Democratizing Flawed AI Amplifies Risks, Warns Subbarao Kambhampati

• Subbarao Kambhampati warns that democratizing AI models without ensuring their technical correctness amplifies risks.
• He states that wide accessibility does not automatically translate into reliability, safety, or societal benefit.
• Kambhampati emphasizes that 'correctness' is a primary concern, separate from ethical or inclusiveness debates.
• He highlights that models can appear accurate for specific tasks while having fundamentally flawed internal representations.
• The danger lies in hidden errors within models that may only surface upon deeper examination of their internal 'world model'.