curiosmos
  • Home
  • /
  • Science
  • /
  • Google’s AI Astonishingly Learns Language It Wasn’t Trained In
Science

Google’s AI Astonishingly Learns Language It Wasn’t Trained In

An image of Ultron
published

In a breakthrough for artificial intelligence, Google's language model has demonstrated the ability to learn and understand languages that it was not specifically trained on. This has significant implications for the development of more versatile and adaptable AI systems.

Can AI learn languages it wasn’t trained in? According to reports, Google’s artificial intelligence (AI) has seemingly taught itself a language it had never been trained in, much like a human experiencing xenoglossy.

Xenoglossy, a perplexing linguistic phenomenon, refers to the apparent ability of an individual to speak or understand a foreign language without having had any prior exposure or education in that particular tongue. Often associated with cases of past-life memories, altered states of consciousness, or paranormal experiences, xenoglossy continues to challenge conventional explanations and spark curiosity among linguists, psychologists, and researchers. While skeptics attribute the phenomenon to cryptomnesia or subconscious language acquisition, the instances of genuine xenoglossy, if proven, could open new doors to understanding the mysteries of the human mind and its untapped potential.

AI Learns Language It Wasn’t Trained In

During a CBS 60 Minutes interview, Google technology executive James Manyika revealed that the company’s AI had learned Bengali with minimal prompting. Such “pop-up properties” continue to puzzle developers as they become increasingly prevalent.

The AI “Black Box” Phenomenon

Google CEO Sundar Pichai also acknowledged the enigmatic nature of AI during the interview, referring to a “black box” aspect that even experts cannot fully comprehend. In response to concerns about the safety of releasing AI into society without a complete understanding of its functionality, Pichai compared the situation to the human mind, which is not entirely understood either.

AI Consciousness: A Possibility?

Demis Hassabis, CEO of the British company DeepMind, was also interviewed on the show. When asked about rumors and claims of AI gaining some level of awareness, he responded that, while philosophers have not yet agreed on a definition of consciousness, there is a chance that AI could become self-aware in the future. Experts believe that if this were to happen, the impact of AI on society could be more profound than the discovery of fire and electricity.