Debating De-Growth: Carbon Economies and the Club of Rome

Joe Weisenthal

Noel King

Odd Lots

Why the Degrowth Movement Is Having a Moment

- The fundamental tenet of economics is often seen as growth, but there is a debate about whether it should be the sole goal.
- The de-growth movement advocates for a planned reduction of energy and resource use to bring the economy into balance with the environment and reduce inequality.
- Some argue that growth can be sustainable and decoupled from carbon emissions, while others believe that less economic activity is necessary to address environmental concerns.
- The de-growth movement has gained popularity in recent years, with influential figures and publications discussing its ideas.
- The movement is mainly concentrated in Europe, but its influence is spreading globally.
- Personal experiences and concerns about the environment have led some individuals to support the de-growth movement.

Cisco Buys Splunk: Intrest Rates, M&As and IPOs

Entrepreneur, angel investor, author, and podcaster. Founder and CEO of Inside.com and Launch

Jason Calacanis

This Week in Startups

Jason Unplugged: IPO market, WaPo tactics against Dave Portnoy + Live Q&A | E1815

- Raising interest rates is aimed at increasing unemployment to cool down the economy and prevent inflation.
- M&A activity is expected to increase as large companies look for growth and smaller companies consider cashing in.
- Cisco plans to acquire cybersecurity company Splunk for $228 billion in cash.
- Three companies recently went public: Arm, Klaviyo, and Instacart.
- Instacart's business model relies heavily on ad revenue and targeting consumers during the checkout process.
- Instacart's valuation is around $8 billion, with potential for growth based on its ad business.

Chammath Breaks Down the Current IPO Market: ARM, Instacart, etc

Entrepreneur, author, and investor in internet technology firms

David Sacks

Avi Friedberg

Entrepreneur, businessman, and angel investor. Founder of The Climate Corporation

David Friedberg

All-In with Chamath, Jason, Sacks & Friedberg

E146: Did the Fed break the VC model? Plus IPOs, M&A, revaluing unicorns & more

- Avi Friedberg suggests that the IPO market would be better served if banks were forced to be locked or if allocations were forced to be locked for a longer period of time, such as six or seven years.
- David Friedberg points out that the pricing of a direct listing can lead to price volatility and may not serve anyone better. He suggests that companies should float at least 15 to 20% and do a better job of allocation.
- Avi Friedberg agrees that the focus should be on the business fundamentals rather than the stock price, but acknowledges that many people get caught up in the price and create distractions.
- The conversation shifts to discussing the success of Instacart and the returns for investors. David Friedberg mentions that the fund he is in has already been paid back many times over and expects a multiple of 100-200x on his investment.
- They also discuss the dilution and inefficiency of companies in the late-stage investment environment, and how it may have been bad for both late-stage and early-stage investors.
- Avi Friedberg shares a visualization that compares the returns of investing in Instacart versus investing in the S&P 500, highlighting the concept of alpha and the decision to invest in a specific company versus the market.
- They conclude that the current market conditions and alternative investment options may impact the appetite of limited partners to invest in venture capital.

Sam Harris Covid Postmortem: Trust in Institutions (Airline Travel vs. Covid)

Sam Harris

Making Sense with Sam Harris

#335 — A Postmortem on My Response to Covid

- Flying on an airplane requires trust in the competence of others and the systems in place for safety.
- While there are risks involved, flying has been remarkably safe and most people judge it to be good enough.
- Trust in institutions and procedures is necessary for normal life to be possible.
- During the pandemic, trust in institutions and following the best available information was important.
- The official number of COVID deaths in the US is approximately 1.1 million, and a quarter of those were under age 65.
- The fixation on health and fitness as a defense against viruses is a delusion, as there are factors beyond individual control.

Transformer Paper Author Illia Poloshukin on Blockchain, AI & Pivoting to Web3

Co-host of the podcast No Priors and CEO of Minion AI

Elad Gil

Illia Polosukhin

No Priors: Artificial Intelligence | Machine Learning | Technology | Startups

How the World’s First Blockchain Operating System Is Democratizing WEB3, With Near’s Illia Polosukhin

- Ilya Polosukhin was involved in the original Transformers paper in AI.
- The initial mission of NIR was to teach machines to code, but they shifted to blockchain due to challenges in paying developers.
- Nira aims to be a blockchain operating system, providing an environment for Web3 experiences without the need to think about low-level implementations.
- The overlap between AI and blockchain/Web3 includes marketplaces for resources and the potential for AI agents to become economic agents that can communicate and pay others for work.
- AI could be used to run organizations, replacing middle management and creating a new layer of management that combines blockchain and AI.
- Blockchain and AI can be used in the context of cancer research to coordinate experiments, funding, and information sharing.
- AI could provide on-the-job feedback in a depersonalized way, potentially starting with DAOs.
- The intersection of blockchain and AI raises questions about alignment and dealing with misinformation at scale.
- Web3 and blockchain provide primitives such as cryptography, authentication, identity, and reputation to address these challenges.

Cloud Computing in AI: Unlocking Tons of Latent Hardware

Legendary software and hardware engineer, leading projects at Apple, Tesla, Google, SiFive, and Modular AI, including the development of Swift, LLVM, Clang, MLIR, CIRCT, TPUs, and Mojo

Chris Lattner

Expanding AI chip capabilities beyond Nvidia with Modular CEO Chris Lattner | E1808

- The hardware used for AI jobs is often underutilized.
- Training and deployment in AI are two different problems.
- Training jobs are massive and often use time-sharing systems.
- Production deployment requires scale-out and utilization of thousands of nodes.
- The industry focuses more on training than on getting models into production.
- Deploying ML involves using multiple tools and point solutions.
- Hardware makers tend to build solutions for their own hardware, creating compatibility challenges.
- AI software is still in its early stages and lacks production-quality solutions.
- Modular aims to provide a unified AI stack for hardware makers.
- Nvidia is a supportive partner of Modular.

How is NVIDIA Able to Maintain their Profit Margin vs TSMC and ASML?

Co-host of The Cognitive Revolution podcast

Nathan Labenz

Erik Torenberg

"The Cognitive Revolution: How AI Changes Everything"

Hollywood Strike Update and AI Roundup with Trey Kollmer

- Nathan Labenz discusses the need for exploration and contribution in the field of AI and model behavior.
- He mentions the use of abstraction and variable approaches to improve performance in reasoning tasks.
- Erik Torenberg asks about the value capture in the chip manufacturing industry, specifically regarding NVIDIA, TSMC, and ASML.
- Nathan Labenz notes the significant increase in NVIDIA's market cap compared to TSMC and ASML, and speculates on possible reasons for this.
- They discuss the potential pain of switching fabs and the value capture in the intelligence supply chain.

Currently co-executive producer for the TV show Ghosts, was described to me as perhaps the single most knowledgeable guild member on the topic of AI

Trey Kollmer

Musk's Superpower: Question, Fire, Hire  - Walter Issacson

Host of the podcast show 'Lex Fridman Podcast'

Lex Fridman

Author of biographies on Elon Musk, Steve Jobs, Einstein, Benjamin Franklin, Leonardo da Vinci, and many others

Walter Isaacson

Lex Fridman Podcast

#395 – Walter Isaacson: Elon Musk, Steve Jobs, Einstein, Da Vinci & Ben Franklin

- Elon Musk fired 85% of employees in three rounds, causing concern among his team.
- Musk has an algorithm that involves questioning every requirement and deleting unnecessary parts.
- He may have overdone it by 20% but is now trying to hire more people.
- Hiring and managing great teams are crucial to Musk's and Steve Jobs' success.
- Musk has a talent for finding the right people and promoting them.
- Trustworthiness, excellence, and drive are important qualities in Musk's hires.
- Musk's time management involves focusing intensely on one task at a time and switching between different projects throughout the day.

Chris Christie & Sacks: Heated Debate on Ukraine

Entrepreneur, VC, engineer, and philanthropist. Founder and CEO of Social Capital

Chamath Palihapitiya

Former NJ Governor and Republican Presidential Candidate

Chris Christie

E145: Presidential Candidate Chris Christie in conversation with the Besties

- Chris Christie supports Ukraine and believes Putin is responsible for the invasion
- Chamath Palihapitiya also holds Putin responsible but believes US inaction contributed to the situation
- Chamath does not believe Ukraine should be admitted to NATO until they win the war
- Chris Christie thinks Ukraine's admission to NATO is almost inevitable now
- Chamath believes it was a mistake to express intentions to bring Ukraine into NATO without a clear plan
- Chris Christie would have been more aggressive in providing military hardware to Ukraine
- Jason Calacanis questions the effectiveness of US support and the lack of ammunition
- Chris Christie agrees that the Pentagon needs to be more efficient and effective but does not support cutting the defense budget
- Chamath suggests reallocating funds based on priorities and implementing zero-based budgeting principles

Baliji: US Decline, America First, India & "The Sine Waves of Civilization" 

Entrepreneur and investor, former chief technology officer at Coinbase, former general partner at the venture capital firm Andreessen Horowitz, and founder of companies with nine figure exits

Balaji Srinivasan

"Upstream" with Erik Torenberg

E24: Balaji Srinivasan on Social Tribalism, Cloud Cartography, Internet Values

- India has experienced significant growth and unity in recent years, with improvements in infrastructure and achievements in various sectors.
- The US, on the other hand, is showing signs of decline and disunity, resembling the India of the past.
- The US is losing ground in terms of attracting high net worth individuals and talented individuals, with other countries such as Canada, Switzerland, Dubai, Singapore, Australia, and New Zealand gaining popularity.
- The global center of the economy is shifting back to Asia, with talent and capital flowing towards Asian countries.
- The concept of "Western values" may no longer accurately describe the values shared by countries like the US, Australia, New Zealand, Japan, Singapore, and Dubai. Instead, these countries are more aligned with "internet values."
- Prioritizing "America first" can lead to a breakdown of global trade and a loss of credibility for the US. It is not a win-win ideology and may result in America being selfish towards other countries.
- Simply being proud of one's country is not enough to ensure success, as seen in the case of Argentina's economic boom and subsequent decline.

Elon Musk & ADL in Twitter Conflict Over Hateful Posts

Kevin and Casey

Kevin Roose

Casey Newton

Hard Fork

Escape From Burning Man + Musk vs. the A.D.L. + Listener Questions

- Elon Musk engaged in a conflict with the Anti-Defamation League (ADL) on Twitter.
- The ADL has been pressuring social media platforms, including Twitter, to address hate speech and content moderation.
- The ADL reported an increase in anti-Semitic and other hateful posts on Twitter after Elon Musk took over the company.
- The ADL asked advertisers to pause their spending on Twitter to pressure the platform to clean up its content.
- Elon Musk threatened to sue the ADL and blamed them for a drop in advertising revenue on Twitter.
- The ADL denies calling Elon Musk anti-Semitic and claims they are not trying to kill Twitter.
- The ADL had a productive meeting with Linda Iaccarino, the CEO of Twitter, but it is unclear what led to Elon Musk's conflict with the ADL.

Cerebras CEO Andrew Feldman: The Chip Game in the AI Boom

Sarah Guo

Andrew Feldman

The World’s Largest AI Processor with Cerebras CEO Andrew Feldman

- Strategic partners are excited about properly representing their culture and language in language models
- Starting a chip company is challenging due to the long development time and high cost
- Chip companies require a different approach to engineering, with a focus on QA and simulation
- Software is still a significant component in chip companies, with a need for low-level software and compile stacks
- NVIDIA's dominance in the AI accelerator market is being challenged due to high costs and supply issues
- Large enterprises and generative AI startups are potential customers for AI accelerator companies
- The semiconductor layer may see increasing specialization for different model types
- The chip market has an inflexible supply chain, leading to supply crunches and forecasting challenges
- The AI market is accelerating rapidly and forecasts are likely to underestimate its growth
- Scaling models has trade-offs between accuracy, size, and cost
- Fine-tuning models with human feedback is important for improving their performance
- Different large enterprises have interesting data sets that can benefit AI training

Self Driving is Finally Getting Somewhere: Waymo, Tesla & Uber

Host of Bloomberg's podcast, 'CHIPS and Science Act of 2022'

Tracy Alloway

Long-time technology journalist and analyst, author of the Understanding AI newsletter

Tim Lee

Self-Driving Cars Might Finally Be For Real This Time

- Waymo is owned by Google and Cruise is owned by GM, while Tesla is in a different market with driver assistance technology.
- Tesla's driver assistance product requires the driver to be present and attentive.
- Uber hired members of the Carnegie Mellon University robotics department for their self-driving car program, but it ended after a fatal accident.
- Self-driving cars have the potential to be a profitable business model, with the cost of labor being a significant factor.
- Self-driving cars could reduce the need for parking spaces and enable congestion charging.
- Self-driving cars could be used for package delivery, but the market for that may be different.
- The safety of self-driving cars compared to human-driven cars is difficult to assess due to the limited number of miles driven by self-driving cars.
- The insurance and liability aspects of self-driving cars are still uncertain.
- The regulatory landscape for self-driving cars varies by state, with some states having more regulations than others.

Yoko Li

Why a16z Built a Town for AI People

- The previous project called Companion App was based on the idea of AI companions that can have conversations and learn from humans.
- The project implemented a memory system that allowed the AI companions to remember interactions.
- The community showed interest in having AI companions in group settings and playing games with them.
- Yoko and her husband built a prototype for AI Town, an environment where AI companions can interact with each other and humans.
- The project was open-sourced and can be used for commercial projects.
- The technology behind AI Town includes memory implementation and language models.
- The project aims to explore the future of relationships and human interaction with AI.

a16z's Town for AI People

Venture capitalist from Benchmark and podcast host

Bill Gurley

The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch

20VC: Why Small Funds Outperform Large Funds & AUM is a Vanity Metric | Why 99% of Investments in AI Startups Will Go To Zero | Being a "Traction First" VC & Investing Lessons from Investing in Canva and Missing Figma with Nikhil Basu-Trivedi

- Bill Gurley believes that investing in star-studded teams with large funding rounds is not always the best strategy. He prefers teams with products already in the market and early signs of success.
- He thinks that many multi-stage funds are wasting money on AI seed companies, and that large investments in seed rounds don't make sense.
- Gurley believes that companies with long runways but no product-market fit need to find a way to deserve their existence. He emphasizes the importance of momentum and winning in startup culture.
- He advises having honest conversations with founders about returning money if necessary, but also acknowledges the importance of giving teams a chance to iterate and experiment.
- Gurley ranks traction, early signs of product-market fit, and people as more important than market when making investment decisions.
- He admits to being seduced by founders who are great at fundraising but not necessarily good at building a business. He emphasizes the importance of substance and understanding beneath the surface.
- Gurley suggests asking founders about their main challenge or discussion topic in a hypothetical first board meeting to assess their clear-eyed thinking and transparency.
- He believes that great founders are confident in what is working but also self-aware and clear-eyed about what is not working yet.
- Gurley acknowledges his own insecurities as an investor and founder, but feels that he has found his voice and authenticity through his writing and the creation of Footwork.

Co-Founder & General Partner at Footwork, an early-stage focused venture firm investing its first fund. In his venture career, he has invested in the early rounds of several companies that have exited or are currently valued at over $1B, including Athelas, Canva, ClassDojo, Color Health, Frame.io, Imperfect Foods, Lattice, and The Farmer's Dog. Prior to Footwork, Nikhil was a Managing Director at Shasta Ventures, on the investment team at Insight Partners, and on the founding team at Artsy.

Nikhil Basu Trivedi

Are VCs Burning Money on AI?

"Elon Musk's Superpower: Taking Calculated Risks" - Tim Urban

Stand-up comedian and co-host of the podcast

Konstantin Kisin

Francis Foster

Author of the blog Wait But Why and a new book What's Our Problem?: A Self-Help Book for Societies

Tim Urban

TRIGGERnometry

How the Media Creates Delusional Tribes - Tim Urban

- Tim Urban believes that it is important to look at the bigger picture and consider the long-term impact of changes in social media companies like Twitter.
- He thinks that Elon Musk's mission with Twitter and X is to improve social media and create an "everything app" similar to WeChat in China.
- Elon Musk's ability to reason from first principles and challenge conventional wisdom is his superpower and allows him to disrupt industries.
- Tim Urban believes that people often overestimate the risks involved in taking calculated risks and that failure is not as detrimental as we think.
- Elon Musk's wealth and reputation allow him to take calculated risks without risking his family's future or his own comfort.

"Republicans Want an Attack Dog" + Interview w/Vivek

Vivek Ramaswamy

"Moment of Zen"

E39: Bull/Bear Case for Vivek and Erik’s interview with Vivek Ramaswamy

- Vivek Ramaswamy is in favor of merit-based immigration and has a vague platform with notable policy proposals.
- Vivek's platform is similar to that of DeSantis, with a focus on anti-woke and administrative state issues.
- Both Vivek and DeSantis have legal backgrounds and want to limit the power of regulatory agencies.
- Vivek is seen as wonky and intelligent, but his success lies in his ability to create viral moments on Twitter.
- However, Vivek's lack of coolness and cringeworthy moments may not resonate with the average Republican primary voter.
- Vivek's appeal lies in his anti-establishment stance and high integrity, which may attract voters who felt guilty about supporting Trump.
- The Republican base wants a bully who can own the libs and drive their enemies crazy, which is what Trump provided.
- It is uncertain if Vivek can fill this role effectively, and it is unlikely that a Republican candidate can be considered cool in the current media landscape.
- The Republican primary focus should be on winning the primary to gain power, rather than elevating the integrity and intelligence of the candidate.
- Trump's potential re-election in 2024 could impact the Republican Party's future, but if Biden remains president, the field is wide open.
- Vivek may be the second-best option after Trump for driving the enemies, but it is uncertain if he can maintain relevance in the long term.
- Politics is unpredictable, and other candidates can emerge, so it is too early to predict Vivek's future prospects.
- The rise of internet-native candidates like Vivek is exciting and shows that anyone can come out of nowhere and become a candidate with the right charisma and views.

"Since 2020 Anthropomass exceeded Earth's Biomass" + Engineering with Nature

#394 – Neri Oxman: Biology, Art, and Science of Design & Engineering with Nature

- Lex Fridman discusses the importance of talk therapy and deep conversation.
- He recommends BetterHelp for accessible therapy services.
- He promotes House of Macadamias for high-quality macadamia nut snacks.
- He mentions InsideTracker for tracking biological data.
- He mentions ExpressVPN for private and secure internet browsing.
- Lex Fridman interviews a guest, Neri Oxman, about the intersection between humans and nature.
- They discuss the concept of material ecology and the potential for technology to enhance nature.
- They explore the idea of augmenting nature's wisdom with computational tools.
- They discuss the goal of increasing information and decreasing entropy in nature.
- They consider the potential benefits and challenges of providing nature with extra dimensions through technology.

Designer, engineer, scientist, and artist working on computational design, synthetic biology and digital fabrication, previously at MIT, and now at OXMAN

Neri Oxman

All In on BRICS: Structurally Challenging the West 

E144: Biden targets Elon, BRICS challenges the West, Tiger hit piece & more

- BRICS countries want to have a bigger voice in the world and reduce dependency on the West.
- They are motivated by the weaponization of the US dollar and want economic sovereignty.
- They do not want US involvement in their transactions and want to create independence.
- The BRICS organization has a poor track record of legislative coordination.
- There are growing regional rivalries among the BRICS countries.
- The BRICS countries want to maintain their sovereignty and be independent from the US.
- They have agreed not to sanction each other as part of BRICS.
- BRICS needs to figure out a way to settle transactions and handle surplus.
- The goal of BRICS is not to create a common currency or a reserve currency for the public.
- The US should consider including India in the G7 and focus on including democracies in the group.

Structural Shifts in the Global Financial System: Emerging Bond Risks 

Economic advisor and head of research at the Bank for International Settlements

Hyun Song Shin

Hyun Song Shin on the New Financial Stability Risks

- Tracy Alloway and Joe are discussing the challenges of living with high levels of public debt and high inflation in a global financial system built on bonds.
- They mention the volatility of bond yields and the recent spikes in mortgage rates and long-term yields.
- They talk about the ease of finding guests at Jackson Hole compared to New York.
- They introduce Hyun Song Shin, Economic Advisor and Head of Research at the Bank for International Settlements, as a guest to discuss the changing structure of intermediation since the global financial crisis (GFC).
- Hyun explains that the GFC was focused on credit risk in the banking sector, but now even safe securities can be at the center of financial stress events.
- He mentions the shift from banks to non-bank financial intermediaries (MBFIs) and the importance of infrastructures like central counterparties and exchanges.
- Tracy asks about the different flavors of non-bank investors, and Hyun mentions asset managers, life insurance companies, pension funds, and hedge funds.
- They discuss the lengthening of duration in the bond market and the potential risks associated with interest rate moves.
- They mention the advantages of lengthening duration for mitigating financial stability risks, but also the need to pay for the additional risk.
- Tracy asks about the next safe asset if not bonds, and Hyun mentions that the safest asset is money issued by the central bank, but there are limits and concerns in fragile economies.
- The conversation ends with Tracy and Joe promoting Interactive Brokers' Bond Marketplace and margin loan rates.

Deep Dive into Vivek Ramaswamy: GOP's Breakout Star in 2024?

Michael Barbaro

The Daily

A Breakout Moment for Vivek Ramaswamy

- Vivek Ramaswamy emerged as a rising Republican candidate in the 2024 primaries, surpassing Ron DeSantis.
- Ramaswamy's background includes being a successful biotech entrepreneur and author.
- He gained attention by criticizing woke culture in corporate America and politics.
- Ramaswamy's presidential campaign focuses on addressing the "darkness" in American society and proposing conservative policies such as cutting the federal workforce, attacking Mexican drug cartels, and ending birthright citizenship.

political correspondent for The New York Times

Jonathan Weisman

Adam Posen: Potentially Deep Problems with America's new Industrial Policy 

Former member of the Bank of England's Monetary Policy Committee who now serves as president of the Peterson Institute for International Economics

Adam Posen

Adam Posen Has a Warning on the Danger of Bidenomics

- The conversation at Jackson Hole is focused on how central bankers can understand and respond to the real economy, particularly in relation to supply chain issues and fiscal spending.
- There is a concern about the potential negative outcomes of government involvement in industrial policy, such as creating national champions and corruption opportunities.
- The Biden administration's domestic policy and trade choices have been criticized, particularly in relation to industrial policy and trade rhetoric.
- The importance of ensuring access to green technology and avoiding political loyalty tests in the context of climate progress is highlighted.
- Good industrial policy should prioritize public investment in infrastructure and public goods, and allow for competition, including international competition.
- There is a discussion about the nostalgia for manufacturing and the belief that it is important for a rich country, but it is noted that the benefits of manufacturing were not experienced by everyone and that adaptability and understanding of economic changes are crucial.

Sam Lessin: The End of Factory System for VCs?

GP of Slow Ventures and former VP Product at Facebook

Sam Lessin

"Turpentine VC" | Venture Capital and Investing

E4: Sam Lessin on Why Seed and the “Factory System” Era of Venture is Dead

- Sam Lessin believes that the factory system of venture capital is no longer effective and that the next generation of successful businesses will be different and unique.
- He criticizes the factory system for being lazy and producing mediocre results.
- He suggests that the current backup in the venture capital system will take a long time to resolve and questions if the factory line will ever restart.
- Sam believes that the future of venture capital lies in non-consensus ideas and industries, and that seed capital should be used to build a working business rather than just de-risking for a Series A round.
- He expresses bearishness towards YC and the accelerator model, as well as multi-stage firms like A16Z.

Jackson Hole: How will Central Bankers Confront High Public Debt Levels? 

University of California at Berkeley economist

Barry Eichengreen

Barry Eichengreen on the New Era of High Government Debt

- The Jackson Hole conference is focused on the changing nature of the global economy, but media interest is primarily on whether there will be a rate hike in November or December.
- One emerging theme is the increase in global debt, which has led to credit rating downgrades and volatility in the treasury market.
- There are two dynamics to consider regarding public spending and debt: technical factors and the shift towards fiscal authorities taking a more active role in economic management.
- Central bankers will need to consider the impact of high public debt on monetary policy and inflation going forward.
- Governments borrow for both good reasons (emergency responses) and bad reasons (inability to agree on fiscal balance).
- The Kansas City Fed invites researchers to present papers at the Jackson Hole conference, providing them with topics and allowing them to produce their work.
- Central bankers will face challenges due to limited fiscal response capacity and pressure to be more active in debt management.
- High levels of debt can be a drag on economic growth, but if the debt is used for productive investments, it may be less of a problem.
- Historical analogies show that crises can galvanize action to reduce debt, but broad political consensus is hard to achieve.
- The United States would need a party with a sweeping effort to address the debt issue, but such consensus is unlikely.

Is Google Losing Its Edge? It's AI Plans & Integrations (with PaLM's Paige Bailey) 

E57: Google's PaLM 2 with Paige Bailey

- Copilot extension has the potential to greatly impact developers and generate code and math
- GPT-4 is the most attention-grabbing model, but other models like Palm-2, Claude-2, and Lama-2 are also relevant
- OpenAI is continuously improving models and exploring different approaches
- The pace of advancements in generative AI is accelerating, with surprising and sudden advances in capabilities
- Models are still limited in reasoning abilities and rely on pattern matching and data learned during training
- Researchers encourage exploration and testing of model capabilities to improve and expand their understanding
- Multilingual capabilities have been observed in models like Palm-2
- Distinguishing between reasoning and statistical noise in model outputs can be challenging

Lead Product Manager of Generative Models at Google Deepmind

Paige Bailey

Instacart & Klaviyo: Fewer Funds, Fewer Exits, a Needed Win for VC

NEWS: Instacart and Klaviyo file for IPO, State of VC distributions, AI lawsuits | E1798

- LPs are being more cautious with their investments in venture capital funds, leading to fewer funds being raised and less capital being deployed.
- The lack of venture-backed tech IPOs in recent years has contributed to the decline in the exit market for LPs.
- The regulatory environment and the reluctance of public markets to buy shares in companies are also factors affecting the IPO market.
- The LP market tightening has led to founders having to fight for every dollar and potentially losing LPs.
- Despite the challenges, some funds, like the one being raised by the speaker, are still able to secure meetings with LPs due to their public following and reach.
- The speaker believes that many funds and founders will quit due to the current market conditions, leading to a trimming of the herd.

AI: From Containment to Cambrian Explosion? (with DeepMind Founder Mustafa Suleyman)

#332 — Can We Contain Artificial Intelligence?

- The cost of chips has exponentially declined, leading to mass proliferation and a potential Cambrian explosion of productivity.
- There is a debate about whether AI will lead to massive labor disruption or create new opportunities for human creativity.
- AI has been slowly climbing the ladder of human cognitive abilities and is now capable of tasks previously thought to be unique to humans.
- The concept of containment suggests that technologies should always be accountable to humans and within our control to prevent catastrophic outcomes.
- It is not too late for containment efforts, and open-source development allows for scrutiny and accountability.
- The idea of keeping AI in a box and testing its safety is unrealistic, as inventions quickly proliferate and become open-source.
- The trajectory of exponentially increasing compute and model size suggests that powerful AI capabilities will be widely available in the future, raising questions about accountability and control.

Co-founder of DeepMind and now co-founder and CEO of Inflection AI

Mustafa Suleyman

Runway AI CEO: Beyond LLMs & BS Startup Advice

CEO and co-founder of Runway, the company that trains and builds generative AI models for content creation

Cris Valenzuela

20VC: Why AI Models are not a Moat, Where Does the Value in AI Accrue; Startups or Incumbents, What the World Has Got Wrong About AI, Why AI Needs a New Story and Who is the Right People to Tell it with Cris Valenzuela, Co-Founder & CEO @ Runway

- Runway is a company that trains and builds generative AI models for content creation
- CEO Chris Valenzuela has raised over $285 million for the company from investors like Lux Capital, Felicis, KOTU, Amplify, and NVIDIA
- Runway's customers include Academy-nominated movies, TV shows, media companies, and creatives across industries
- Valenzuela and his co-founders met in art school and started tinkering with AI models, eventually realizing there was something special in building creative tools
- Valenzuela feels like an outsider due to his diverse background in econ, business, art, and programming
- Being an outsider allows him to reason from first principles and merge different domains, leading to the creation of Runway
- Valenzuela is running away from the rigid traditions of his home country, Chile, and seeks to pursue his passions and curiosity
- He fears people not pushing themselves intellectually and creatively, as it leads to mediocrity
- High performance for Valenzuela is defined by a vision of "just figure it out" and the ability to learn and adapt quickly
- Valenzuela approaches learning by being hands-on and active, trying things, reading, and finding brilliant people to learn from
- Runway has learned a lot from their video generation models, including the assumption that UIs matter less and the need to test and uncover the possibilities of this new medium

Many Startups No Longer VC Eligible + GOP Debate

E143: Nvidia smashes earnings, Arm walks the plank, M&A market, Vivek dominates GOP debate & more

- David Friedberg discusses the metrics that make a company eligible for VC funding, emphasizing the importance of year-over-year growth.
- He advises slower-growing companies to consider a PE model and cut their burn to become cash positive.
- Friedberg provides guidance on VC eligibility criteria, including growth rates, gross margins, net dollar retention, CAC payback, and burn multiples.
- He mentions that many startups are currently in the "danger zone" and may need to consider alternative options.
- Friedberg shares his positive experience with a PE firm and how their approach can be beneficial for startups.
- He mentions a profitable company called Stone Algo and expresses interest in investing in similar companies.
- The conversation shifts to discussing the GOP debate, with Nikki Haley, Vivek, and Chris Christie being highlighted as winners.
- Vivek is seen as a potential change agent without the baggage of Trump.
- The Republican Party is noted for having debates and showcasing different factions within the party.

Woody Hoburg

Vivek Ramaswamy: From Biotech CEO to GOP Presidential Candidate

E22: Vivek Ramaswamy on his Presidential Run, Intellectually Oppressive Culture, Otherism

- Vivek Ramaswamy is a Republican presidential candidate and former biotech entrepreneur.
- He believes that people in his hometown in Ohio have a better understanding of the Constitution than many Harvard or Yale Law School graduates.
- He argues that people at MAGA rallies are not necessarily unintelligent and may be more educated on climate change than commonly believed.
- He criticizes the bias in climate science and suggests that talent pool and bias affect the quality of scientific research.
- Vivek Ramaswamy discusses his journey from being a CEO to becoming a presidential candidate.
- He talks about his focus on fighting cultural issues and his belief in the importance of a national identity.
- He emphasizes the need to define what the GOP stands for and the importance of leading the way in defining the agenda.
- Vivek Ramaswamy highlights his commitment to ending race-based affirmative action and abandoning the demands of the climate cult.

Frank Slootman

Howie Liu

20Product: Enterprises are not Adopting AI Yet, When Will AI Break Into Enterprise, What are the Blockers, What Do Enterprises Need from AI & Why Services Companies Will Win in the Next 10 Years of AI Implementation with Howie Liu, Founder & CEO @ Airtabl

- AI should be presented in a specific use case rather than as an abstract concept for better understanding and adoption by enterprises.
- A million-dollar logo is considered a threshold for being a significant enterprise account.
- Enterprises are now focusing on proving ROI, usage, and value of products, leading to a shift in team structures towards customer success.
- Grounding into business outcomes and value delivery is more important than just measuring active to paid ratios in enterprise software.
- Building a business goes beyond finding product-market fit and involves many challenges.
- Thinking on longer timescales and setting milestones along the way can help in building a successful company.
- Valuation is an outcome metric influenced by execution efforts, and the focus should be on driving durable growth in the business.
- The goal is to be remembered as a leader who built a great product and business through efficient and consistent growth.

Enterprise Adoption of AI: A Guide

OpenAI's 3.5 Turbo Fine Tuning: 90% Cheaper, Improved Performance

E56: OpenAI’s GPT3.5 Turbo and the State of AI in Coding, Healthcare, and Education

- OpenAI has released the OpenAI 3.5 Turbo fine tuning model, which is 90% cheaper than their previous offering.
- The previous fine tuning option had limitations in terms of quality and price.
- The new model is based on GPT 3.5 Turbo and offers improved performance and cost savings.
- Other fine tuning options, such as Lama 2, have limitations and issues with false positives.
- OpenAI's fine tuning product is convenient and reliable compared to other options in the market.
- The release of the new model has made fine tuning Lama 2 less necessary for many organizations.
- OpenAI is continuously improving their product offerings, with function calling and GPT-4 fine tuning coming soon.
- Overall, the new fine tuning model from OpenAI is a strong and cost-effective option for customers.

Jakob Uszkoreit: Unlocking New Potential of Transformer Models

Jakob Uszkoreit

AI-Powered Biological Software with Jakob Uszkoreit, CEO of Inceptive

- Jakob Uszkoreit played a key role in developing the idea of attention in deep learning, which was the basis for the Transformer model.
- Language understanding is somewhat hierarchical, and it is not necessary to process a sequential signal from beginning to end to understand it.
- The Transformer model leverages parallel computing to understand pieces of an image and then combine them to disambiguate the overall understanding.
- The success of the Transformer model is due to its efficient implementation on available hardware accelerators.
- There may be other architectures with interesting properties at scale, but the lack of funding and hardware compatibility hinders their exploration.
- The design of accelerators is influenced by the requirements of large-scale transformer architectures.
- Simplified architectures like the Mixer MLP have shown comparable performance to vision transformers, indicating the potential for alternative designs.
- Optimism and hope have played a significant role in the success of the Transformer model, driving experimentation and progress.
- Improving the performance of systems like language models requires addressing the issue of compute allocation, where the amount of compute depends on the length of the prompt rather than the complexity of the problem.
- Elasticity and flexibility in models, such as adjusting compute based on different resolutions or durations of input data, are important research areas.
- Current models lack the ability to elegantly adjust compute based on the specific requirements of the problem, leading to inefficient resource utilization.

Deep Learning Solves Biology: Inceptiv, RNA, & Medicine

- The speaker discusses two problems in biology: lack of complete understanding of all mechanisms and the inability to come up with predictive theories.
- Deep learning is seen as a potential solution to these problems, as it can treat everything as a black box and make useful observations at scale.
- The speaker mentions Inceptiv, a company applying deep learning and other techniques to biology, specifically in the context of designing better RNA molecules for various medicines.
- The potential of RNA, particularly mRNA, in medicine is highlighted, with the expectation of it becoming a major modality in terms of revenue and impact.
- The scalability of RNA manufacturing and distribution is emphasized, with the ability to print any given RNA molecule in the lab.
- The speaker discusses the use of deep learning to generate RNA programs for personalized cancer vaccines and other interventions, highlighting the limitations of screening approaches.
- The focus is on whether understanding and discovery are hindrances in the field of medicine, with an emphasis on empirical evidence of efficacy.
- The historical framing of medicine is mentioned, where many important discoveries were made without a complete understanding of their mechanisms.
- The speaker acknowledges the controversy surrounding the black box approach of deep learning and the challenges of extracting a theory from it.
- The parallels between deep learning and other forms of biology, such as genetic screens, are noted.
- The speaker expresses bullishness on human augmentation in the long term but questions the evolutionary pressures for significantly boosting computational capacity in the brain.

Highlights from the GOP Debate - a dominant Ramaswamy

Political correspondent for The New York Times

Maggie Haberman

A Fiery First Republican Debate — Without Trump

- Donald Trump chose not to participate in the debate because it was not in his interests as the frontrunner in the Republican primary.
- Instead, he did a competing event with Tucker Carlson.
- During the debate, the candidates avoided talking about Trump and focused on issues.
- Vivek Ramaswamy emerged as a dominant figure and defended Trump's agenda.
- Ron DeSantis, who is polling second, received less attention and seemed less interested in standing out.
- The debate eventually turned to Trump and his indictments, with some candidates reluctantly raising their hands when asked if they would still support him if convicted.
- Chris Christie took the opportunity to criticize Trump.

Waymo's Driverless Cars Take Over San Francisco - NYT

Sabrina Tavernise

Technology correspondent for The New York Times

Cade Metz

Ready or Not, Driverless Cars Are Here

- San Francisco has become the capital of driverless cars
- The cars are outfitted with sensors and are easily spotted on the streets
- Getting a driverless car involves using an app and being picked up at a designated spot
- The experience of riding in a driverless car is cautious and slow
- There have been concerns about accidents and the impact on daily life in San Francisco
- The development of driverless cars started in the 60s for military purposes
- Google's founders, Larry Page and Sergey Brin, became interested in driverless cars and started a lab called Google X
- The challenge of driverless cars is dealing with the chaos and unpredictability of the real world
- Google started testing driverless cars on public roads in 2010, sparking a tech arms race
- There was hype and high expectations, but the reality of building the technology was difficult
- A fatal crash involving an Uber test car in 2018 set the industry back and led to a change in approach
- Waymo, a spin-off of Google, focused on building a robo-taxi service in a limited geographic area
- San Francisco became a testing ground for driverless cars, with hundreds of cars driving around the city.

Multimodal vs Specialist Models with Google's Med-PaLM

Research scientist at Google Health. Vivek leads the Google Brain moonshot behind Med-PaLM, Google’s flagship medical large language model, featured in The Economist, The Scientific American, CNBC, and Forbes.

Vivek Natarajan

- The conversation discusses the challenges and possibilities of training generalist AI agents in the biomedical field.
- The team at Google believes in the importance of collaboration and including contributions from various roles, including researchers, engineers, clinicians, and legal experts.
- The MedPAM-M project involved a cross-functional effort and collaboration with authors of previous models.
- The team received help from in-house clinicians to provide insights and craft prompts for tasks.
- The project required a lot of compute resources and the use of the pathways framework, which is an ML orchestration system optimized for training large models on TPUs.
- The JAXX XLA system and TPUs were used for efficient computation and communication in training the models.

Apple Podcasts

Spotify

All Other Apps

Android & Windows

The AI Wrap-Up

Subscribe on your podcast app:

Apple Podcasts

Spotify

All Other Apps

Android & Windows