Countries must own their AI destiny, need to decentralise deployment, open source an important tool: Mistral AI CEO | Business News

“We are facing too much concentration of power in artificial intelligence. We don’t want to be in a world where three or four enormous companies actually own the deployment and the making of AI …in a world where multiple digits of the GDP are going to be produced by AI in the coming years, we need to ensure that everyone that runs AI workloads, actually has access to the turn on and turn off button. It can actually make sure that there is a business continuity and that they are not dependent on external providers that can actually turn off that button. The global AI landscape is evolving rapidly but too often the narrative is shaped by a focus on control, surveillance and geopolitical leverage.” he said while speaking at the India AI Impact Summit.

Stating that open source is “not a radical idea”, Mensch said it has played a role in building cloud infrastructure and a secure internet. He said a dichotomy is being faced in the AI world between those who support open source and those who oppose it. “A few companies like Mistral are working on making the world knowledge compressed into models that can be used as the basis for creating applications and for serving the enterprises, the states, education, healthcare, and another world, where all these models are actually compressed by a few large private corporations that actually use them as leverage against their users,” the CEO of the France-based AI startup said.

Given how AI is going to change the equilibrium in the world in the next few years, there is a need to “ban excessive leverage” to have that equilibrium to remain sustainable and stable, Mensch said. “…and in that India is a leader in betting on self-reliance, which is absolutely critical for a country of this size,” he said.

Noting the diversity of 22 languages in India, Mensch said to make better open source models, there is a need to collaborate with local ecosystems, research institutes to make sure that languages make their way into the AI models. Mistral AI models have focused on making rare languages more usable, he said. “We like to collaborate with companies and with research labs that are working on acquiring the content that is needed to make the models better,” he added.

India’s diversity of culture and its ambition, the size of the market, and the size of its industry provides immense leverage when it comes to building AI that will be differentiated, that will be a major product of export, and that will provide them with even more leverage on the international scene, Mensch said. “By controlling its own AI, India can become a global hub for innovation and can also lead the way for the entire world,” he added.

Governments can improve the quality of their services, the lives of their citizens, and the quality of life of communities, he said. However, it will only work if everybody can access AI and if everybody understands how to use it. “That’s why multilingual AI is actually a very strong necessity. That’s also why we need to focus on training people and on bringing them the technology at an affordable price,” he said.

The true potential of AI is through the level-playing field, said Mensch. To ensure that, it’s important that the technology being built does not widen the digital divide, he said, adding that there is a need to ensure that no language and no culture are left behind. “We need to ensure that the technology that we’re building is affordable and therefore our focus on open source is a very important aspect,” he said, adding that the governments need to invest in AI infrastructure that they own and need to make sure to distribute it to local and regional providers.