# 190 Where the ‘Translator Still Feels Like a Translator’ with Bureau Works’ Gabriel Fairman

New Section

In this section, Gabriel Fairman discusses his perspective on generative AI and its role in translation.

Generative AI in Translation

• Gabriel views generative AI as a contextual manager that can emulate linguistic inferential capability, rather than just a replacement or enhancer for machine translation.

Gabriel's Background and Company Overview

Gabriel provides insights into his background, the founding of Bureau Works, and an overview of the company's services.

Background and Company Introduction

• Gabriel Fairman is the Founder and CEO of Bureau Works, a cloud-based translation management system.

• Bureau Works focuses on helping companies translate content with maximum leveraging from AI or other contexts, ensuring quality, cost-efficiency, and governance.

Evolution of Bureau Works

Gabriel delves into how Bureau Works evolved from a traditional translation agency to a tech-enabled business.

Evolution Journey

• The company functions as a CAT tool, business management system, translation management system, and quality management system but emphasizes an integrated approach towards a single goal.

• Gabriel advocates for abstract thinking in tool development to ensure longevity and effectiveness.

Origin Story and Growth

Gabriel shares the origin story of Bureau Works and its growth trajectory.

Origin Story

• The genesis behind starting the company was getting fired from a corporate job in his early 20s. His passion for languages led him to venture into translation.

• With a background in linguistics emphasizing decision-making processes, Gabriel started translating after being encouraged by his mother who was an interpreter.

Transition to Tech-Enabled Business

The discussion centers around how Bureau Works transitioned into becoming a tech-enabled business focused on information management.

Tech Transformation

• Despite starting as a translation agency, Gabriel realized the core challenge was information management. This realization prompted the pivot towards becoming tech-oriented.

Translation Industry Workflow and Large Language Models

In this section, the speaker discusses the rigid structure of the translation industry workflow and reflects on the adoption of large language models within the localization and translation sector.

Translation Industry Workflow Challenges

• The speaker critiques the complex structure of the translation industry workflow, emphasizing that people primarily seek efficient solutions rather than adhering to traditional processes.

• There is a need to reevaluate existing practices in the industry, moving towards a more abstract perspective to determine essential requirements.

Adoption of Large Language Models

• The emergence of large language models is disrupting traditional workflows in the translation industry, with new companies entering the market without preconceived notions.

• Despite experimentation and research, a significant portion of the industry has not fully embraced large language models for practical use.

Understanding Generative AI in Localization

• Generative AI is viewed as a contextual manager with linguistic inferential capabilities rather than just an enhancement layer for machine translation.

• Many individuals approach large language models cautiously, opting to minimize their impact until further clarity emerges on its implications.

Impressions of Generative AI Implementation

This segment delves into how generative AI is implemented in production within localization processes and its transformative potential.

Key Aspects of Generative AI Implementation

• The success of generative AI implementation lies in its seamless integration into existing workflows while preserving translator autonomy and decision-making.

• Generative AI facilitates a novel mode of interaction between translators and computers, akin to social media engagement where learning from behavior enhances output quality imperceptibly.

Enhancing Translator Experience

• Generative AI enables translators to personalize content through rewording options, ensuring that their style influences future translations consistently.

• Unlike traditional methods focused on syntax, generative AI operates semantically, allowing for inferred meanings and enhanced suggestions based on context.

New Section

In this section, the discussion revolves around the flexibility of the user interface in translation tools and how translators interact with the system.

User Interface Flexibility and Translation Process

• The UI is highly configurable to cater to individual preferences, allowing users to adjust settings like suggestions and translation smells. -

• Translation smells act as guardrails by flagging semantic issues such as inaccuracies, wordiness, and omissions for translators' benefit. -

• Translators have the freedom to modify the initial feed they receive, enabling them to teach the engine new patterns for future translations. -

New Section

This segment delves into the frustrations faced by translators during machine translation post-editing and contrasts it with a more adaptive approach.

Translator Experience in Post-Editing

• Traditional machine translation post-editing often leads to repetitive corrections without improvement in feeds, causing frustration among translators. -

• The adaptive nature of the new approach fosters a sense of collaboration between translators and machines, enhancing productivity and authorship feelings. -

New Section

Here, there is a discussion on linguists' perspectives regarding translation technologies and their varying attitudes towards adoption.

Linguists' Perspectives on Translation Technologies

• Many translators exhibit divided opinions on technology adoption, with a significant group expressing resentment towards it. -

• Some translators view technology users negatively while others struggle with complex workarounds due to perceived necessity for complication in adopting new tools. -

New Section

This part explores initial prejudices among translators towards new technologies followed by their eventual curiosity and increased productivity.

Translator Adoption Journey

• Translators initially face prejudice but transition into curiosity upon experiencing enhanced productivity while maintaining a positive outlook on using new tools. -

New Section

In this section, the speaker discusses the benefits of a streamlined workflow and enhanced performance data in decision-making processes.

Benefits of Streamlined Workflow

• Enhanced performance data leads to better decision-making.

• Simplified processes result in more streamlined situations.

• Emphasis on results over methodology for improved outcomes.

• Simplification allows individuals to focus on core tasks effectively.

New Section

The conversation shifts towards cost implications and benchmarks associated with utilizing language models.

Cost Implications and Benchmarks

• Concerns about increased costs when running large volumes of words through models.

• Cloud computing optimization crucial for cost efficiency.

• Language models likened to mathematical complexity rather than linguistic understanding.

New Section

The discussion delves into the evolving sophistication of language models and their potential impact on future applications.

Evolving Sophistication of Language Models

• Transformation of language into numerical sequences enhances complexity.

• Balancing excitement and apprehension regarding the rapid evolution of language models.

• Understanding the intricate correlations within language models beyond word prediction.

New Section

Exploring the perception of language engines as sources of truth and their role in interacting with natural language through numbers.

Perception of Language Engines

• Language engines viewed as instruments for interacting with language through numerical representations.

• Appreciating the sophistication and possibilities offered by advanced language structures.

New Section

In this section, the speakers discuss the challenges of translation and the complexity involved in achieving accurate translations.

Why Translation is Challenging

• The difficulty of translation lies in the complexity of capturing context and associations beyond literal meanings.

• Translating well requires accessing decades worth of information non-linearly, making it hard to replicate effectively.

• Viewing translation as a tool to enhance human experience rather than replace human translators can lead to better results by incorporating additional context like translation memories and terminology.

• Comparison between GPT models and traditional translation engines reveals that these models were not primarily designed for translation tasks.

Impact on Industries

This part delves into how technological advancements, such as HeyGen's capabilities, are reshaping industries and consumer perceptions regarding translation services.

Technological Advancements in Translation

• Technologies like HeyGen offer cost-effective solutions for creating multilingual content seamlessly, impacting industries by changing how people perceive translation services.

• The accessibility and affordability of advanced technologies may lead to a shift in how businesses approach multilingual communication strategies.

Future Challenges

The speakers explore future challenges related to maintaining relevance in an evolving industry landscape where technology influences traditional practices.

Balancing Technology with Human Expertise

• Striking a balance between leveraging technology for efficiency while respecting human expertise and brand identity poses a challenge for industry players.

Intertextuality and Changing Dynamics of Text

In this section, the speaker discusses the evolving relationship with text and introduces the concept of intertextuality, highlighting how user behavior can shape textual outcomes dynamically.

Intertextuality Concept

• The speaker emphasizes the idea of intertextuality, suggesting that reader reactions to text can influence subsequent content outcomes.

• Contrasts high-tech frameworks with service-oriented industry structures, foreseeing a challenging transition but potential positive outcomes.

• Explores the impact of new technologies on linguistic expertise balance, questioning automation's role in creative language tasks.

Evolution of Translation Challenges and Expertise

This segment delves into the shifting landscape of translation tasks due to technological advancements and highlights the importance of subject domain expertise and linguistic finesse.

Technological Impact on Translation

• Discusses automation's influence on creative language tasks versus linguistic experts' roles in translation processes.

• Emphasizes the value of literary finesse over subject domain expertise in translation tasks for superior language crafting.

• Examines how improving baselines affect performers differently, stressing the increasing significance of top performers in translation endeavors.

Future Trends in Translation Decision-Making

This part explores future trends in translation decision-making processes as technology evolves, emphasizing dynamic changes and increased weight on decisions made by skilled individuals.

Decision-Making Evolution

• Foresees a shift towards more dynamic decision-making processes in translations with far-reaching consequences.

• Predicts a transition to decisions impacting multiple pages dynamically, necessitating expert knowledge for effective outcomes.

Living Experience and Relatability

In this section, the importance of real-life experiences in improving skills and the significance of relatability in content creation are discussed.

The Value of Living Experience

• Living experiences are crucial for skill enhancement.

• Deep understanding requires linguistic, situational, and cultural awareness.

Relatability in Content Creation

• University provides space for profound absorption into writing and text analysis.

• Emphasis on literary analysis, semiotics, and developing a deeper appreciation of meaning production.

Machine Translation Quality Estimation

This segment delves into machine translation quality estimation (MTQE), its challenges, and the shift towards qualitative assessment.

Challenges in MTQE

• Difficulty in quantifying translation quality accurately.

• Transition from quantitative to qualitative assessment for better results.

Qualitative Approach

• Introduction of "translation smells" for qualitative evaluation.

Human Integration in Translation Process

Discusses the integration of human experience with text translation to enhance productivity without solely relying on MTQE.

Human-Centric Translation Process

• Advocacy for integrating human experience with text translation.

• Investment needed to develop solutions that benefit from human intervention efficiently.

Software Development Impact

Explores how software development is influenced more by vision and clarity than by financial resources or team size.

Software Development Dynamics

• Scale impact on sales versus software development dynamics.

New Section

In this section, the speaker discusses the approach to funding and growth in comparison to private equity or venture capital expectations.

Funding Approach and Growth Strategy

• The speaker emphasizes the difference in their approach by focusing on creating a great product with a great experience rather than solely seeking exponential returns. -

• They highlight that their company is already profitable, allowing them to prioritize product quality over chasing additional revenue. -

• The discussion shifts towards avoiding unnecessary complexities such as connectors, favoring simplicity and efficiency in operations. -

New Section

This part delves into recent developments and upcoming plans for the year 2024.

Recent Developments and Future Plans

• A new feature called learning terms has been introduced, enhancing the platform's ability to infer terminology during runtime for improved efficiency. -

• The focus remains on simplification across project creation and management processes to streamline operations further. -