Earlier this week at the Milken Global Conference in Beverly Hills, five key figures from the tech world sat down to discuss the current state of artificial intelligence. Christophe Fouquet from ASML, Francis deSouza from Google Cloud, Qasar Younis from Applied Intuition, Dimitry Shevelenko from Perplexity and Eve Bodnia from Logical Intelligence shared their insights on various challenges facing AI.

The first bottleneck identified is the physical limit of chip production. Despite significant efforts, Fouquet believes that for the next few years, there will be supply limitations in chips, meaning that major tech companies won’t get all they pay for. DeSouza highlighted Google Cloud's revenue growth and backlog to underline this point, expressing the reality of high demand.

Data collection is another constraint mentioned by Younis. Applied Intuition builds systems for various industries and stresses that real-world data can't be fully simulated, leading to a long time before synthetic models suffice. This highlights the need for extensive testing in actual environments.

Energy constraints are also discussed, with Google considering space-based data centers as part of their solution. DeSouza explained how efficiency through integration pays dividends, and Fouquet echoed similar points about energy consumption growing alongside increased computing power.

Bodnia's company, Logical Intelligence, proposes a different approach to AI by focusing on energy-based models (EBMs). These models are designed to update knowledge as data changes, making them suitable for domains requiring physical rule understanding. This could signal a shift in the paradigm of how we design and use AI.