Artificial intelligence (AI) is expected to generate most online content in the near future, and without deliberate efforts to incorporate Middle Eastern cultures into AI models, those cultures risk fading from the digital landscape, an industry expert has warned. In an interview with 'Gulf Times', Amit Jain, co-founder and CEO of Luma AI, said cultures that are not represented in AI systems today may be erased from tomorrow’s digital record. “If we combine three things, our identity living online, AI generating most of that content, and models trained only on existing internet data, then cultures that are underrepresented risk vanishing,” Jain said. “The alternative is that it simply disappears from the internet.”
Luma AI, a Palo Alto-based foundation lab, recently raised $1bn in Series C funding led by HUMAIN and announced plans to build a 2-gigawatt AI supercluster, Project Halo, to train next-generation “world models”. These multimodal systems are designed to understand video, audio and language simultaneously, not just text.
“Text is humanity’s interpretation of the universe. Video and audio are records of the physical world. When you combine them, you start building intelligence that understands how the universe functions — and how humans interact with it,” Jain explained.
Unlike companies focused solely on text or image models, Luma is building systems capable of reasoning across formats. “We don’t train image models or video models,” he said. “We train models that are intelligent.”
The implications stretch far beyond advertising and entertainment, although Luma’s technology has already been used to produce Saudi Arabia’s National Day broadcast film with a team of just three people.
“It would have taken six to eight months and up to a million dollars. Three people did it in about a month. It aired on national television,” Jain said.
Yet for Jain, the bigger issue is representation: “Right now, our models know the US very well. They know China very well. They know India very well. But the Arabic world is just coming online. Our models don’t fully understand how you look, your customs, your mannerisms.”
To address this, Luma is partnering with regional institutions to gather and annotate culturally aligned data, employing local experts to ensure accuracy. “It’s really important that AI models understand local customs and cultures. Otherwise, all the content generated will be oblivious to Arabic culture,” he stressed.
Jain sees the Mena region and Qatar in particular as critical to AI’s global development. “This region is showing leadership. There is hunger, energy and forward thinking. Nations will look very different because of AI,” he said.
Beyond media production, Luma’s roadmap includes smart cities, infrastructure monitoring and robotics. Its world-understanding models could detect pipeline corrosion, monitor industrial facilities or power intelligent urban systems. “You don’t programme them with code. You tell them what to watch for corrosion, anomalies, and they alert you,” Jain said.
Jain’s advice for startups in the region: “Building AI products is very different from building software; you need to be deeply technical. It’s not sufficient to buy an AI domain and say you’re an AI company. You need to know your space.” As generative systems accelerate, Jain believes the stakes are civilisational, saying: “In 500 years, people won’t know we existed from archaeology. They’ll know from the internet. And most of that internet will be generated by AI.”
