Will AI Erase All Language Barriers? | Smartling's Olga Beregovaya
04/23/25 • 40 min
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
Check out Galileo
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
Check out Galileo
Previous Episode
AI, Low-Code, and Shaping the Next Generation of Apps | OutSystems' Rodrigo Coutinho
What if you could turn a requirement document into a full enterprise application in just minutes?
Rodrigo Coutinho, co-founder and AI Product Manager at OutSystems, joins hosts Conor Bronsdon and Atin Sanyal to explore this new reality of AI-driven development. Rodrigo shares insights from OutSystems' nearly 25-year journey, detailing their early adoption of AI and the development of their AI platform, Mentor. Discover how the pairing of AI and low-code empowers developers, accelerates the creation of enterprise applications, and shortens the cycle from idea to deployment.
But this newfound speed brings its own set of challenges. The discussion addresses the hurdles of managing AI-generated code, contrasting experiences with traditional versus low-code approaches. Learn why a dev's focus pivots from syntax to strategy, pinpointing human creativity and ideation as the crucial limiter in today's development lifecycle.
Chapters
00:00 Welcoming Rodrigo Coutinho of OutSystems
01:30 OutSystems' Early AI Journey (Pre-LLM)
03:30 The LLM Revolution & OutSystems Mentor Emerges
07:30 The Critical Need for Validating AI-Generated Apps
12:00 The Shifting Role of the Modern Developer
13:30 Quality Control & Accountability in the AI Era
16:00 Low-Code's Edge in AI Validation
18:30 OutSystems Mentor: A Deeper Look
23:30 Choosing the Right AI Models (In-House vs Public)
27:30 Future Opportunities: Speed, Experimentation & Multimodal AI
37:00 The Use Case Hurdle & Final Thoughts
Follow the hosts
Follow Atin
Follow Conor
Follow Vikram
Follow Yash
Follow Today's Guest(s)
Website www.outsystems.com
OutSystems Mentor
LinkedIn Rodrigo Sousa Coutinho
Check out Galileo
Next Episode
Why Enterprises Need a Different Approach to AI Agents | Lyzr’s Siva Surendira
Agentic AI exploded in 2025, but how do businesses move beyond prototypes to deploy reliable, valuable agents at scale?
Join host Conor Bronsdon and Lyzr AI CEO Siva Surendira as they discuss the complexities of building and managing AI agents for enterprises. Siva shares his journey creating Lyzr, focusing on making powerful agent frameworks accessible and trustworthy for enterprise developers. They discuss the critical hurdles businesses face, including productionization challenges, ensuring responsible AI, and bridging the gap between rapid innovation and the stringent requirements of regulated industries.
Listen as Siva explains Lyzr's approach to embedding safety guardrails natively and learn about the nuances of multi-agent orchestration, including managerial, DAG, and hybrid flows. Siva also offers insights into the limitations of "vibe coding" for enterprise use cases and stresses the crucial role of robust evaluation (evals) and choosing the right models—from local open-source options to frontier LLMs. Explore the bottlenecks hindering adoption, like custom application integration and data readiness, and learn why Siva believes the biggest opportunity for agent companies may not lie in replacing SaaS platforms but rather in automating the mundane work currently performed by humans.
Chapters
00:22 Introduction and Guest Welcome
00:52 Enterprise Agent Framework
02:48 Building Enterprise-Friendly AI Frameworks
04:56 Enterprise Concerns with Vibe Coding
09:23 Safe and Responsible AI Implementation
11:05 Multi-Agent Orchestration
14:13 Challenges in Multi-Agent Systems
14:22 Enterprise Integration Bottlenecks
17:37 The Role of Low-Code and No-Code Solutions
19:55 Inter-Agent Communication Standards
21:49 Future of AI Agents in Enterprises
29:37 Evaluating AI Agents
36:34 Conclusion and Final Thoughts
Follow the hosts
Follow Atin
Follow Conor
Follow Vikram
Follow Yash
Follow Today's Guest(s)
Website: lyzr.ai
LinkedIn: Siva Surendira
Check out Galileo
If you like this episode you’ll love
Episode Comments
Generate a badge
Get a badge for your website that links back to this episode
<a href="https://goodpods.com/podcasts/chain-of-thought-601485/will-ai-erase-all-language-barriers-smartlings-olga-beregovaya-89859819"> <img src="https://storage.googleapis.com/goodpods-images-bucket/badges/generic-badge-1.svg" alt="listen to will ai erase all language barriers? | smartling's olga beregovaya on goodpods" style="width: 225px" /> </a>
Copy