SlatorPod - #219 PR Wars over LLM Translation Quality

#219 PR Wars over LLM Translation Quality

07/19/24 • 32 min

1 Listener

Florian and Esther discuss the language industry news of the week, with a nearly USD 4m settlement involving Language Line Services and ongoing disputes in Canada, where freelance interpreters are protesting against unpaid breaks during debates.

Florian highlights a YouTube video about UN simultaneous interpreting and the rigorous standards and working conditions for UN interpreters.

Esther shares a significant procurement opportunity from the UK's Crown Commercial Service, which plans to issue a GBP 250m tender for various language services.

In tech news, the duo talk about DeepL’s new LLM launch that claims to outperform competitors in translation quality. They also touch on the trend of adding translation features to various SaaS platforms, like Airtable and Happy Scribe.

A tweetstorm by venture capitalist Olivia Moore from Andreessen Horowitz introduces the concept of "AI Scribes," suggesting a significant market opportunity for AI-driven transcription services.

In an M&A and funding corner, AI video startup Captions raised USD 60m at a USD 500m valuation and Mantra, a Japanese manga translation startup, raised nearly USD 5m. Toppan Digital Language acquired Austria’s Meinrad and Universal Music Group partnered with SoundLabs AI for voice cloning and language transposition in songs.

Florian highlights a YouTube video about UN simultaneous interpreting and the rigorous standards and working conditions for UN interpreters.

Esther shares a significant procurement opportunity from the UK's Crown Commercial Service, which plans to issue a GBP 250m tender for various language services.

A tweetstorm by venture capitalist Olivia Moore from Andreessen Horowitz introduces the concept of "AI Scribes," suggesting a significant market opportunity for AI-driven transcription services.

Previous Episode

#218 How Large Language Models Replace Neural Machine Translation with Unbabel’s João Graca

João Graça, Co-founder and CTO of language operations platform Unbabel, joins SlatorPod to talk about the present and future of large language models (LLMs) and their broad impact across all things translation and localization.

First, the CTO explains how Unbabel was founded to address language barriers for people using services like Airbnb, combining MT with human validation to improve translation quality.

João believes that LLMs are quickly replacing neural MT models as much more R&D is going into LLMs vs NMT. He highlights that LLMs can handle more complex tasks like automatic post-editing, source correction, and cultural adaptation, which were previously difficult to achieve with traditional models.

He also tells the backstory of the company's decision to develop TowerLLM. João shares how Unbabel's approach involves using open-source LLMs, fine-tuning them with multilingual data, and applying techniques like retrieval-augmented generation to improve translation quality in production settings.

Despite the advancements, João acknowledges that human intervention is still necessary for high-stakes translation tasks.

The podcast concludes with the hiring environment for AI talent and the future directions for LLM development, with João expressing optimism about the continued progress and potential of these models.

Next Episode

#220 Building an AI Video Localization Startup with Supernative Co-founder Paul Ghio

Paul Ghio, Co-founder of AI video localization startup Supernative, joins SlatorPod to talk about how language AI is reshaping how we create and consume content globally.

Paul shares how his passion for helping creators and his experience with internationalization at Shutterstock and Yelp led him to start Supernative, with a mission to democratize video localization using AI.

The Co-founder outlines Supernative's technology stack, which includes well-known models like Whisper for speech-to-text, ChatGPT for translation, and ElevenLabs for speech generation. He highlights the importance of improving these core capabilities, such as voice cloning and lip-syncing, to deliver high-quality dubbing at a competitive cost.

Paul discusses the current landscape and challenges of the AI localization market. He notes that while there is significant interest and potential in AI-driven localization, retaining users remains a challenge due to the novelty factor wearing off.

He elaborates on Supernative's strategy to introduce collaborative features to better align with the workflows of content creators and businesses, which could be a more sustainable strategy for user retention and growth.

Paul also touches on their long-term vision, including more comprehensive localization tools that cater to cultural and contextual nuances, and the possibility of localizing video content itself.