In conversation with Sam Altman

Welcoming Sam Altman to the show! (00:00:00)

• Sam Altman is a successful entrepreneur and investor.
• He co-founded Loopt, a local mobile app, and was part of the first class of Sequoia Scouts.
• Altman invested in Stripe, while Jacob invested in Uber.
• Sequoia's highest multiple returning fund came from a tiny experimental fund that Altman and Jacob were part of.
• Altman served as president of Y Combinator from 2014 to 2019.
• In 2016, he co-founded OpenAI with the goal of ensuring the beneficial use of artificial general intelligence.
• Altman left YC in 2019 to join OpenAI full-time as CEO.
• OpenAI launched ChatGPT on November 30, 2022.
• Microsoft invested $10 billion in OpenAI in January 2023.
• In November 2023, rumors spread that Altman was fired from OpenAI, leading to speculation about the company's achievements in artificial general intelligence.
• Altman was reinstated as CEO of OpenAI a few days later.
• In February, Altman reportedly sought to raise $7 trillion for an AI chip project.
• He was also reported to be raising $1 billion from Masayoshi Son to create an iPhone competitor with Johnny Ive.
• ChatGPT has become a household name and has had a significant impact on work and productivity.
• It reportedly reached 100 million users in just two months, the fastest product to achieve this milestone.
• OpenAI's revenue has also grown rapidly, reaching $2 billion in annual recurring revenue (ARR) last year.

What's next for OpenAI: GPT-5, open-source, reasoning, what an AI-powered iPhone competitor could look like, and more (00:02:28)

• OpenAI plans to release an advanced language model, potentially deviating from the GPT-5 naming convention.
• OpenAI prioritizes continuous improvement and aims to make advanced AI technology accessible to free users despite cost challenges.
• The company believes in the value of both open-source and closed-source models and will continue to open-source some of its work.
• OpenAI's ultimate goal is to develop artificial general intelligence (AGI) and ensure its benefits are widely distributed.
• Open-source models are crucial, especially for running AI on devices like phones, but staying ahead of open-source competition requires building a useful intelligence layer and creating enduring value through great products and pricing.
• The industry faces hardware constraints, but algorithmic gains and increased efficiency can help alleviate these limitations.
• The future of devices is uncertain, but there's potential for new form factors and advancements beyond current smartphone capabilities.
• A different interaction paradigm is needed to make AI more accessible, with voice interaction showing promise but requiring improvements in latency and smoothness.
• Combining visual or video data with voice can create powerful multimodal experiences, but societal and interpersonal issues surrounding wearable devices need consideration.
• The ideal AI system should be an always-on, low-friction assistant that provides constant help and improves the user's life.
• Sam Altman envisions a separate entity acting as a super capable assistant, capable of reasoning, pushing back on requests, providing explanations, and behaving like a competent person rather than a sycophant.

How advanced agents will change the way we interface with apps (00:21:56)

• Sam Altman discusses the future of user interfaces and suggests an ideal interface usable by both humans and AI agents, with smooth handoffs and interpretability.
• Voice chat could replace traditional app interfaces, but visual user interfaces remain valuable for certain tasks.
• AI-powered interfaces need improvement before fully replacing traditional methods, but promising examples of AI-powered apps and experiences are emerging.
• Altman highlights the potential of AI tutors and the transformation of healthcare through AI.
• Specialized models or network models working together can address larger problem sets and use reasoning, raising the question of whether reasoning can be generalized.
• The protein modeling example involves training data, sequence data, and a set of processes, similar to interconnected systems in biology.
• Startups focusing on specialized data and training models for specific tasks may face challenges.
• Chat GPT can provide updated information with additional context without retraining, but specialized simulators, connectors, and data are needed. A generalized model may acquire and utilize these resources.
• The Sora video model generates moving images and differs architecturally from language models.
• Altman believes generalized reasoning is crucial for faster progress in AI, requiring specialized models like video models rather than current language models as the starting point.
• The future of AI involves a generalized reasoning system that can render different types of models, such as text, image, and video, by understanding how to perform the tasks.
• There is an ongoing debate in the AI community about the best approach to building AI models, with some favoring regressive models for text and diffusion models for images and videos.

Fair use, creator rights, why OpenAI has stayed away from the music industry (00:33:01)

• Sam Altman expresses concerns about the fairness and fair use of AI-generated art, particularly in the style of a specific artist, as it differs from traditional copyright issues.
• He suggests that the current legal framework may be inadequate and proposes exploring new approaches, especially for art and creative works.
• Altman emphasizes the shift from training data to inference time as AI systems advance, drawing parallels to sampling in music as a potential economic model for fair compensation.
• While acknowledging inspiration from human artists, he argues that using an artist's name or likeness in prompts requires specific permissions or economic arrangements.
• Altman and his team decide not to pursue music generation due to ethical concerns and the challenge of defining boundaries, despite the potential to create a great music model.
• He stresses the importance of preserving human artistic expression and highlights OpenAI's efforts to prevent intellectual property infringement with DALL-E 2.
• Altman acknowledges the complexity of specifying exact limits and views the discussion around AI-generated content as ongoing, requiring collaboration and input.
• He anticipates the music industry's interest in AI as an opportunity for innovation, leading to potential partnerships and legal considerations.

AI regulation, UBI in a post-AI world (00:42:02)

• Sam Altman, the CEO of OpenAI, expresses concerns about proposed AI regulations, particularly in California, due to their complexity and the lack of a clear definition of AI.
• He suggests international oversight for powerful AI systems that could cause significant global harm, similar to the oversight of nuclear weapons and synthetic biology.
• Altman emphasizes the need to balance regulatory overreach and underreach, respecting trade secrets and intellectual property while protecting society.
• He criticizes proposed legislation requiring government auditing of AI models before deployment, advocating instead for an agency-based approach to AI regulation that can adapt to the rapidly evolving field.
• Altman stresses the importance of safety testing for AI models and proposes collaboration within the industry to establish international safety testing standards for potentially harmful models.
• In 2016, Altman initiated a five-year study on Universal Basic Income (UBI) to address societal changes brought by AI and improve the social safety net.
• He believes that providing people with money directly could empower them to make good decisions and reduce poverty.
• Altman introduces the concept of "Universal Basic Compute" as an alternative to UBI, where everyone would receive a portion of GPT-7 compute that they can use, resell, or donate, promoting ownership of productivity rather than monetary compensation.

Sam breaks down how he was fired and re-hired, why he has no equity, dealmaking on behalf of OpenAI, and how he organizes the company (00:52:23)

• Sam Altman, the former CEO of OpenAI, was fired in November 2022 due to a culture clash between the nonprofit board members and those with startup experience. Despite his disagreement with the decision, Altman respects the board's commitment to AGI safety.
• OpenAI's mission to create artificial general intelligence (AGI) raises concerns for some, while others see it as a necessary step in technological advancement. Altman believes AGI development is inevitable and beneficial but acknowledges the significant challenges it poses.
• Altman initially declined equity in OpenAI due to the nonprofit's structure, which required a majority of disinterested directors. However, this decision led to questions about his motivations and potential conflicts of interest, which he finds frustrating.
• OpenAI is exploring building AI infrastructure with a different cost structure to meet growing global needs. Altman emphasizes the importance of clear communication to avoid misunderstandings and regrets not taking equity earlier to address these concerns.
• OpenAI's centralized approach to AI research has proven successful, allowing them to concentrate resources and collaborate on complex projects. Altman stresses the significance of coordination and resource concentration for achieving ambitious goals in AI research.
• Altman invites the interviewer to return after OpenAI's next major launch for more detailed discussions on progress and specific developments.
• Altman recalls a memorable poker game with the interviewer and Chamath Palihapitiya, showcasing his unique playing style.

Post-interview recap (01:05:33)

• Sam Altman believes OpenAI will be one of the four major companies in the AI space, but the economics around AI models are still unclear.
• The scaffolding and infrastructure built around AI models will create opportunities for businesses.
• Altman emphasizes the importance of reasoning beyond just language models and highlights other areas of development within OpenAI.
• There is a cultural clash on the OpenAI board between those from the nonprofit and startup worlds.
• Altman sees significant potential in mobile AI beyond just chat apps and envisions a more comprehensive personal assistant with executive functions.
• He also recognizes opportunities in biology and scientific discovery, particularly in relation to AlphaFold 3.

All-In Summit announcements, college protests (01:10:38)

• Sam Altman announced that Elon Musk and Mark Cuban will be guests at the upcoming All-In Summit.
• The well-received All-In tequila bottle design is undergoing feasibility analysis for manufacturing.
• Peter Thiel was barricaded by protesters at Cambridge, but details of the incident remain unclear.
• Altman stressed the significance of peaceful protests and the right to assembly, while condemning violence and disruption of classes.
• Excessive police force against peaceful protests is deemed dangerous and unnecessary.
• Peaceful protests should be allowed as long as they don't infringe on others' rights.
• The Overton window has expanded, enabling a broader range of ideas to be discussed.

Signs of innovation dying at Apple: iPad ad, Buffett sells 100M+ shares, what's next? (01:19:06)

• Apple's recent iPad ad faced criticism for its portrayal of destroying creative tools, raising concerns about the company's brand identity.
• Berkshire Hathaway's sale of $20 billion worth of Apple shares, despite Warren Buffett's previous praise, raises questions about Apple's future direction.
• Apple's limited experience in large-scale mergers and acquisitions, coupled with its risk-averse nature, hinders its growth and innovation potential.
• Apple's focus on the iPad's thinness as a primary selling point highlights a potential decline in innovation.
• Despite its history of creating groundbreaking products like the iPad, iPhone, and Apple II, Apple has missed opportunities to innovate in areas such as self-driving cars, gaming consoles, and televisions.
• Apple's credibility and trust could drive significant changes in industries like transportation if it adopts a methodical and safe approach to autonomous vehicles.
• The lack of innovation at Apple is attributed to a lack of will and drive rather than a lack of imagination.
• Sam Altman proposes Apple develop a smart home system that integrates various home automation devices, with Apple TV serving as the central hub, offering convenience and a unified interface for users.

Google unveils AlphaFold 3.0 (01:29:41)

• AlphaFold 3, developed by Google, can predict the three-dimensional structure of proteins from their DNA sequences, including all small molecules involved in biological processes. This capability enables the design of new drugs and proteins with specific functions, reducing the risk of side effects and accelerating drug development.
• Google's subsidiary, Isomorphic Labs, retains the intellectual property rights for AlphaFold 3 and monetizes its use, while providing limited access to scientists for non-commercial research purposes.
• AlphaFold 3 demonstrates the potential of AI beyond language models, opening up new possibilities in chemistry, biology, and drug design.
• The AI system can predict molecules that bind and promote certain DNA sequences, potentially leading to breakthroughs in aging research.
• The AI's capabilities extend to predicting how viruses interact with antibodies, paving the way for more effective vaccines and treatments.
• The All-In Summit will take place in Los Angeles from September 8th to 10th. Scholarships will be available soon.
• The All-In Podcast is hiring a researcher to work with the team on primary research, data analysis, and finance.