Latest Interviews
Showing 526–540 of 1,195 transcripts.
Clear all filters- RAISE Summit18 min
Why the most demanding companies entrust their AI to OCI & NVIDA
NVIDIA and Oracle announced the upcoming arrival of the Blackwell GPU architecture and the GB200 rack-scale system on Oracle Cloud Infrastructure next year, delivering training speeds four times faster than the Hopper generation while introducing the versatile L40S chip by June. To support massive-scale AI workloads, OCI leverages a high-performance interconnect architecture that matches InfiniBand speeds with a 20-40% cost advantage and now hosts enterprise clients like Microsoft for Bing inference. The strategic partnership further expands into sovereign cloud regions in Frankfurt and Madrid to address data privacy concerns, while combining Oracle Database 23c with NVIDIA Inference Microservices to create an end-to-end pipeline for RAG-enabled AI deployment.
- RAISE Summit11 min
Machine Unlearning: Making AI Forget | RAISE Summit 2024 | Paris
Hirundo addresses the critical compliance gap identified by Stanford research, where 83% of foundational models violate the EU AI Act by trapping irreversible personal data and sensitive information. The company offers a neuro-surgical unlearning platform that enables engineers and compliance officers to selectively erase specific data from trained models in minutes, achieving 95% efficiency gains over traditional retraining. Backed by Alpha Intelligence Capital and led by a team including the former Dean of Computer Science at the Technion, Hirundo positions selective unlearning as the essential standard for navigating global data privacy regulations without sacrificing model performance.
- Y Combinator12 min
When Should You Trust Your Gut?
Startup guidance at YC distinguishes between founders bringing deep prior expertise, who should trust their gut to build products that impress themselves, and those building expertise, who must rely on rapid learning and customer discovery to avoid over-planning. The most frequent failure mode involves role confusion where experts chase investor trends instead of leveraging their unique vision, or novices apply complex roadmaps without validating assumptions. Consequently, advisors are urged to tailor feedback dynamically, shifting from instinct-driven validation for experts to iterative market learning for generalists to align strategy with the founders' actual capabilities.
- Goldman Sachs20 min
Global insurers turn to private credit — and AI
Mike Siegel, Matt Armas, Alison Nathan
The Goldman Sachs Insurance Survey, recorded on April 4, 2024, reveals that global insurers are reallocating $26 trillion in assets toward private credit and high-yield fixed income to counteract recession risks and declining interest rates. While the industry accelerates AI adoption to improve underwriting accuracy and drive consolidation, it simultaneously reduces commercial real estate exposure and pivots ESG strategies toward climate transition financing. These shifts reflect a strategic effort to balance portfolio resiliency against economic downturns with the operational efficiencies promised by new technology.
- All-In Podcast14 min
All-In Summit 2023 Recap
The Awning Conference in Los Angeles convened high-profile figures including Ray Dalio and industry grinders to discuss geopolitical shifts, the transformative potential of AI, and radical scientific advancements like viral brain cancer therapy. Despite logistical friction involving key personality lateness and pre-event resource shortages, the event fostered unscripted debates on macroeconomic trajectories, the rise of "luxury beliefs," and the necessity of separating Silicon Valley from Washington's regulatory influence. Participants ultimately emphasized that despite global instability, collective agency and open dialogue remain the defining mechanisms for addressing humanity's upcoming crises of meaning and governance.
- Sequoia Capital7 min
AI-augmented game development with Inworld co-founder Kylan Gibbs
Kylan In-World has launched a vertically integrated platform combining a local execution runtime engine with an AI studio to resolve the industry tension between technical performance and creative control in game development. Backed by strategic partnerships with NVIDIA, Microsoft Studios, and Ubisoft, the system powers dynamic narratives by processing multi-modal inputs in parallel to generate real-time quests, adjust environmental states, and modify mission objectives based on player dialogue. By prioritizing local latency reduction over cloud-based models, the technology enables non-scripted interactions where emotional states and relationship dynamics evolve in the backend to directly influence gameplay mechanics.
- Dwarkesh Patel10 min
Hitler put Russians on Death Ground - Sarah Paine
The analysis posits that extreme existential threats, such as Germany's invasion of the Soviet Union and Japanese brutality in China, acted as catalysts that forged unified national identities and hardened resistance against invaders who initially faced no organized opposition. While Japanese military leadership remained paralyzed by cultural obligations and fear of execution until Emperor Hirohito intervened in 1945, the United States avoided the "death ground" dynamic seen in Europe by not threatening the wholesale extermination of civilian populations. Ultimately, the historical record suggests that the specific strategies of unconditional surrender and the unique socio-cultural frameworks of the combatants were decisive in determining the durability of these national fronts and the eventual outcomes of the conflict.
- Y Combinator10 min
How to Survive the Crypto Boom & Bust Cycle
Cointracker co-founder Chandan Lodha guides the portfolio tracking and tax compliance platform through volatile crypto cycles by overcoming a 2020 liquidity crisis and a subsequent product obsolescence period. To regain market leadership, the company paused feature development for nine months to rebuild its offering around user obsession and long-term execution, successfully launching Cointracker 2.0 in January. Lodha emphasizes that navigating the industry's boom-bust patterns requires founders to maintain humility during bull markets and conviction during bear markets.
- Sequoia Capital11 min
AI-powered workflow automation with Zapier co-founder Mike Knoop
Zapier is launching Zapier Central, a new platform that replaces its traditional workflow canvas with natural language "AI Bots" capable of executing over 500,000 autonomous actions. These self-healing agents utilize zero-shot inference to automatically configure triggers, resolve parameters, and repair broken steps when external variables shift without manual intervention. The system now supports all 7,000 existing integrations while allowing users to refine bot behavior through direct feedback and a threading feature for step-level approval.
- Dwarkesh Patel9 min
AI progress is about to rapidly accelerate in 2025 – Sholto Douglas & Trenton Bricken
Sholto Douglas, Trenton Bricken
Current AI research progress is primarily constrained by compute availability rather than engineering effort, with scaling experiments suggesting a fifty percent elasticity where doubled resources significantly accelerate discovery. Top teams distinguish themselves through ruthless prioritization and rapid iteration cycles that balance experimental inference against frontier-scale training, effectively treating model development as a greedy evolutionary optimization. This approach transforms the intelligence explosion from a self-writing code phenomenon into a collaborative workflow where AI augments researchers to navigate imperfect information and drive emergent architectural breakthroughs.
- Sequoia Capital7 min
What's next for AI agents ft. LangChain's Harrison Chase
Harrison Chase positions LangChain as the dominant generative orchestration platform, while defining agent systems as loop-based entities that utilize language models to plan, act, and observe external tools. To address current reliability gaps in complex reasoning, the discourse advocates for "flow engineering" strategies that offload planning logic to human-designed state machines and introduces a "rewind and edit" user experience to facilitate human-in-the-loop correction. Furthermore, the next generation of agent applications is expected to integrate distinct procedural and personalized memory architectures, enabling systems to retain specific workflows and user preferences for enhanced personalization and task completion.
- Dwarkesh Patel11 min
How They Became Leading AI Researchers in Just 1 Year – Sholto Douglas & Trenton Bricken
Sholto Douglas, Trenton Bricken
An interpretability team, originally comprising five members, has scaled significantly by prioritizing engineers with extreme agency and multi-disciplinary backgrounds over narrow specialization. Leaders like James Bradbury and Tristan Hume facilitated this growth by recruiting individuals who demonstrated "maniacal" execution and a refusal to be blocked by structural roadblocks. Recent efforts now focus on converting early experimental signals into scalable results by combining biological sparsity insights with rigorous technical investigation across NLP, computer vision, and robotics.
- Goldman Sachs19 min
Are the largest US stocks too dominant?
Ben Snider, Peter Callahan, Allison Nathan
Ben Snyder and Peter Callahan analyze record-breaking market concentration where the S&P 500's top ten stocks now hold 33% of total capitalization, a level unmatched since the 1930s despite current valuations being supported by robust earnings growth rather than speculation. While the "Magnificent Seven" face pressure to sustain revenue gains amid rising AI infrastructure costs, active managers are capitalizing on increasing sector dispersion and finding alpha outside the dominant tech tier. Historical precedents and ongoing investor strategies suggest that such concentrated rallies often conclude with broader market participation rather than immediate collapse, provided the economic backdrop remains stable.
- Sequoia Capital14 min
What's next for AI agentic workflows ft. Andrew Ng of AI Fund
AI agents are driving a paradigm shift from single-step prompting to iterative workflows that combine reflection, multi-agent collaboration, tool use, and planning to achieve results that can surpass larger, faster models running in zero-shot mode. This approach allows systems using smaller language models like GPT-3.5 to outperform GPT-4 on complex benchmarks such as HumanEval by enabling self-correction loops and specialized role delegation. While reflection patterns are now robust enough for immediate integration, emerging capabilities in planning and multi-agent debate are expected to dramatically expand the scope of autonomous tasks over the coming year.
- Dwarkesh Patel10 min
What a GPT-7 Intelligence Explosion Looks Like | Carl Shulman
The discussion outlines mitigation strategies for early detection of hostile AI motivations alongside a defined productivity threshold where AI contributions match or exceed human researcher output. It details operational mechanisms such as voting algorithms, cost-effective scaling of smaller models, and self-generated curricula that enable intelligence explosions without relying solely on brute force capability. These approaches combine hard physical constraints with empirical verification to ensure safety while accelerating innovation through distributed compute and structured learning environments.