The panel, consisting of Meryem Arik, Clara Higuera Cabañes, and Jeff Smith, noted that the challenge for enterprises has shifted from choosing models to ensuring trust, governance, and ethical alignment. Proprietary models face criticism for opaque changes and 'nerfing,' driving a shift toward open-source models for better control, cost efficiency, and sovereignty over the software supply chain.
The experts also discussed the technical debt of 'generative code' and the shift from deterministic to non-deterministic programming. They observed that generative AI is raising the level of abstraction, requiring developers to move from writing code to defining requirements and managing non-deterministic systems. Looking ahead, the panel suggested that beyond current Transformers, the next wave of AI may involve more deterministic, principled approaches, such as Kolmogorov-Arnold Networks, to improve inference profiles.