Quantum Computing Could Address Rising AI Training Costs, Says IBM

Earlier this month, a report from the Wall Street Journal revealed that a third of nuclear power plants are in discussions with tech companies to supply energy for new data centers. At the same time, Goldman Sachs projects a 160% increase in power usage by data centers due to AI by 2030, leading to a significant rise in carbon dioxide emissions. Notably, each ChatGPT query consumes about 10 times the energy of a typical Google search. This raises a crucial question: will the rapidly increasing cost of training AI models eventually hinder AI's growth?

At VB Transform 2024, a panel led by Hyunjun Park, co-founder and CEO of CATALOG, addressed this issue. Panelists included Dr. Jamie Garcia, director of quantum algorithms and partnerships at IBM; Paul Roberts, director of strategic accounts at AWS; and Kirk Bresniker, chief architect at Hewlett Packard Labs, HPE Fellow, and VP.

Unsustainable Resources and Inequitable Technology

Kirk Bresniker emphasized the urgency of the situation, noting, “The 2030 timeline is close enough for us to make necessary changes but far enough to feel the real impact of our current actions. By 2030, the cost to train a single AI model could exceed the U.S. GDP and global IT spending. Therefore, decisions must be made now to avoid hitting a hard ceiling.”

He further explained, “Sustainability and equity are intertwined. If a technology is unsustainable, it’s inherently inequitable. We must explore how to make this technology universally accessible, which may require significant changes.”

Corporate Responsibility in Tackling the Crisis

Some companies are proactively addressing this looming environmental and financial disaster. AWS is taking steps towards more sustainable practices, such as implementing Nvidia’s liquid cooling solutions and exploring alternative fuels like hydro vegetable oil.

Roberts mentioned, “We’re enhancing our infrastructure to reduce carbon usage and exploring alternative chips. Our silicon, Trainium, and Inferentia chips offer improved efficiency and performance, reducing costs and energy consumption for AI training.”

The Role of Quantum Computing

Dr. Jamie Garcia from IBM highlighted the potential of quantum computing in revolutionizing AI training. Quantum computing can provide resource savings and speed benefits. Garcia explained that quantum machine learning could be applied in three ways: quantum models on classical data, quantum models on quantum data, and classical models on quantum data.

“Quantum computing holds promise for areas with limited or sparse data and interconnected data, such as healthcare and life sciences,” Garcia noted. IBM is actively researching quantum machine learning's potential, with applications spanning life sciences, industrial applications, and materials science.

IBM is also developing Watson Code Assist to help users unfamiliar with quantum computing utilize it effectively. Garcia added, “We’re using AI to optimize quantum circuits and define problems suited for quantum computing.”

Infrastructure Challenges and Future Prospects

While quantum computing offers promising solutions, significant infrastructure improvements are needed. Garcia emphasized, “Reducing power consumption and enhancing component engineering are crucial. Physics research is necessary to meet quantum computing’s infrastructure requirements.”

Transparency and Informed Decision-Making

Bresniker stressed the importance of transparency in decision-making. “We need radical transparency to understand the true costs and returns of these technologies. Understanding energy, privacy, and security characteristics will help us make informed decisions.”

Roberts added, “Organizations need to choose performance characteristics that fit their applications, considering sustainability and energy usage. This approach allows for control and cost-efficient deployment.”

Bresniker concluded, “We must consider alternatives to avoid a hard ceiling. Transparency and understanding the true cost of AI models are essential for sustainable and equitable integration of these technologies.”