Science and TechnologyScience and Technology
Mon, March 10, 2025
[ Today @ 01:05 AM ] - MSN
Nedbank Cup semi-final draw
Sun, March 9, 2025

How Much Compute and Video to Solve Real World Superintelligence


Published on 2025-03-10 00:42:26 - NextBigFuture
  Print publication without navigation

  • Yann LeCun is both a giant in the field of AI but also a major skeptic of the ultimate potential of large language models. He is a major skeptic about large

The article from NextBigFuture discusses the computational and data requirements for achieving real-world superintelligence, particularly focusing on the advancements in AI and machine learning. It highlights the ongoing debate about how much compute power and video data are necessary to train AI systems to reach superintelligent levels. The piece references various expert opinions and research, including insights from AI researchers like those at DeepMind and xAI, suggesting that while current models like those from OpenAI require vast amounts of data and computational resources, future models might become more efficient. It also touches on the idea that real-world superintelligence might not just depend on scaling up current methods but could involve breakthroughs in algorithms, data quality, and possibly even new forms of data like synthetic or simulated environments. The article concludes by noting the speculative nature of these estimates, emphasizing that the path to superintelligence involves not only hardware and data but also significant software innovations and understanding of intelligence itself.

Read the Full NextBigFuture Article at:
[ https://www.nextbigfuture.com/2025/03/how-much-compute-and-video-to-solve-real-world-superintelligence.html ]