Arvind Narayanan
Showing 1–2 of 2 transcripts.
Zico Kolter: OpenAI's Newest Board Member on The Biggest Questions and Concerns in AI Safety | E1197
Zico Kolter, Harry Stebbings, Sam Altman, David Cahn, Arvind Narayanan
The discussion addresses the accelerating erosion of shared objective reality due to AI's impact on human skepticism, while analyzing how massive untapped multimodal data and architectural shifts now drive model capabilities beyond mere data scarcity. Key figures argue that future AI safety depends on mitigating immediate cyber threats and prompt injection vulnerabilities rather than restricting specific model architectures, as scaling laws suggest performance will continue to improve with sufficient compute. The speaker anticipates industry consolidation and a definition of AGI within 40 to 50 years, emphasizing that survival for enterprises will rely on augmenting human judgment through intelligent systems rather than replacing labor.
Arvind Narayanan: AI Scaling Myths, The Core Bottlenecks in AI Today & The Future of Models | E1195
Arvind Narayanan, Harry Stebbings
The event analyzes a fundamental shift in the AI industry where the era of unlimited scaling is ending due to data scarcity, prompting a pivot toward smaller, on-device models and agent-based applications. It highlights a growing strategic divergence between companies like OpenAI and Anthropic, alongside concerns over market concentration and the need for regulation that targets specific harms rather than the technology itself. Speakers clarify that while job displacement fears are currently overblown, the long-term societal impact on education and cognitive development requires immediate attention alongside the transition to a "layer above" model innovation.