After our short summer break and the end of the ELE 2 project, we’re happy to be back with a new edition of our newsletter packing a variety of LT news with a focus on the latest developments in Europe. Read about European countries that have launched their own national language technology platforms, OpenAI’s road ahead in Europe, European LLM alternatives, and the latest in AI regulation.

In the context of the news updates, be sure to (re)visit the exciting presentations of this year’s META-FORUM, available in full length on our YouTube channel. 

We’re also taking a closer look at the STOA technical feasibility study for the development of a European streaming platform for European national news accessible in all EU languages in a dedicated section.

With best regards

The European Language Data Space initiative that was started in January 2023 recently launched its monthly newsletter, providing information on the latest developments in secure, privacy-preserving language data sharing and use across Europe. 

We’d like to invite you to subscribe to the newsletter for updates on LDS implementation, success stories, events, and more!

• “5 European Countries Have Already Launched Their National Language Technology Platforms” – Slator, 25 July 2023

• “Artificial intelligence and jobs: Evidence from Europe” – CEPR, 29 July 2023

• “Viewpoint: how Europe can catch the US on artificial intelligence” – Science|Business, 22 August 2023

• “Can Basque, Catalan and Galician really become EU languages?” – Euronews, 24 August 2023

• “Meet the guy responsible for helping Meta, Google and Amazon prepare for new laws in Europe” – CNBC, 25 August 2023

• “Protecting human rights and democracy in the era of AI” – Innovation News Network, 25 August 2023

• “Spain launches AI regulation agency in bid to become industry leader” – Cointelegraph, 25 August 2023

• “Top Language AI Researchers Propose New Way to Auto-Evaluate Machine Translation” – Slator, 25 August 2023

• “A consortium by SiloGen and TurkuNLP develops Europe's largest open-source LLMs” – Tech.eu, 30 August 2023

• “Meet Aleph Alpha, Europe’s Answer to OpenAI” – Wired, 30 August 2023

• “ChatGPT Isn’t Made For European Union” – Wired, 31 August 2023

• “UK government sets out AI Safety Summit ambitions” – gov.uk, 4 September 2023

• “DeepMind’s Path to Better Large Language Models Runs Through Machine Translation” – Slator, 5 September 2023

• “After Gizmodo Fires Spanish Writers, Google Now Ranks Raw Machine Translation” – Slator, 6 September 2023

• “G7 countries commit to AI code of conduct” – Politico, 7 September 2023

• AI vs. Human Writing: Experts Fooled Almost 62% of the Time” – Neuroscience News, 7 September 2023

• “Reddit users on mobile can now translate posts into other languages” – TechCrunch, 8 September 2023

• “OpenAI confirms that AI writing detectors don’t work” – Ars Technica, 8 September 2023

• “What are governments doing to try to regulate AI?” – Euronews, 11 September 2023

• “What to know about Congress’s inaugural AI meeting” – MIT Technology Review, 11 September 2023

• “More writers sue OpenAI for copyright infringement over AI training” – Yahoo Finance, 11 September 2023

• “YouTube Launches a Multilingual Health Video Creator Program Using Machine Dubbing” – Slator, 11 September 2023

• “The Man Keeping Musk, Zuckerberg and Big Tech in Line in Europe” – Yahoo Finance, 12 September 2023

• “Websites With Fact-Checked Content Should Block Scraping and build a European LLM” – DataEthics, 12 September 2023

• “What Jobs Are Most at Risk From Generative AI Automation [Charts]” – Voicebot.ai, 13 September 2023

• “EU countries to slow Spain´s new official languages bid” – Euractiv, 15 September 2023

• “Riding the AI tsunami: The next wave of generative intelligence” – VentureBeat, 17 September 2023

• “Catalans are not demanding favourable treatment, just linguistic equality” – Euractiv, 18 September 2023

• Read the European Commission’s Twitter thread on building a new global AI framework. 

• When the LLM hallucination corrects itself.

• The keynotes and discussions of META-FORUM 2023 are now up on our Youtube channel!

On 20 August, Silo AI, in collaboration with TurkuNLP at the University of Turku, announced the launch of a significant initiative to promote open and trustworthy Large Language Models. The goal is to ensure digital sovereignty in Europe and to make LLMs accessible to all. This initiative includes the development of the world's largest open-source language model, covering all official European languages.

To achieve this, the consortium will have access to approximately 15 million GPU hours and will ensure that data used in these models accurately represent European languages. The project will adhere to European regulations and collaborate closely with key European institutions and agencies. Additionally, this open-source initiative will democratise LLM access and enable use-case specific downstream applications.

The consortium has access to high-performance computing resources, including the LUMI supercomputer.

A team of social scientists working with global management consulting firm Boston Consulting Group published a working paper on Artificial Intelligence and its role in reshaping the future of work. Titled “Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality”, the paper details the results of the team examining the performance implications of AI on realistic, complex, and knowledge-intensive tasks. After establishing a performance baseline on a similar task, subjects were randomly assigned to one of three conditions: no AI access, GPT-4 AI access, or GPT-4 AI access with a prompt engineering overview. The study suggests that the capabilities of AI create a “jagged technological frontier” where some tasks are easily done by AI, while others, though seemingly similar in difficulty level, are outside the current capability of AI. 

Consultants using AI finished 12.2% more tasks on average, completed tasks 25.1% more quickly, and produced 40% higher quality results than those without.

In addition to the paper itself, one of the team members has documented the paper’s results in an easy-to-follow blog post.

In our previous newsletter edition, we briefly introduced the technical feasibility study for the development of a European streaming platform for European national news accessible in all EU languages that was published by the European Parliament’s Panel for the Future of Science and Technology. In this section, we want to take a closer look at it.

The study, prepared and conducted by a team of 21 researchers from six different research labs of the German Research Centre for Artificial Intelligence (DFKI), envisions a digital public space where EU citizens can easily access national news in their own languages, eliminating language barriers and providing access to reliable news content from across the EU. 

To make this vision a reality, support is needed from high-quality European broadcasters, including public broadcasters, media companies, non-governmental organizations, and various stakeholders, including citizens. These entities will collaborate to share broadcasts, media-related content, and data, as well as to develop innovative solutions, thus creating a multilingual focal point that bundles together all high-quality European news broadcasts in one platform. It would enable EU citizens to discover and consume news in multiple languages seamlessly, thanks to advanced AI-based language technologies. These technologies include automatic speech recognition for transcription and indexing, machine translation for subtitles, and even automatic speech-to-speech translation for fully automated dubbed programs.

Ensuring the platform's resilience against external manipulation is of utmost importance. The platform would leverage a diverse set of artificial intelligence technologies to process audiovisual content and provide fair, balanced search, discovery, and recommendation mechanisms. This endeavor represents a significant step forward, enabling communication across language barriers at a continental scale.

Funding opportunity for research organization and SMEs – Development and Application of Deep Models for eXtended Reality:

HE-UTTER FINANCIAL SUPPORT FOR THIRD PARTIES (FSTP) CALL

The Horizon Europe project, UTTER, invites project proposals from research organizations and SMEs to develop and/or pilot applications of large pretrained language models with a focus on enabling human-human and human-machine interaction. Successful applications will receive up to € 60 000 each, and run for 6-9 months. The call closes on October 15, 2023. 

Below is an exhaustive compilation of essential documentation pertaining to the Submission & Evaluation process:

• FSTP- Key Facts

• UTTER- Call documentation

• UTTER- Grant Agreement

• UTTER- Consortium Agreement

• A1- guide for applicants

• A2- third party agreemetns

• A3- project proposal template

• A4- Evaluation criteria

Proposals are to be submitted via UTTER’s proposal management portal at: https://utter-fstp.science.uva.nl

UPSKILLS Teaching and Learning Content – This month’s resource is a collection of modular teaching and learning content created in the UPSKILLS project ( UPgrading the SKIlls of Linguistics and Language Students) and downloaded from the Moodle platform in .mbz format. The learning content can be reused and adapted by curriculum designers, lecturers, and instructors of courses in linguistics and language-related subjects. Different blocks or individual units within a block can be combined to create new learning paths at the BA and MA levels. Some of the learning content is also suitable for the PhD level. Students can also use the content for self-study, considering this is not a MOOC (Massive Open Online Course). 

• European Big Data Value Forum - Data and AI in Action: Sustainable Impact and Future Realities, 25 - 27 October, Valencia, Spain

• NGI Forum 2023: Unlocking the power of Digital Commons, 15 - 16 November, Brussels, Belgium

If you have an event that you think the European language technology community should know about, get in touch with us to have it featured in this newsletter.
 

The next ELT newsletter will be sent out on 24 October 2023. Until then, follow our ELT social media accounts (as linked below) for the latest news!