The surge in generative AI has led to a significant increase in demand for GPUs, the powerful chips that drive large language models and other AI applications. However, this demand may not be sustainable in the long run, according to Erick Brethenoux, Gartner’s chief of research for AI. Brethenoux, who has observed AI for 45 years, points out that specialist hardware for AI workloads has historically been a temporary necessity during the “brute force” phase of AI development.
Once programming techniques are refined and general-purpose machines can handle the tasks, the need for specialist hardware diminishes. “If you cannot find the elegant way of programming… it [the AI application] dies,” Brethenoux said at Gartner’s Symposium in Australia.
Generative AI’s long-term viability questioned
He believes that generative AI will follow this trend, and organizations can benefit from AI without relying heavily on generative AI. Brethenoux noted that generative AI currently accounts for “90 percent of the airwaves and five percent of the use cases.” Many organizations have realized that their existing AI applications, such as machine learning for predictive maintenance, are already making significant contributions to their business. Gartner vice president and distinguished analyst Bern Elliot echoed these sentiments in another session titled “When not to use generative AI.” Elliot emphasized that generative AI lacks reasoning powers and is best suited for content generation, knowledge discovery, and conversational user interfaces.
Despite improvements in reducing “hallucinations” – responses with no factual basis – Elliot cautioned that generative AI is still unreliable and not a sign of the technology’s maturity. He recommended using composite AI, which integrates generative AI with established AI techniques, as a safer approach. As businesses navigate the generative AI landscape, they will need to carefully consider the long-term viability of their investments in specialist hardware and focus on leveraging AI in ways that provide lasting value to their operations.
Cameron is a highly regarded contributor in the rapidly evolving fields of artificial intelligence (AI) and machine learning. His articles delve into the theoretical underpinnings of AI, the practical applications of machine learning across industries, ethical considerations of autonomous systems, and the societal impacts of these disruptive technologies.
