SlatorPod

ICON plc’s Senior Vice President of Commercialisations & Outcomes, Brian Mallon, joins SlatorPod to shed light on the crucial role of language services within clinical research organizations (CROs). Brian recounts ICON's growth journey, from a team of five in Dublin back in 1990 to its present-day workforce of over 41,000 across 50 countries. Brian outlines the clinical cycle for drugs or medical treatments, which consists of several phases aimed at evaluating safety, efficacy, dosage, and potential side effects before making them available to the public.Brian reveals how language services play a vital role in improving clinical trials by enhancing the quality of data, reaching a more diverse patient population, and increasing retention and access to trials. He shares how medical translation and linguistic validation in clinical research require expertise in medical terminology and cultural nuances of the target language.Brian talks about how ICON led an industry-wide shift towards insourcing language services instead of relying on external language service providers. Key motivators include cost savings, increased efficiency, faster turnaround times, and the importance of data security and confidentiality.Brian explains how the cost of failure in translations is high, as errors can impact patient safety, regulatory compliance, and the overall success of the research. To manage these risks, ICON’s Language Services team implements stringent quality assurance processes, audits, and ongoing training.When it comes to the EU Clinical Trial Regulation and Clinical Trial Information System, Brian believes it has not had a major impact on language services in terms of volumes. However, the expectation is that over time, these regulations will improve the efficiency of conducting trials in Europe and make the EU more attractive for large-scale clinical trials.The pod rounds off with Brian’s view on the role of AI in the life sciences market and ICON’s  "string of pearls" approach when it comes to broadening its capability in line with industry trends in the life sciences sector.

Florian and Esther discuss the language industry news of the week, covering key takeaways from SlatorCon Remote such as lip sync AI dubbing demos from NeuralGarage and how FIFA and DeepL tackle sports localization. After unveiling its first-half results for 2023, RWS CEO Ian El-Mokadem emphasized the advantages of utilizing AI and large language models (LLMs) to enhance existing products. In M&A, Unbabel has completed its second acquisition of 2023 by acquiring Bablic, to expand its website and marketing localization solutions.AI video generation platform, Synthesia, has secured USD 90m in funding at a billion-dollar valuation for its multilingual avatars. One other unicorn, Cohere, has raised USD 270m in a Series C funding round, bringing the total raised amount to USD 445m. The duo do a SWOT analysis of what LLM capabilities mean for the language industry, specifically for LSPs, TMS providers, machine translation providers, and linguists. 

Farbod Mansorian, CEO of Unilingo, discusses launching the leading dubbing provider for Youtube creators, connecting with popular creators at YouTube conferences, challenges faced by voice actors, pros and cons of YouTube's multitrack audio feature, and the importance of grabbing viewers' attention in the first 10-20 seconds of a video.

Nic McMahon, CEO of United Language Group (ULG), joins SlatorPod to talk about the language service provider’s (LSP) mission in connecting language and culture to create meaningful outcomes.Nic recounts his journey in the language industry, from junior project manager at SDL to ascending to the role of CEO of ULG. He highlights ULG’s goal to go beyond translation services and prioritizing outcomes, cultural connectivity, and bridging gaps for better accessibility and engagement.The CEO explains that one specific area driving demand for ULG is the welfare cycle of non-English speaking communities, where the LSP aims to ensure access, equality, transparency, safety, and privacy in their language services. When working with municipalities and private sector healthcare providers in the US, Nic emphasizes the importance of understanding cultural influences within the target audience's network. Nic shares that HIPAA is a significant factor in the healthcare industry, particularly in relation to translation services, and is intertwined with transparency and patient safety.Nic discusses how ULG uses workflow automation, translation memory, and call management technology to help remove manual processes and improves the efficiency of translation and interpreting cycles. AI and machine translation have seen rapid adoption in recent years, with AI-driven neural machine translation (NMT) being used in 60% of projects at ULG.The pod rounds off with ULG’s outlook, where they are particularly interested in leveraging AI, such as optical character recognition, language recognition, and NMT to enhance outcomes and engagement.

This week Diego Bartolome, Founder of sintetic.ai, joined us on SlatoPod to talk about empowering companies to achieve more with generative artificial intelligence, including Slator with Slator Answers.Diego discusses how he collaborated with Slator to create a chat interface that allows Slator subscribers to question an AI model specifically trained on our research, podcast transcripts, and thousands of articles. Florian gives a brief demonstration of Slator Answers’ natural language querying abilities and puts the model on the spot by asking a question in German.Esther takes a look at a graphic from Slator’s 2023 Language Industry Report on company names in the language industry and reveals top terms, trends, and themes. The article concluded that the top three terms are language, translations, and services.Florian touches on Google's recent focus on large language models (LLMs) and their efforts to compete with OpenAI. In a research paper, Google claims that the PaLM 2 model outperforms their own Google Translate in machine translation across all locales.Google's chatbot, Bard, was released in February and is expanding to 180 countries, aiming to support 40 languages. When testing Bard, the LLM declined to translate from Japanese to English, stating that it is still learning. Additionally, Bard provided slightly different responses when asked about its abilities, such as generating subtitles.K-pop star, Lee Hyun, used YouTube’s early access multi-track audio feature to release his song in six languages simultaneously. The singer performed in all six languages, with AI used to correct pronunciation and ensure a smooth and accent-free result.US-based media localization company, Visual Data, has acquired Eva Localisation, a France-based dubbing and accessibility provider. The acquisition expands Visual Data's global presence, as Eva brings dubbing studios in France, Germany, and Spain, along with expertise in audiobooks and podcasts.

Joining SlatorPod this week is Steven Piantadosi, Associate Professor of Psychology at UC Berkeley. Steven also runs the computation and language lab (colala) at UC Berkeley, which studies the basic computational processes involved in human language and cognition.Steven talks about the emergence of large language models (LLMs) and how it has reshaped our understanding of language processing and language acquisition.Steven breaks down his March 2023 paper, "Modern language models refute Chomsky’s approach to language”. He argues that LLMs demonstrate a wide range of powerful language abilities and disprove foundational assumptions underpinning Noam Chomsky's theories and, as a consequence, negate parts of modern Linguistics.Steven shares how he prompted ChatGPT to generate coherent and sensible responses that go beyond its training data, showcasing its ability to produce creative outputs. While critics argue that it is merely an endless sequence of predicting the next token, Steven explains how the process allows the models to discover insights about language and potentially the world itself.Steven acknowledges that LLMs operate differently from humans, as models excel at language generation but lack certain human modes of reasoning when it comes to complex questions or scenarios. He unpacks the BabyLM Challenge which explores whether models can be trained on human-sized amounts of data and still learn syntax or other linguistic aspects effectively.Despite industry advancements and the trillion-dollar market opportunity, Steven agrees with Chomsky's ethical concerns, including issues such as the presence of harmful content, misinformation, and the potential impact on job displacement.Steven remains enthusiastic about the potential of LLMs and believes the recent advancements are a step forward to achieving artificial general intelligence, but refrains from making any concrete predictions.

Florian and Esther discuss the launch of Slator’s flagship 2023 Language Industry Market Report. The 140-page report provides a comprehensive view of the global language services and language technology industry, which Slator estimates grew over 4% to nearly USD 28bn in 2022 on the back of a strong first half.The duo talk about OpenAI’s new endeavor to enforce its branding guidelines to prevent companies from using GPT in their names or products. OpenAI believes that using GPT in branding confuses end-users, as it may imply a partnership or endorsement where there isn't one.Esther examines Appen’s financial update, where they reported a decline in revenue and gross profit in the first third of 2023. She expects that the data-for-AI provider will continue to face headwinds from the broader technology market slowdown, impacting revenues for FY23.In a roundup of Swiss-centered NLP news, Florian discussed how researchers at the University of Zurich have created SwissBERT, a pre-trained language model specifically for processing Switzerland-related text. The model was trained on over 21 million Swiss news articles in German, French, Italian, and Romansh, and outperformed previous models on natural language understanding tasks related to Switzerland.The Swiss Parliament has rejected a proposal to allow Swiss-German to be used in official federal political debates alongside the country's four official languages. Opponents argued that it would pose translation challenges and lead to gaps between what is spoken and written.Meanwhile, Zurich-based machine translation (MT) company Textshuttle has launched a free MT service for the general public covering all four national languages, including Swiss-German.

Florian and Esther discuss the language industry news of the week, covering the latest in large language models with Custom.MT’s Chat GPT in Localization Part II, SlatorPod’s The Great ChatGPT and Translation Debate, and Hugging Face’s launch of HuggingChat.Belgium-based language service provider (LSP) Jonckers sells a majority stake to investment firm Mayfair Equity Partners, following strong organic growth of 75% in 2022. The sale will enable Jonckers to continue its growth trajectory, with plans to invest in technology, expand into new markets and pursue bolt-on M&A opportunities.Super Agencies Welocalize and Lionbridge have announced new C-level appointments and partnerships. Paul Carr has been named as the new CEO of Welocalize, succeeding co-founder Smith Yewell, who has held the role since 1997. Menaka Thillaiampalam has been appointed as the new CMO of Lionbridge, bringing over 20 years of experience in technology firms to the role. Additionally, Lionbridge has signed a multi-year contract with Phrase, a translation management system company, to integrate its computer-assisted translation tool into the LSP's workflow.Over in the UK, RWS released a trading update for the first half of the financial year 2023. While revenues grew by 2.5%, sales declined by 6.8%. The market responded with shares in RWS plunging over 16% in one day. Media localization provider, ZOO Digital, has raised USD 15.5m in a share placement, with the aim of using the proceeds to acquire a “media localization subsidiary of a leading Japanese technology company.” ZOO Digital's rationale behind the proposed acquisition was to deliver Japanese language services in-house to achieve better margins.

In this week’s SlatorPod, Fireflies.ai CEO Krish Ramineni joins us to talk about scaling the AI meeting assistant and building on the latest advances in large language models.Krish starts with his journey to co-founding Fireflies, which began as a drone delivery service and as a result of conversations with customers and investors, evolved into an AI meeting assistant to solve their own pain point.The CEO shares how they found their product-market fit after focusing on automated transcripts over human-assisted note-taking. He discusses the early days of AI investment and how with the rise of APIs and large language models (LLMs), you no longer need multiple PhDs to attract investors. Krish explains how Fireflies leverages technologies like Whisper to improve their language transcription, allowing them to be more accessible to global companies. He talks about their decision to improve their Super Summaries feature through GPT technology.The CEO shares his excitement about the potential for LLMs and how Fireflies are building a Chrome extension that uses LLMs to summarize any article or video on the internet. He advises that simply building a wrapper on top of OpenAI is not a defensible moat for companies, but rather you should build a unique platform with a unique angle into the industry you’re selling to.Kirsh talks about the current fundraising environment where there is a lot of money being thrown around for generative AI companies, but only a few will weather the storm. When it comes to hiring machine learning talent, Krish doesn't believe in prompt engineering and also holds the view that machine learning companies may no longer need to hire large cohorts of ML PhDs to scale.The pod rounds off with the company’s roadmap for 2023, which includes creating an ecosystem of extensions on top of Fireflies. These extensions will offer powerful functionalities to users in different sectors like healthcare and recruiting. 

Tony Abrahams, CEO and Co-founder of Ai-Media, joins SlatorPod to talk about the journey to building a market leader in multilingual live captioning.Tony discusses his transition from working in finance to co-founding Ai-Media with Alex Jones and introducing large-scale captioning to Australian Pay TV. He gives an overview of Ai-Media’s technology stack, which delivers high-quality automatic captioning through three key elements: encoding, the iCap network, and LEXI.The CEO talks about the use of respeaking versus LEXI in settings where captioning accuracy is critical, and where there are multiple speakers, mixed-quality audio, or background noise. He discusses how Ai-Media measures live-captioning quality using the NER model, which weights the types of errors as editing errors or recognition errors.Touching on the multilingual component of Ai-Media, Tony explores the possibility of using AI instead of respeakers and having a fully-automated translation product in the near future. He believes that large language models are an opportunity as the technology has enabled them to interpret sentences more accurately, resulting in a better outcome with LEXI 3.0. Tony gives his thoughts on growing through M&A and the strategy behind acquiring EEG to gain a competitive advantage in terms of its technology and product suite. He shares his rationale for taking AI-Media public.The CEO reveals Ai-Media’s roadmap for 2023, such as improving the iCap network and launching the LEXI Library, which allows customers to search their media library by captions.