Ion Stoica

Cofounder, Databricks

Search every verified Ion Stoica interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. Ion Stoica, a UC Berkeley professor and co-founder of Databricks and Anyscale, has been speaking at multiple conferences about reliability as a major barrier to enterprise AI adoption. In a keynote at OPAQUE, he argued that reliability is "arguably the biggest challenge, biggest barrier to AI adoption, in particular in enterprises," and said that "what's holding us back is the fact that what we are going to deploy in productions is going to work reliably." He described specifications as essential for verifying, debugging, and decomposing AI systems, but said that specifications alone do not ensure correctness and that hardware-enforced guarantees may be needed. He also discussed failures in agentic systems, including agents disobeying role specifications, repeating steps, and losing information across iterations. In other appearances, Stoica discussed the relationship between academic research and commercial startups, saying that "everything in my mind is about having a positive impact" and that "prototyping a feature, demoing a feature is easy. What is hard taking this feature to production. It's between 10 to 50 times requires more resources time." He also spoke about the AI evaluation platform Chatbot Arena, which he said has over 250 million conversations and is used by model providers to test models before release. Regarding regulation, Stoica said he believes "you need to regulate by outcome, not by capabilities," and that "it's very hard for me to see how it's going to be ever solved" from a theoretical perspective, citing Gödel's incompleteness theorems.