SlatorPod

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.

This week, SlatorPod hosts its very first panel debate with guests Adam Bittlingmayer, CEO of ModelFront, Varshul Gupta, Co-founder of Dubverse, and Mihai Vlad, General Manager of Language Weaver. To start off, the panel participants reflect on their recent experience with ChatGPT since its launch in November 2022 and how this shapes their views on large language models (LLMs). Varshul and Adam talk about how clients view ChatGPT.Mihai agrees with the idea that the language services industry is exceptionally well-prepared for the launch of ChatGPT due to its experience with human-machine interaction. Varshul discusses how LLMs have influenced startups like Dubverse to build prototypes that can handle edge cases.Mihai shares the challenges of deploying LLMs in large enterprises. Adam and Varshul highlight how parameters such as security, data privacy, latency, throughput, and cost are essential to consider in an enterprise setting.Varshul and Mihai talk about the potential of multilingual content generation from scratch and how it will affect production costs. Varshul shares how they continue to attract users throughout this AI hype and the importance of adding a UX on top of LLMs.Adam discusses the potential for LLMs to assist translators in their work, although the implementation of this tech may take some time to become the new normal. Varshul and Mihai debate how services-focused companies should react to the rapid advancements in LLMs, whether you wait to see how things pan out or go all in to stay ahead of the curve.The panel rounds off with emerging use cases for LLMs, from building prompt-based systems for more concise translations to addressing long-tail languages that are often overlooked by machine learning due to the fragmentation of the language industry.

In this week’s SlatorPod, we are joined by Christian Federmann, Principal Research Manager at Microsoft, where he works on machine translation (MT) evaluation and language expansion.Christian recounts his journey from working at the German Research Center for Artificial Intelligence under the guidance of AI pioneer Hans Uszkoreit to joining Microsoft and building out Microsoft Translator.He shares how Microsoft Translator evolved from using statistical MT to neural MT and why they opted for the Marian framework.Christian expands on Microsoft’s push into large language models (LLMs) and how his team is now experimenting with NMT and LLM machine translation systems. He then explores how LLMs translate and the role of various prompts in the process.Christian discusses the key metrics historically and currently used to evaluate machine translation. He also unpacks the findings from a recent research paper he co-authored investigating the applicability of LLMs for automated assessment of translation quality.Christian describes how Microsoft’s custom translator fine-tunes and improves the user’s MT model through customer-specific data, which degrades more general domain performance. He shares Microsoft’s approach to expanding its support for languages with the recent addition of 13 African languages. Collaboration with language communities is an integral step in improving the quality of the translation modelsTo round off, Christian believes that the hype around LLMs may hit a wall within the next six months, as people realize the limitations of what they can achieve. However, in a year or two, there will be better solutions available, including LLM-enhanced machine translation.

In this week’s SlatorPod, we are joined by Nikola Nikolov, an experienced researcher, engineer, YouTuber, and consultant in natural language processing (NLP) and machine learning.Nikola talks about the evolution of large language models (LLMs), where the core technology remains the same, but the number of parameters has grown exponentially and the capacity to fine-tune models on human data via reinforcement learning from human feedback has turbocharged the models’ capabilities.Nikola unpacks the rapid increase in front-end use cases with companies like Google and Microsoft already integrating LLMs into their products. At the same time, he speculates about what will happen to the hundreds of startups that are using APIs to build similar tools like writing assistance or summarization.Nikola shares the limitations of an API-only approach, which include using a model limited in data it has collected from the internet and that is not fine-tuned to a domain or specific use case. He discusses how LLMs perform when it comes to machine translation (MT). Although GPT is trained on large amounts of multilingual data, it’s not specialized in translation, so machine translation providers will retain their edge over ChatGPT for now.Nikola predicts two different scenarios when it comes to the future of LLMs: the first is where large corporations quickly integrate LLMs into their products, competing with startups and putting many of them out of business. The second scenario is where startups will create novel use cases and integrate multimodal technology to build something completely new and different from big companies.

Florian and Esther are back for a packed news pod. The two catch up with OpenAI announcing the release of its much anticipated GPT-4 large language model. They unpack OpenAI’s research on GPT or GPT-powered systems’ impact on the labor force, including specifics around how exposed translators and interpreters are to large language models.Over at the University of Edinburgh, researchers looked at hallucinations in large multilingual translation models. The research looked at hallucinations across models of different scale, translation directions, and data conditions.Lilt announced the launch of Contextual AI Engine, a new GPT-style model which Lilt claims has higher accuracy than Google Translate and GPT-4 for many enterprise contexts. Esther talks about Acolad appointing Bertrand Gstalder as the Super Agency’s new CEO.At the Google for Games Developer Summit 2023, the Tech Giant announced that developers now have access to free machine translation (MT) for their Android apps. Keynote Speaker, Greg Hartrell, also revealed an early access program for machine translation in the Play Console.Meanwhile, Disney lawyers have filed a subpoena request to force Reddit to identify the person or people responsible for leaking a movie dialogue transcript. It was revealed that a Reddit community shared a Google Doc transcript in Portuguese of the film “Ant-Man and the Wasp: Quantumania” weeks before it was released.