# 178 LSP Performance Check, Llama 2, Phrase, Language I/O, LanguageWire

AI Solutions Replacing LSP Contracts

Introduction to the Discussion

• The podcast opens with a discussion on the increasing trend of AI solutions replacing Language Service Provider (LSP) contracts, suggesting this may become more common in the future.

• Hosts Esther and Florian mention their excitement about SlatorCon Zurich, scheduled for October 4th at Sheraton Zurich West, highlighting its unique high-rise venue.

SlatorCon Zurich Details

• A list of partners for the conference is provided, including We Localize, RWS, LanguageWire, Tarjama, Supertext, DeepL, Phrase, and Language Inspired.

• There are still a few slots available for exhibitors; interested parties are encouraged to contact Andrew as tickets are expected to sell out soon.

AI Replacing Traditional Translation Services

Case Study: Baltimore's Emergency Services

• A notable case is discussed where Baltimore police replaced part of their contract with LanguageLine with NWN Carousel for AI-enabled multilingual emergency services.

• The contract is valued at $6.5 million and aims to provide automated text translation in over 140 languages for emergency calls.

Impact on Emergency Response

• The new system claims to reduce emergency response times by 90%, transitioning from traditional live interpreters to an AI-driven model.

• This instance marks a significant shift in how LSP contracts can be supplanted by technology solutions; however, it raises questions about the reliability of such systems in critical situations.

Large Language Models (LLMs): Opportunities and Challenges

Open Source vs. Proprietary Models

• Discussion shifts towards large language models like Meta's Llama series which have been open-sourced; Llama 2 was released in late July 2023.

• Yann LeCun from Facebook describes Llama's launch as significant due to its open-source nature allowing free use for research and commercial purposes.

Limitations of Current Models

• Despite advancements, there’s a warning regarding Llama’s training corpus being predominantly English (over 90%), potentially limiting effectiveness in other languages.

Survey Insights and Financial Results Overview

ALC Survey Presentation

• The team has completed the ALC survey, which will be presented at the ALC conference in Portland from September 20th to 22nd.

• Esther will deliver an exclusive presentation on the survey findings during this event.

Financial Statements of Language Service Providers (LSPs)

• Recent financial results from listed LSPs have been reviewed, highlighting trends and market reactions.

Straker's Q1 Financial Performance

• Straker reported a 2% revenue growth to 13 million New Zealand dollars for Q1 of FY24, with annual revenues last year around 37 million US dollars.

• Commentary indicated that IBM, their major client, saw over 20% revenue growth in Q1 but remained below peak levels from the previous financial year.

• Europe emerged as Straker's fastest-growing region with a 25% increase linked to strong demand from institutional clients; however, North America faced challenges.

Market Valuation Concerns

• Straker's shares are near all-time lows with a market cap around 26 million dollars despite annual revenues of about 37 million US dollars.

• The company needs to achieve solid profitability consistently for stock value improvement; current metrics used for profitability are considered technical and infrequent.

ZOO Digital's Strong Year Amid Challenges

• ZOO Digital released its full-year results for March 2023, showing a remarkable revenue increase of 28% to 90 million US dollars and more than doubling adjusted EBITDA to over 15 million US dollars.

Media Localization Growth

• Within media localization services, subtitling grew by 15%, while dubbing surged by an impressive 73%, attributed partly to lower comparative figures from the pandemic period.

Current Industry Challenges

• Despite past successes, ZOO faces disruptions due to Hollywood strikes affecting production volumes and cost-saving measures by major content owners.

Strategies for Content Delivery

Understanding the Concerns of Actors and Writers in Hollywood

Key Issues Surrounding Compensation and AI

• The discussion highlights two primary concerns for striking actors and writers: compensation issues and the implications of AI technology. Both aspects are interconnected, particularly regarding how individuals are compensated when AI is involved.

• Currently, an actor's likeness or voice can be reused without their consent or compensation. This raises significant ethical questions about ownership and rights in the context of advancing AI dubbing technologies.

• As voice cloning technology improves, there is a potential future where actors' voices could be cloned to dub shows in multiple languages without their direct involvement, leading to calls for fair compensation.

• The conversation touches on how this technology could eliminate traditional dubbing artists from the process, as actors might perform their own dubbing through AI-generated clones instead.

Rights and Ownership Concerns for Writers

• Writers are also concerned about rights and ownership related to content used for training large language models (LLMs). If a script is utilized to train an AI that generates new content, questions arise regarding legal ownership of that generated material.

• The ongoing strikes have led to production halts in Hollywood, significantly impacting companies like ZOO Media Localization, which has seen a drastic drop in stock value due to these uncertainties.

Market Impact and Company Performance

• Despite challenges faced by media localization businesses due to strikes, some companies remain profitable. For instance, ZOO's shares dropped significantly but may recover if resolutions are reached quickly.

• Keywords Studios reported strong first-half revenues with organic growth despite industry challenges. Their localization services still contribute substantially to overall revenue despite shifts away from certain business areas.

Company Revenue Stability and Market Movements

Overview of Company Performance

• The speaker reflects on a company known for 15 years, noting its stagnant revenue level around $80 million since 2008.

• In the first quarter of 2024, the company reported revenues of approximately 2.6 billion Japanese yen (around $17.7 million), marking a modest year-on-year growth of about 4%.

Core Business Insights

• Translation services constitute over 75% of total revenues, with patent translations showing significant growth at 14%.

• While some business segments are experiencing softer performance, overall growth remains modest at around 4%.

Mergers and Acquisitions in Language Services

ElaN Languages Expansion

• ElaN Languages acquired Lingua-World to expand geographically into the German market, leveraging synergies in translation and interpreting services.

• Backed by French private equity firm Andera Partners since April 2022, ElaN aims to accelerate growth in France and Europe.

Financial Projections

• Lingua-World is projected to generate consolidated revenues of approximately $48 million for 2023, up from just over $22 million in 2022.

LanguageWire's Strategic Moves

Nordic Market Consolidation

• LanguageWire's acquisition of Delingua strengthens its position in the Nordic region, focusing on both private and public sector contracts.

Revenue Expectations

• LanguageWire anticipates combined revenues reaching around $88 million for 2023 following recent acquisitions.

Funding Developments in the Language Industry

Phrase's Financing Strategy

• Phrase secured €12 million in venture debt from CIBC Innovation Banking to further develop its Phrase Localization Suite.

Product Development Focus

• Phrase Translate serves as a machine translation aggregator connecting various MT engines; it aims to enhance workflow management tools like Phrase Orchestrator.

Technological Advancements and Future Prospects

Leveraging Machine Learning Models

• Sophisticated TMS platforms are well-positioned to utilize large language models (LLMs), enhancing their service offerings through advanced integration capabilities.

Emerging Trends

Using Machine Translation for Customer Support

Overview of Language I/O's Series A1 Funding

• Language I/O is utilizing machine translation (MT) to assist monolingual customer support agents, primarily English speakers, in communicating with customers.

• The funding round is referred to as Series A1, which is a new trend; it mirrors the size of previous Series A rounds at approximately $8 million.

• This approach serves as a bridge in funding, contrasting with traditional expectations of a larger Series B round, which would typically be around $30 million.

• In mid-2023, securing Series B funding has become increasingly challenging for tech companies.