'Cost of digital tutors will be much cheaper': How AI is making language learning more accessible

Founder of Bright Al, an app that facilitates the mastery of languages through generative Artificial Intelligence (AI) breakthroughs, Dmitrii Basalkin is a UAE-based entrepreneur with over 12 years of experience in building online education services. In 2017, Dmitrii launched Bright AI — a publisher of language learning apps that help beginners streamline their path to learning English, Spanish, French, Japanese, and most recently, Arabic.

Leveraging his experience in language learning, gamification, and AI, Dmitrii and his team have achieved 25+ million downloads of the language-learning apps, signifying a shift towards AI-powered educational frameworks. “I dedicated myself to language learning because it empowers millions of people to enhance their lives—a challenge I am determined to tackle,” says the Russian entrepreneur. “Language learning presents an ideal arena for AI application, innovation, and enhancement.”

This ambition has since transformed into a thriving business comprising 25 million users worldwide and “With 750,000 paid customers, we’re now ranked among the top 10 language learning app publishers globally,” he adds.

How AI can help us learn a language

The previous generation of language apps helped learners with grammar and vocabulary expansion, but the current era of AI-driven language learning operates on a different level, explains Dmitrii. “It enables streamlined and effortless language learning for beginners, adopting a method similar to how young children learn: through conversations. These conversations are designed to be suitable — easy and familiar enough to encourage engagement yet challenging enough to promote progress. AI is responsible for carefully selecting the appropriate level of difficulty.”

AI's capabilities for personalisation and gamification are leveraged as the key to efficient language learning. “With GPT-4 under the hood, AI dynamically builds a personal course for each student with the latest AI technologies instead of following one given unified set of learning materials for everyone. Generative AI has the potential to create unlimited, engaging exercises for fresh learning and better control of the previously learned material,” he adds.

Additionally, as feedback and repetition are crucial for language learning, AI can also provide accurate feedback on mistakes, says Dmitrii. “AI can remember every previous answer and every mistake a student has ever made, fixing it over time.

Expats learning Arabic

Learning Arabic is one of the most pressing desires of a lot of the expat population in the region. Can AI-powered learning make it easier for people to learn Arabic?

“In a country where Arabic is the official language and home to millions of foreigners, the chance of a lifetime might come through speaking Arabic,” says Dmitrii. “The lack of language knowledge may result in missed opportunities and poor communication may nullify all the positive impressions of an effective service.”

“Expats are mostly adults, and adults are very conscious about their goals, which often involves practical uses of language. Educational apps are usually the most efficient and convenient way for busy individuals to stay connected with the language,” he adds.

While this sounds appealing, one might wonder about the cost. “Based on the average tutor’s rate per hour, the average cost of digital tutors will be up to 100 times cheaper when compared to one-on-one lessons with a human tutor. And they will potentially deliver the same level of efficiency,” says Dmitrii.

How can AI teach us Arabic?

There is one issue specific to the Arabic language that AI can tackle effectively, says the founder. “The structure and mechanics of the Arabic language drastically differ from English and other languages (e.g., word formation, sentence structure). Students may need to gain an understanding of these differences, and how these concepts vary from their native language.”

“AI is great at explaining these differences in their native languages, coming up with relevant and effective examples that elevate understanding these unfamiliar patterns,” he adds.

Learning in context

Students usually dump learning languages for two reasons: lack of motivation and a rigid educational process that doesn't suit their goals. The language app, as a result, uses generative AI, speech recognition, and synthesis to develop written and spoken examples and vocabulary based on students’ personal goals to pave their path to fluency.

“There is a sure thing we are concerned about most and can relentlessly talk about — ourselves, our goals, achievements. To navigate this challenge, wise tutors know how to benefit from this, but their abilities may be limited. Until the current AI breakthrough, it was impossible to design a whole learning programme based on this aspect of human nature,” says Dmitrii, highlighting this as a key advantage of AI-driven language learning models.

AI Assistants

AI language learning apps can also utilise virtual educational assistants designed to enhance users' interaction with the chat-based interactive form, significantly transforming their learning experience.

“This innovative feature is now available to students in the UAE, catering specifically to native speakers of Arabic, French, Spanish, and Russian. The app leverages the latest breakthroughs in Generative AI to transform the educational process into an interactive dialogue. With the introduction of the assistant section, students can now ask questions in their native language to clarify complex aspects,” says Dmitrii.

Future of public education

Whether it’s Arabic, English or any other language, AI can drastically boost language learning in public education, believes Dmitrii. “For kids, AI-powered digital tutors will help them become more successful in mastering the school curriculum and provide extra practice to build strong language skills.”

According to the AI app builder, this will be a major shift in overall public education. “In regions with a shortage of human teachers (as is the case in the Mena region) AI-powered assistants will even play the role of educators themselves. Moreover, there are significant opportunities for improvement in English proficiency in the UAE and Saudi Arabia. Speaking broadly, the region will massively benefit from this development.”

The advent of Digital Tutors

AI-enhanced experiences can mirror interactions with a live tutor, who guides the student through unfamiliar territory using feedback, progressively increasing the difficulty level in vocabulary and grammar. “Technologies necessary to develop effective Digital Tutors are already here,” says Dmitrii. “With recent advancements in speech recognition and voice generation, generative AI can now hear and speak as well as humans in many languages. I’m certain that we will see the first full implementation before 2024 ends.”

“Of course, integration with the national curriculum and pilot projects in schools will take a bit more time. But I expect millions of adults and children to use Digital Tutors daily for language learning by next year,” he adds.

Challenges in implementation

A key challenge in developing AI solutions within language learning is a phenomenon known as ‘AI hallucinations’. “This term describes instances when AI systems produce incorrect or nonsensical information despite being trained on extensive datasets,” explains Dmitrii. “In the realm of language learning, such errors can mislead learners or hinder the learning process."

To address this challenge, the Bright AI team has implemented cross-validation techniques, wherein another AI model verifies the AI’s language outputs for accuracy before presenting it to learners. “Additionally, we’ve also established feedback loops that enable users to report errors. This input is then leveraged to refine our prompts, algorithms, and models for future improvements,” says Dmitrii.

Another aspect to keep in mind, he adds is that AI models are not fundamentally dedicated to language learning and hence, are not designed for it. “They serve a different purpose. To pursue educational goals, any AI model needs an educational framework: a methodology that always originates from direct experience in the subject matter. So, no doubt, there’s a long journey ahead when it comes to perfecting the use of AI in Arabic learning.”

Economic growth and innovation ecosystem

Recently, Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum, Crown Prince of Dubai, shared his vision on AI marking the onset of a new chapter in human history, and how Dubai endeavours to lead this revolution. Aligned with this perspective, Dmitrii seeks to leverage existing AI capabilities to enrich language acquisition in the region, encompassing both English and Arabic.

“I believe we’ll see many more tech companies choosing the UAE for their headquarters, development centres, and capital markets,” says Dmitrii. “I want to contribute to this movement because a lot is happening in the field. For instance, on August 30, 2023, Jais, the world's highest-quality Arabic Large Language Model, was introduced and open-sourced. We look forward to applying large language models developed in the UAE to benefit everyone worldwide.”

According to the entrepreneur, deep personalisation in Arabic is the next big thing to be achieved. “Given its complexity, we are currently examining contemporary tools and models to determine which can benefit students the most.”

somya@khaleejtimes.com

ALSO READ:

• From code to conversation: Can AI teach robots Arabic?

• Look: Forest inside Dubai Metro stations? Designer uses AI to show transport of the future

• Watch: AI robot patrols Dubai beach to monitor e-scooter violations

• Business
• Newsletters
• Dubai