Are we on the verge of removing all language barriers with AI?

Olga Beregovaya, VP of AI at Smartling, joins host Conor Bronsdon to tackle this question, discussing the evolution from rule-based NLP to today's powerful LLMs. Together, they confront the persistent challenges that stand in the way, like the English-centric nature of AI, domain-specific inaccuracies, and the unpredictability of model hallucinations. Olga unpacks the difficulties faced when striving for accurate, nuanced translation across all languages, especially under-resourced ones.

Beyond these hurdles, the conversation explores the cutting-edge opportunities and technical innovations driving progress, including RAG, the rise of purpose-built models, agentic AI workflows, and the potential of multilingual multimodality. Olga shares insights into boosting translator productivity, achieving more predictable quality, and the path toward human parity in translation, examining how technology and human expertise will shape the future of global communication.

Chapters

00:00 Introduction and Guest Welcome

01:14 Evolution of NLP: From Rule-Based to Machine Learning

02:40 Challenges in AI Translation

04:21 Biases in Language Models

05:28 Inference Time and Latency

05:44 English-Centric AI Models

08:53 Opportunities in AI Translation

09:14 Industries Benefiting from Language AI

10:36 Human-in-the-Loop Translation

12:06 Architectural Innovations in Language AI

16:20 Success with RAG Architectures

17:58 Multilingual Vectorization

19:54 Agentic AI in Translation

24:35 Data Sets and Data Privacy

28:30 Using Smaller, Purpose-Built Models

32:10 Future of AI in Translation

36:37 Conclusion and Farewell

Follow the hosts

Follow⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Atin⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Conor⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠ Vikram⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow⁠⁠⁠⁠⁠⁠⁠⁠ ⁠⁠⁠⁠Yash⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠

Follow Today's Guest(s)

LinkedIn Olga Beregovaya

LinkedIn ⁠Smartling

Check out Galileo

⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠Try Galileo⁠⁠