggles, and empathize with their feelings. While AI can mimic human interaction, it's never ever real. Moreover, teachers can understand non-verbal language, see nuances in the way we communicate, and pinpoint where students are struggling and why. So while AI can be a great tool for language learning and much more, it can't substitute good teachers and their human knowledge.
Learn Languages With Ling! It Excels Where Artificial Intelligence Fails
Want to learn one of the languages AI can’t teach well? Time to give the Ling app a try.
Since its lessons are created by actual people, you’ll follow a more structured approach that makes sure you’re learning what you need and doing it step by step.
Plus, the Ling app has audio by native speakers in all its lessons, which no AI can match. It also has many review functions to make sure you remember what you learn, a spaced repetition system for better retention, fun and engaging exercises in the form of mini-games, and much more.
So if you’re ready to truly start learning your target language with a trustworthy teacher, download the Ling app. You can try it out for 7 days for free before making a final decision. Start your language learning journey with Ling today!
