In this episode of Embedded, host Zain Verjee speaks with Asme Teka Hadgu, co-founder and CTO of Lesan AI, about the critical challenge of language accessibility in artificial intelligence. Based in Berlin, Lesan AI is pioneering machine translation for Ethiopian languages, notably outperforming tech giants in Amharic and Tigrinya translations. Asme discusses their unique approach to data collection, working directly with communities to preserve linguistic authenticity, and the broader implications of language barriers in AI development. The conversation explores how AI can be leveraged to preserve and revitalize languages while ensuring ethical development and cultural preservation.
Key Highlights
🌍 Lesan AI's innovative approach to machine translation for African languages
📊 Over 9 million translations completed for Ethiopian languages
🤖 Limitations of current AI models like ChatGPT for African languages
👥 Community-centered approach to data collection
🔐 Ethical considerations in language technology development
🌿 Preservation of cultural context and linguistic nuance
Episode Timestamps
00:00 - Introduction and welcome
01:20 - Personal Journey and Mission
02:51 - Language Barriers in AI
04:51 - Lesan AI's Approach
09:16 - Speech Recognition Project
13:26 - Technology Access Challenges
16:19 - Language Models Discussion
18:05 - AI Language Limitations
20:19 - Data Privacy and Ethics
24:43 - Bias Prevention and Cultural Context
25:55 - Advice for African Innovators
About Our Guest
Asme Teka Hadgu is the co-founder and Chief Technology Officer of Lesan AI, a Berlin-based startup revolutionizing machine translation for Ethiopian languages. With a background in computer science and a deep commitment to education, Asme leads the development of translation tools that outperform major tech companies in Amharic and Tigrinya translations.
Connect With Embedded
Website: www.therundown.studio
Twitter / X: @Rundown_Studio_
LinkedIn: https://www.linkedin.com/company/zvgtherundown/
Host: @Zain_Verjee