IBM to help Saudi Arabia train AI models in Arabic

SPA

• ALLaM model will use IBM platform
• Ability to understand Arabic dialects
• IBM to release open source models

Saudi Arabia has signed a deal with US tech giant IBM to develop the kingdom’s generative artificial intelligence (AI) capabilities.

The Saudi Data and Artificial Intelligence Authority (SDAIA) said on Tuesday that its Arabic language AI model, ALLaM, will be trained on IBM’s platform to run more efficiently.

This will also add new language skills to IBM’s offerings, including the ability to understand multiple dialects of Arabic.

The announcement was made at the Think Boston 2024 event.

---

---

IBM’s watsonx, a commercial tool that facilitates the training, deployment, validation and governance of AI models, will serve as the cornerstone for ALLaM’s training.

The partnership aims to achieve two critical objectives. 

Firstly, it seeks to broaden the availability and applicability of Arabic large language models (LLMs). An LLM is a type of generative AI programme that can recognise and generate text, among other tasks.

Secondly, the tie-up fosters responsible AI development by providing organisations with governance frameworks for using Arabic generative AI.

“This collaboration ensures that both public and private sectors can benefit from deploying Arabic language models, aligning with the cultural and linguistic needs of the region,” said Ayman AlRashed, regional vice president at IBM Saudi Arabia. 

• Aramco launches world’s first industrial AI model
• Pressure on Saudi tourism exposes mixed feelings about AI
• Core42 unveils advanced Arabic language AI model

IBM says it is committed to expanding its investments in Saudi Arabia. 

The New York-headquartered company announced an investment of $250 million to establish a global software lab in Saudi Arabia during the Leap conference earlier this year.