Middle East Leaders Bow Out of AI Race as U.S. and China Take the Lead

(Vmeetsolutions News) — Just months after ChatGPT was launched at the end of 2022, research laboratories in the United Arab Emirates announced they had created formidable competitors. Leading this pack was Falcon, an extensively used open-source AI system backed by the government, along with Jais, which is named after the nation’s tallest summit.

“The future of AI is not a far-off vision, but a current truth,” stated Peng Xiao, CEO of the Emirati technology giant G42.
said
In a statement following the company’s launch of Jais in 2023,

However, the UAE’s aspiration for locally developed, competitive AI models is still distant. Falcon trails considerably behind prominent offerings from American firms when it comes to user counts and public ratings. On another front, G42 has redirected its efforts away from Jais; instead, they are concentrating their resources on developing specialized functionalities atop pre-existing AI platforms created by external entities such as OpenAI.

In the beginning, we were unsure about the potential of foundation models and what was needed to advance them,” Xiao shared with Vmeetsolutions News in April. “Given our country’s size, it wouldn’t be practical for us to take on such an endeavor alone.

After over two years of the generative AI craze, the worldwide contest to create more advanced models seems to be turning into a rivalry between two nations. Several U.S. companies still dominate in AI advancement by investing massive amounts in hardware, data facilities, and skilled personnel to construct superior and larger systems. In contrast, China is quickly closing the gap and making significant strides forward.
flooding the market
By utilizing inexpensive, open-source models to broaden its scope. Many other countries, despite being affluent such as the UAE, find themselves positioned somewhere in between.

An increasing number of formerly promising AI enterprises in the Middle East and Europe have either faltered or nearly surrendered. Germany’s Aleph Alpha, which was once touted as Europe’s counterpart to OpenAI, has faced similar challenges.
made a similar decision
To G42 last year. Britain’s Stability AI, one of the early pioneers in AI models, has
petered out
Following management problems, even firms such as France’s Mistral—which secured considerable venture capital and received support from the nation’s government—has demonstrated
little evidence
with significant commercial appeal or developer engagement.

Across the Middle East, just like in various other regions, businesses are reassessing the value proposition of developing advanced AI models from the ground up. According to Kiril Evtimov, who heads the cloud division known as Core42 at G42, Jais has the potential to become a leading “frontier model” should G42 continue investing in it. However, he questions whether this approach represents the most effective business strategy for capturing the market, suggesting that the likely response would be negative.

In 2023, the UAE introduced a new company called AI71, which was promoted as a commercial entity for Falcon—a model created by a governmental research organization. According to a representative, AI71 ultimately joined G42 and Aleph Alpha in developing specialized AI tools for various businesses, utilizing several models such as Falcon.

Although Falcon continues to be the standout product from the UAE, it has found itself lagging behind innovations from open-source rivals at Meta Platforms Inc. and China’s DeepSeek. In 2023, TII, which developed Falcon,
touted
The AI system achieved first place among open-source models listed on Hugging Face, which serves as an important indicator for the sector. Up until last week, Falcon was not even within the top 500 positions on the site’s leaderboards.

A spokesperson for TII mentioned that Falcon currently boasts over 55 million downloads. This figure represents only a portion of the total downloads for Meta’s suite of Llama models.
over 1 billion times
The representative mentioned that integrations with cybersecurity firms, robotic enterprises, and cloud corporations will be unveiled shortly. They stated, “A model’s worth isn’t solely determined by its preliminary test scores; rather, it hinges on its capacity to foster and drive wider-ranging innovations as time progresses,” according to an official statement.

G42, meantime, has explored alternative methods to capitalize on the present AI surge aside from developing models. Their data center operations are also part of this strategy.
growing in the Gulf area
And MGX, an investment fund formed jointly by the company, has
backed US AI
developers OpenAI and xAI.

Several nations continue to compete with the US and China in advancing artificial intelligence technology. Earlier this year, in January, the Indian government
said it would
Support up to 18 distinct initiatives aimed at developing core AI models. A minister promised these efforts would “stand alongside the world’s finest.” Additionally, Saudi Arabia’s AI authority joined forces with another entity.
International Business Machines Corp.
Last year, they introduced their domestically developed national model, ALLaM, through the technology firm’s cloud services.

Nations pursue these so-called sovereign models to have more control over how the AI systems are trained, either to influence how they work or lessen dependence on foreign tech, Michael Bronstein, a professor of AI at the University of Oxford, said at the Machines Can See conference in Dubai. But he said the prevalence of open-source models means these approaches have little odds of remaining competitive.

Several of these models were proposed as methods to more effectively represent languages and demographics not adequately addressed by tech giants like those in Silicon Valley. Most of the training undertaken by organizations such as OpenAI and similar entities was primarily conducted using data sourced mainly from the internet, where content is predominantly in English. However, according to Nour Al Hassan, who serves as the CEO at Tarjama—a translation service located in Dubai—an efficient solution tailored for Arabic does not necessitate building a comprehensive language model entirely anew.

A newly established startup called Arabic.AI has been nurtured through an incubator program at her company.
released an AI system
What Al Hassan mentioned was developed by refining various large-scale models, which she believes works well for cost-conscious business customers. “You don’t require enormous models to accomplish a particular job for a financial institution,” she stated.

Jais from G42 was developed with the capability to fuel chatbots in Arabic. Additionally, in September, G42 launched Nanda, a Hindi-language model, as part of their growth strategy within India. According to Andrew Jackson, who leads G42’s AI division called Inception, both models will be continually updated by the company along with its research collaborators.

But commercial success isn’t our main priority,” he stated. “We still need to generate revenue.

Leave a Comment Cancel reply