Why Tejal Patwardhan stopped underestimating the models - Episode 21 – OpenAI Podcast

Why Tejal Patwardhan stopped underestimating the models - Episode 21 – OpenAI Podcast — Transcript

Market views

Bullish

OPENAI— OpenAI's reasoning models (o1, o3) represent a paradigm shift in AI capabilities, with rapid progress across multiple domains including coding, science, and real-world task execution. Models are already passing human baselines and OpenAI research interviews, with no signs of progress slowing.
AILABS— AI labs using their own models internally are developing faster and more capable models, creating a compounding advantage. The capability overhang means models are already capable of far more than people realize or are currently using them for.
BIOTECH— AI models are demonstrating breakthrough capabilities in wet lab optimization and protein synthesis, beating human baselines and setting state-of-the-art efficiency. This suggests significant acceleration potential for drug development and clinical trials.
GINKGOBIOWORKS— Ginkgo Bioworks' automated wet lab robots are being used to validate AI model capabilities in real-world protein synthesis optimization, positioning them at the intersection of AI and biotech automation.

Bearish

BENCHMARKING— Traditional AI benchmarks are becoming saturated and less useful for differentiating model capabilities. Many public benchmarks have quality issues including broken or underspecified problems, making them unreliable for measuring true progress.

Not investment advice.