UAE unveils locally developed AI large language model dedicated to climate intelligence

Jais Climate, the world’s first bilingual large language model dedicated to climate intelligence, has been unveiled at the Mohamed Bin Zayed University of Artificial Intelligence in Abu Dhabi, just days before the Cop28 conference in Dubai.

The large language model (LLM) which is focused on climate, sustainability and Cop28, was developed through a partnership between the university and Core42, a subsidiary of Abu Dhabi-based technology holding company G42.

Jais Climate includes 1.4 million climate-related instructions. It was trained on ClimaInstruct, the largest instruction-based bilingual data set on climate and sustainability-related topics, according to Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI) and Core42.

LLMs have fuelled the growth and interest in artificial intelligence, producing content by training on large amounts of data, which in turn, enables users to get answers and automate tasks in a fraction of the time previously needed.

Jais Climate offers a way to learn more about climate-related issues before and during Cop28, said Andrew Jackson, Core 42’s executive vice president and chief AI officer.

The climate conference will take place from 30 November to 12 December in Dubai.

Mr Jackson said his children had tested Jais Climate on related school work. “They’re my base users,” he said.

“They were able to ask climate questions quickly and get answers back, they truly had a great experience.”

Some of the initial criticisms of LLMs have revolved around the languages they’re based on, with some critics claiming the focus on English potentially skews answers and results.

The launch in August of Jais, an open-source bilingual Arabic-English model, addressed those concerns by bringing Arabic into the AI mainstream, with its developers touting it as more accurate than other Arabic LLMs.

Jais Climate, like Jais, was created using sources and climate information in English and Arabic.

It will make climate data easily accessible to more than 274 million Arabic and 1.4 billion English speakers worldwide, according to MBZUAI and Core42.