Wales partners with OpenAI to improve Welsh language tech

The Welsh government has signed a partnership deal with OpenAI as part of a plan to increase Welsh speakers in the country.

The government believes that artificial intelligence (AI) technologies can help meet its target of one million Welsh speakers and increasing use of the language by 2050.

The deal with OpenAI, a company probably best known for creating ChatGPT, will see it build open data archives to contribute data to the research community, and thereby improve the linguistics performance of AI models and apps.

Cabinet secretary for economy, energy and Welsh language Jeremy Miles said the partnership with OpenAI will help the government “ensure we’re on the right path” to meet the 2025 target.

“All of us use technology one way or another, and increasingly, we’re seeing AI being used in more situations,” he said. “I’m excited to see how the new data partnership with OpenAI will lead to improving how AI-technologies work in Welsh. Working with OpenAI in the past, we’ve been able to share Welsh-language resources and Welsh government-funded components.”

The government will publish a Welsh Language Education Bill in the next few weeks, aiming to help school pupils become “independent and confident Welsh speakers”, with the use of technology being a key part of the bill.

It will also use technology with the aim of increasing language transmission both within families and communities.

In 2018, the Welsh government published its Welsh-language technology action plan, focused on technological developments to “ensure that the Welsh language can be used in a wide variety of contexts, be that by using voice, keyboard or other means of human-computer interaction”.

“We must ensure that we have the necessary Welsh-language digital infrastructure we need, such as speech-to-text, machine translation, relevant corpora and big datasets, and machine learning,” the strategy said. “We must have Welsh-language data to train these systems. We need to make sure that academic research and development in the field of Welsh-language technology receives long-term support.”

The plan has so far funded, created and started work on many of the digital components to increase Welsh language use, however, Miles said there is still more to do.

“Today, I’m inviting people to submit information or ideas about what technological developments would help them use more Welsh, and what needs to happen to make it easier for them to use technology in Welsh,” he said.

The ideas can be submitted on the Welsh government’s website, and the government will open communication with people across Wales and beyond to get further input into future Welsh-language technology work.

OpenAI has already been involved with the Welsh-language technology action plan, but this new deal aims to take the partnership further. Anna Makanju, vice-president of global affairs at OpenAI, said the Welsh government has been a “great partner in creating an open-source dataset for training language models”.

“At OpenAI, we want our models to understand as many languages and cultures as possible, so they can benefit as many people as possible,” she said.