Imagine a world where artificial intelligence (AI) not only processes language, but truly understands and deals with the complexities of the real world.
Gagandeep Reehal, co-founder and CEO of Minus Zero, introduced this revolutionary idea during a keynote at Mint Digital Innovation Summit 2024. Reehal emphasized the need to move from customary vast language models (LLM) to what he called “real-world models.” positioning this inflection point as a unique opportunity for India to lead the AI revolution.
Reehal highlighted the current limitations of LLMs which, despite their advanced capabilities in pattern recognition and language processing, are insufficient for real-world cognition and reasoning.
He illustrated this with several striking examples, including the high-profile problem of AI-generated hallucinations. Unlike humans, who naturally understand and navigate their environments, LLMs often fail, producing erroneous results such as confusing handshakes or misinterpretations of physical interactions. This lack of internal understanding becomes especially problematic in high-stakes applications such as autonomous driving, where AI misjudgments can lead to disastrous consequences.
The crux of the problem lies in the inability of artificial intelligence to reason – a fundamental human characteristic that allows it to make diverse decisions and adapt to elaborate scenarios.
“Basic organisms navigate their environment better than the best AI because they understand and respond to the world in ways that AI currently cannot,” Reehal explained. This gap highlights the urgent need to develop a modern generation of AI models that go beyond simply imitating patterns.
This modern frontier in artificial intelligence research has been actively explored by technology giants such as Meta, Google and other leading organizations. They focus on developing fundamental models that could understand real-world physics and principles, enabling them to make reliable real-world decisions.
These models aim to address current AI shortcomings through deeper environmental knowledge, which is imperative for applications ranging from robotics and stock market investing to delivering drones and navigating complex terrain such as mining sites.
Reehal emphasized that the move towards embodied AI is a massive opportunity for India. “As LLM programs become increasingly commoditized, the real value lies in these core models, where there is still a significant gap to fill,” he noted.
Posted: May 24, 2024 17:17 EST
