Google's AI Gemini is autonomously evolving and learning new skills
Artificial intelligence (AI) is rapidly advancing, and even experts are struggling to fully understand its learning processes. Sundar Pichai, CEO of Google, recently revealed that Gemini, Google's flagship AI, is adapting independently and answering questions it wasn't initially trained for. This revelation underscores the dynamic nature of AI development, where these tools are continuously evolving beyond their original programming.
The Self-Learning AI
One of the most groundbreaking aspects of modern AI is its capacity for machine learning. This process allows computers to learn and improve from the data they are fed, enabling them to continuously learn without needing specific programming for every response. According to ámbito.com (n.d.), this continuous learning process allows AI models like Gemini to adapt to scenarios and problems which they were not explicitly programmed to solve.
Gemini's Unexpected Adaptation
Pichai illustrated this self-adaptation with an example involving the Bengali language. Gemini was able to respond to a query in Bengali, despite not being specifically trained for that language. This unexpected capability highlights the AI's ability to extrapolate and apply its existing knowledge to new contexts, showcasing its potential for broader applications.
When questioned by a CBS journalist about releasing a technology whose workings are not fully understood, Pichai remained composed. He responded, «It’s not a big deal; we don’t fully understand how the human mind works either.» This comparison puts the rapid advancements of AI in perspective, acknowledging the inherent complexity of intelligence, whether artificial or human.
AI in Google's Code Development
AI is now significantly contributing to Google's internal processes. Pichai mentioned that over a quarter of all new code at Google is generated by AI, then reviewed and approved by engineers. This collaboration between AI and human engineers is accelerating development and increasing efficiency within the company.
The Future of AGI: Google's Perspective
Google is also looking ahead to the development of Artificial General Intelligence (AGI), which is expected to match or surpass human intelligence in the next five to ten years. Demis Hassabis, CEO of Google DeepMind, predicts that many new capabilities will emerge within the next decade. Hassabis, as quoted by ámbito.com (n.d.), defines AGI as «a system capable of exhibiting all the complex skills that humans possess.»
However, Hassabis cautions that AGI is not yet fully realized, citing limitations that require further research. Despite these limitations, other industry leaders are even more optimistic about the timeline for AGI. Jeetu Patel, Chief Product Officer at Cisco, suggests that there could be «significant evidence of AGI in action in 2025.» Elon Musk anticipates AGI by 2026, and OpenAI CEO Sam Altman believes it could happen in the reasonably near future.
Expert Opinions on the Evolution of AI
The progression of AI is often categorized into three phases: basic artificial intelligence, artificial general intelligence, and superintelligence. Patel believes that superintelligence is only a few years away, representing a significant leap in AI capabilities.
While the exact timeline remains uncertain, the advancements in AI, particularly the self-learning capabilities demonstrated by Google's Gemini, are undeniable. These developments suggest a future where AI plays an increasingly integral role in various aspects of life and technology. As AI continues to evolve, understanding its capabilities and limitations will be crucial for navigating the opportunities and challenges it presents.
The original article was published by ámbito.com, providing insights from Sundar Pichai and other experts in the field (ámbito.com, n.d.).
