Trending right now
World Collapse Expert (Ian Bremmer): The Real Crisis Is What Comes After Trump
Is the world on the brink of collapse, and what comes next after a seismic shift in global leadership? Ian Bremmer lays out the stark reality.
Alex Imas on Why Economists Might Be Getting AI Wrong
Are economists miscalculating AI's true impact? Discover why the speed of adoption and job task interconnectedness could reshape our economic future faster than we think.
Our Tax System Should Make You Furious
Is the tax system rigged against you? This episode reveals how the ultra-wealthy pay minimal taxes while you pick up the tab.
#240 - Project Glasswing, Claude Mythos, GLM-5.1, emotion concepts
From AI finding zero-day vulnerabilities to data centers becoming military targets – the latest AI news is stranger and more critical than fiction.
Gadget Prices Are Getting Ridiculous
Are you feeling the pinch of rising gadget prices? MKBHD dives into why your favorite tech is getting ridiculously expensive.
Viewing Podcast: Podcast
AI
Arts
Business
Crypto
Finance
Health
History
Interviews
Investing
Macro
Misc
News
Politics
Product
Programming
Science
Social
Startups
Technology
VC
Episode 15 - Inside the Model Spec
- The "model spec" is an evolving document designed to explain OpenAI's high-level decisions on how their models should behave, acting as a public behavioral interface rather than a strict implementation guide.
- The "chain of command" principle within the model spec manages conflicts between user, developer, and OpenAI instructions, prioritizing OpenAI's safety policies as the highest authority to maintain overall safety while allowing for user steerability.
- OpenAI's model spec is publicly available and open source, with feedback mechanisms for the public to contribute to its ongoing development and alignment with model behavior.
Episode 14 - Building AI for better healthcare
- Vibrant Practice is integrating AI into its EHR and practice management platform to offer personalized medicine solutions for independent healthcare providers.
- Sunita Mohanti, CEO and co-founder, leveraged her experience as a product leader at Meta and her personal journey with functional medicine to address burnout and inefficiencies in traditional medical practices.
- The company's AI-driven approach aims to create an "agentic workforce" to reduce administrative burdens for doctors, enabling them to focus on patient care and providing a more efficient, modern practice management system.
Episode 13 - The Thinking Behind Ads in ChatGPT
- OpenAI is introducing ads to the free tier of ChatGPT to align with its mission of bringing AI to all of humanity, using ads as a proven model for consumer product reach.
- User trust is paramount, with strict principles ensuring ads are visually and functionally independent from AI-generated answers, and sensitive conversations will never feature ads or have data shared with advertisers.
- Users will have significant control over ads, including transparency into data usage, the ability to turn off personalization, and the option to upgrade to paid tiers for an ad-free experience.
Episode 12 - State of the AI Industry
- Organizations are currently in the "profit era" of AI adoption, moving from "if" and "when" to "how" AI will impact their operations.
- The evolution of AI follows a three-stage adoption curve: initial uncertainty, emerging certainty with ongoing ripples, and eventual saturation.
- Deciding whether to build or buy AI solutions, particularly large language models, depends on an organization's core business and the significant resources required for development.
Episode 10 - How AI Is Accelerating Scientific Discovery Today and What's Ahead
- AI is set to accelerate scientific discovery by significantly speeding up research processes, enabling scientists to explore more possibilities in less time.
- AI, particularly models like GPT-5, is proving capable of performing novel scientific tasks, from solving complex mathematical proofs to identifying obscure research papers, and even making contributions to theoretical physics.
- The integration of AI into science represents a new era of collaboration, providing researchers with a tireless, knowledgeable assistant that augments human capabilities and opens up new avenues for exploration across various scientific disciplines.
Episode 9 - ChatGPT Atlas and the next era of web browsing
- ChatGPT Atlas is a new type of browser designed to integrate ChatGPT at its core, enabling natural language interaction with the web for task completion and personalized experiences.
- The current era is considered the "sweet spot" for building compelling AI-powered web experiences due to advancements in LLMs and surrounding technologies, making browsers more capable and accessible.
- The future of computing, as envisioned with Atlas, moves towards users simply stating their intent, with AI agents handling the complex execution of tasks and information retrieval on their behalf.
Episode 8 - OpenAI x Broadcom and the future of compute
- OpenAI and Broadcom are partnering to design and deploy a new custom chip and an entire integrated system for advanced AI, with plans to deploy 10 gigawatts of compute infrastructure starting late next year.
- This collaboration aims to create a highly optimized, vertically integrated solution, from chip design to the final system, to achieve significant efficiency gains and lower the cost of AI models.
- The partnership is described as the "biggest joint industrial project in human history," focusing on building the foundational infrastructure to power future AI advancements and ensure compute abundance for global benefit.
Episode 7 - Live from DevDay
- Model progression from OpenAI has significantly improved both the intelligence and cost-effectiveness of AI personal tutors for students.
- Educators are shifting perspectives from banning AI to recognizing its value in improving productivity and preparing students with essential AI usage skills.
- New tools like the agent builder are allowing domain experts to build customized AI solutions, streamlining development and empowering non-technical users.
Episode 6 - The future of coding with AI
- CodeCraft aims to be more than a code generator by serving as a true coding companion, suggesting improvements and refactoring existing code.
- The tool can vastly improve efficiency and agility in development through rapid prototyping and personalized code suggestions, where it learns your coding style.
- While promising, CodeCraft carries a risk of overreliance and potential for bias from its training data, requiring developers to use it responsibly and maintain critical thinking.
Episode 5 - Defining AGI and the road ahead
- The pace of AI development is accelerating rapidly, with researchers like Shyman Sedor at OpenAI expressing amazement at the speed of these advancements.
- Traditional AI benchmarks are becoming obsolete due to benchmark saturation, necessitating a shift towards evaluating reasoning and problem-solving abilities, exemplified by AI's gold-medal level performance on the International Math Olympiad.
- The ultimate goal of AI research is automated science, where AI systems can generate new knowledge, design experiments, and accelerate scientific discovery across all fields.