Five AI Economy Architects Discuss Failures in the System
Image Credits:Milken Institute /
Insights from the Milken Global Conference on AI Supply Chain Challenges
Earlier this week, five influential figures in the AI supply chain gathered at the Milken Global Conference in Beverly Hills. They discussed critical issues ranging from chip shortages and data center innovations to questioning the fundamental architecture of AI technology. The panel included Christophe Fouquet, CEO of ASML; Francis deSouza, COO of Google Cloud; Qasar Younis, co-founder and CEO of Applied Intuition; Dimitry Shevelenko, chief business officer of Perplexity; and Eve Bodnia, a quantum physicist and founder of Logical Intelligence.
Real Bottlenecks in AI Development
The discourse revealed tangible bottlenecks impeding the AI industry’s growth. Fouquet emphasized the ongoing “huge acceleration of chip manufacturing,” but predicted that hyperscalers like Google, Microsoft, Amazon, and Meta would face supply limitations for the next two to five years. These tech giants won’t receive all the chips they are financially committed to, affecting their operations.
DeSouza underlined the gravity of the situation by sharing that Google Cloud recently recorded a revenue of over $20 billion, with a backlog that surged from $250 billion to $460 billion in just one quarter. He asserted, “The demand is real.” Younis provided a different perspective, focusing on data as the primary bottleneck for Applied Intuition. His company, which specializes in autonomy systems for various sectors, highlighted the necessity of real-world data for training models, stating that synthetic simulations are inadequate for accurately capturing reality.
The Energy Dilemma
Beyond chip shortages, energy availability emerged as another pressing concern. DeSouza revealed that Google is seriously considering orbital data centers as a solution to energy constraints. While this would provide access to more abundant energy, operating in space poses unique challenges due to its vacuum environment, which complicates traditional cooling methods.
DeSouza further illustrated Google’s strategic advantage by highlighting the company’s integration of AI components, from custom TPU chips to final models. He argued that this combination results in greater energy efficiency than conventional techniques. Fouquet echoed this sentiment, noting that while the industry is investing heavily in computing power, increased energy needs inevitably come with costs.
A Different Kind of AI Intelligence
While the industry largely debates large-scale language models, Bodnia is charting a different course with her startup, Logical Intelligence. Her approach utilizes energy-based models (EBMs) that mimic the human brain’s understanding of data rules, rather than merely predicting sequential patterns. Unlike conventional large language models with hundreds of billions of parameters, her largest EBM runs with just 200 million parameters but is significantly faster and capable of updating knowledge continuously.
Bodnia argues that EBMs are better suited for applications requiring an understanding of physical rules, such as robotics and chip design. She pointed out, “When you drive a car, you make decisions based on your surroundings, rather than searching for language patterns.” This approach may gain increased attention as the AI sector reevaluates the sufficiency of scale alone.
Trust and Control in AI Deployment
Shevelenko elaborated on Perplexity’s evolution into a “digital worker” entity, shifting from a basic search tool to a sophisticated assistant for knowledge workers. His pitch is compelling but raises the important question of control over such agents. His solution involves granularity in permissions, where enterprise administrators can dictate the data access levels for different agents. For example, Perplexity’s Comet agent requires user approval for actions, reinforcing trust and security—a crucial factor for companies prioritizing client confidentiality.
National Sovereignty and Physical AI
Younis made a significant observation about the relationship between physical AI and national sovereignty. Unlike traditional digital technologies that predominantly face challenges at the application layer, physical AI systems—like autonomous vehicles and drones—interact directly with the real world. This reality raises pressing questions about safety, data management, and regulatory concerns.
Countries are increasingly wary of foreign-controlled physical AI solutions, with many insisting that such systems operate under domestic control. Younis remarked that fewer nations can deploy robotaxis than those with nuclear capabilities, suggesting a shift in global power dynamics. Fouquet framed the debate differently, highlighting China’s advancements in AI constrained by its lack of access to cutting-edge manufacturing technologies. This could create a disparity where Chinese models built on older hardware lag despite their software capabilities.
Concerns About the Next Generation
The panel concluded with a crucial question from the audience: How will these advancements affect the next generation’s critical thinking skills?
DeSouza responded optimistically, noting that enhanced AI capabilities could enable humanity to tackle significant challenges, including neurological diseases and climate change. He believes this evolution could spark new levels of creativity.
Shevelenko pointed out that while entry-level jobs may be declining, the tools for launching independent projects are more accessible than ever, reinforcing the idea that innovation relies on individual initiative and curiosity. Younis differentiated between sectors: while knowledge work is changing, physical labor roles—especially in agriculture and mining—experience chronic shortages. In this regard, physical AI serves to fill gaps in labor rather than replace willing workers.
Conclusion
The discussions at the Milken Global Conference illuminated the multifaceted challenges and opportunities facing the AI industry today. As bottlenecks in chip supply and energy availability persist, the industry is poised for a transformative period influenced by innovative approaches like energy-based models and enhanced digital worker capabilities. Looking ahead, a careful balance of trust, control, and national considerations will shape the future of AI as it continues to intertwine with the fabric of society.
Thanks for reading. Please let us know your thoughts and ideas in the comment section down below.
Source link
#architects #economy #explain #wheels #coming
About The Author
