Unsupervised Learning: With Jacob Effron

Applied Compute CEO on the Limits of RL, the New AI Hyperscaler & Why Post-Training Wins Inference

AI Vibe Check: Unpacking Chinese Open Models, Cursor's Data Bet & Containment

The rapid evolution of the AI ecosystem centers on the narrowing capability gap between closed-source frontier models and open-source alternatives. While distillation contributes to the performance of Chinese models like Kimi K3, it does not fully account for their competitive standing. Geopolitical tensions underscore...

xAI Co-Founder: Igor Babuschkin Unpacks the Future of Model Development

The rapid evolution of AI agents, particularly in coding, signals a shift toward personal AI systems that prioritize individual control and local hardware deployment. Igor Babuschkin, formerly of DeepMind, OpenAI, and xAI, founded River AI to address the concentration of power within large, closed-source model provider...

Top AI Analyst Unpacks Today's AI Hype Cycle

AI's current trajectory mirrors past platform shifts like mobile and the internet, yet its ultimate impact remains difficult to quantify due to the lack of clear physical limits and the "jagged" nature of model capabilities. While foundation models currently excel in software development, their broader enterprise adopt...

AI Pioneer Jürgen Schmidhuber on the State of AI Today

True AGI requires more than just large language models; it necessitates physical robots capable of interacting with the real world to generate their own experimental data. Current AI development remains overly biased toward human-generated web data, limiting its ability to understand physical reality. Massive capital i...

AI Vibe Check: Lab Wars, Why APIs Might Vanish & Future Predictions

The AI landscape is shifting toward longer-horizon coding agents, which are increasingly replacing individual contributor tasks with management-level agent oversight. While frontier labs dominate, compute scarcity is forcing a strategic pivot: open-weight models face an existential threat as labs prioritize proprietary...

AI Research Legend’s Honest Assessment of Where We Are

Transformer architecture co-author Lukasz Kaiser examines the current state of AI, questioning whether reasoning and next-token prediction are sufficient for true generalization. While current models demonstrate impressive capabilities in coding and mathematics through reinforcement learning, they remain data-hungry an...

A Conversation With Demis Hassabis' Biographer

The AI race is not an inevitable outcome of technological progress but a dynamic shaped by the distinct personalities, strategic choices, and shifting motivations of industry leaders like Demis Hassabis. While DeepMind initially aimed to avoid competitive pressures, the reality of the field has forced a transition towa...

Gemini Co-Lead on World Models, RL's Next Domains & Continual Learning

Multimodal models and world models represent the next frontier in artificial intelligence, shifting from simple text-based generation to complex, interactive systems capable of reasoning and physical world simulation. Gemini co-lead Oriol Vinyals highlights the importance of integrating visual and video data to enhance...

Yann LeCun on What Comes After LLMs

Large Language Models (LLMs) excel at language manipulation but fail to achieve human-level intelligence because they lack the ability to predict the physical consequences of their actions. True intelligence requires "world models"—architectures like Joint Embedding Predictive Architectures (JEPA)—that enable agents to...

AI investor + engineer discuss the current state of AI

The AI ecosystem is currently in a phase of rapid capability exploration, with coding agents serving as the primary driver of innovation and market growth. These agents are increasingly breaking containment to automate complex workflows, signaling a shift where software-generating tools fundamentally change how work is...

OpenAI’s Chief Scientist on Continual Learning Hype, RL Beyond Code, & Future Alignment Directions

The podcast explores the progress and future of AI research, focusing on achieving research intern-level AI and fully automated AI researchers. It highlights the explosive growth of coding tools like Codex and advancements in math and physics capabilities, which serve as benchmarks for improving AI reasoning. The conve...

Owning the System of Record, Building as Models Improve, & Why ITSM is Vulnerable

Serval, an AI-native platform for employee support, challenges legacy incumbents like ServiceNow by building a comprehensive system of record rather than a shallow integration layer. This approach prioritizes deep data synchronization and deterministic workflow automation, leveraging code generation to allow non-techni...

The Most Unexpected Breakout Category in AI

Legora has achieved rapid growth in the legal tech sector, recently securing a $550 million Series D round at a $5.5 billion valuation. CEO Max Junestrand emphasizes that transforming legal workflows requires an AI-native approach, where teams must maintain a "low ego" to abandon outdated features as foundation models ...

Ex-OpenAI Researcher On Why He Left, His Honest AGI Timeline, & The Limits of Scaling RL

The podcast explores the current state and future directions of AI, particularly focusing on scaling paradigms, reinforcement learning, and continual learning. Jerry Tworek, former VP of Research at OpenAI, shares his insights on the limitations of current AI models, emphasizing their struggles with generalization and ...

AI Vibe Check: The Actual Bottleneck In Research, SSI’s Mystique, & Spicy 2026 Predictions

The AI industry is currently navigating a plateau in foundational large language model capabilities, shifting focus toward specialized enterprise deployment and algorithmic efficiency. While consumer-facing LLMs show diminishing returns in performance, reinforcement learning and synthetic data strategies offer pathways...

Edwin Chen: Why Frontier Labs Are Diverging, RL Environments & Developing Model Taste

Foundation model development currently suffers from a reliance on flawed benchmarks that incentivize superficial metrics like verbosity and formatting over genuine accuracy. Optimizing for public leaderboards such as LLM Arena frequently results in models that prioritize "clickbait" responses rather than complex proble...

OpenAI's Head of Product: How the Best Teams Build, Ship and Scale AI Products

In this episode of Unsupervised Learning, Jacob Effron interviews Olivier Godement, Head of Product for Enterprise at OpenAI, discussing the advancements and enterprise applications of AI models. They explore the progress in AI for science, the development of scaffolding and harnesses around models, and the potential c...

China Expert: AI Race, Key Policy Decisions & Unpacking Geopolitical Chip Tension

The Chinese AI ecosystem has rapidly pivoted from vision and robotics to large language model development, fueled by private capital and elite engineering talent rather than direct state-led innovation. While the Chinese government maintains a restrictive regulatory environment, it heavily subsidizes hardware indigeniz...

Anthropic's First PM: Opus 4.5, Rethinking Model Scaffolding & Safety as a Competitive Advantage

Anthropic's Opus 4.5 model is the focus, with Dianne Na Penn, head of product for research at Anthropic, discussing its capabilities and the company's research process. Anthropic aims to improve instruction following, coding, and memory in its models. The company focuses on practical applications, such as enhancing Cla...