The Relationship Between AI's Reasoning Abilities and the Cost of Computational Power
In a recent podcast, NVIDIA CEO Jensen Huang discussed the future of artificial intelligence with ARM CEO Rene Haas. Huang highlighted that the key to AI's potential lies in its ability to offer "reasoning" services. However, achieving this goal requires a significant reduction in the cost of computational power.
He explained that next-generation AI tools will respond to user queries through hundreds or thousands of steps, allowing them to reflect on their conclusions. This level of reasoning capability will differentiate future software from existing systems like ChatGPT. Huang shared that he personally uses ChatGPT daily.
Huang emphasized NVIDIA's commitment to improving chip performance by two to three times each year while keeping costs and energy consumption stable. This transformation is expected to significantly enhance AI systems' abilities in pattern recognition and conclusion drawing. He noted, "We can drastically lower the cost of achieving intelligence. The value of this advancement is clear; by reducing costs, we can enable breakthroughs in reasoning."
According to the latest report from Jon Peddie Research, global shipments of AIB graphics cards are projected to reach 9.5 million units by the second quarter of 2024, a 9.4% increase from the previous quarter and a 47.9% year-over-year growth. NVIDIA continues to dominate the market with an impressive 88% share.