20VC Who Wins the AI Race; Startups or Incumbents & Does Having Proprietary Data Really Matter For Startups Today

In a dynamic discussion on the future of AI, leading founders and investors, including Emad Mostak of Stability AI, Vince Hanks of Thrive Capital, and Tom Tungus of Redpoint Ventures, debate whether startups or incumbents will dominate the industry. They explore the advantages of incumbents like Google and Microsoft in terms of scale and resources versus the agility and innovative potential of startups. The conversation touches on the critical role of foundational model companies, the importance of data quality, and the necessity for startups to find their niche and execute flawlessly. Jan Lacoon and Sarah Groh weigh in on the potential for an open platform ecosystem for base large language models (LLMs) and the balance of speed versus data advantages in this rapidly evolving field. The episode also highlights the tools and services like Coda, Brex, and AngelList that support startups and venture ecosystems in this competitive landscape.

• Emad Mostaque predicts there will be only five or six foundation model companies globally within three to five years.
• Foundation model companies are those that create the underlying AI models upon which other technologies and applications are built.
• Emad lists potential foundation model companies: Stability AI, Nvidia, Google, Microsoft, OpenAI, Demeta, and possibly Apple.

"I think that there's only going to be five or six foundation model companies in the world in three years. Five years." "I think it's going to be us, Nvidia, Google, Microsoft, OpenAI, and Demeta. And Apple probably are the ones that train these models."

The quotes indicate Emad's forecast about the limited number of key players in the foundation model industry and identify the companies he believes will be leading in this space.

• The podcast aims to explore who is best positioned to succeed in AI: startups or incumbents.
• The discussion revolves around the advantages of startups (speed and innovation) versus incumbents (scale, resources, distribution).
• Six founders and investors provide their insights on the matter.

"AI and the investing world. Who is best positioned to win? Is it startups with speed and innovation, or is it incumbents with scale, resources and distribution?"

The quote frames the central question of the podcast, setting the stage for the discussion about the competitive dynamics between startups and established companies in the AI industry.

• Speaker B endorses Coda as an essential tool for team collaboration and efficiency.
• Coda is highlighted for its ability to centralize data and information, which aids in productivity.
• Brex is promoted as an all-in-one financial stack for founders, emphasizing its global-first mindset and capabilities.

"Coder allows your team to operate on the same information and collaborate in one place by putting data in one centralized location regardless of format, eliminating roadblocks that can really stop your team in their tracks."

This quote emphasizes the utility of Coda as a collaborative tool that centralizes information, which is crucial for team productivity.

• Brex is introduced as a comprehensive financial solution for startups, offering corporate cards, business accounts, and bill pay.
• Brex's global-first approach is underlined, suggesting it is suitable for startups looking to operate internationally.
• The emphasis is on the importance of a proper financial stack for founders who aim to scale their businesses globally.

"With Brex, you get a high limit corporate card, a high yield business account with up to $6 million in FDIC protection and bill pay, all billed with a global first mindset."

The quote outlines the services provided by Brex, highlighting the benefits for startups that require robust financial tools to support their growth and international operations.

• AngelList is described as a key player in the venture ecosystem, providing services that benefit startups and fund managers.
• The platform is praised for reducing friction in cap table management, banking, and fundraising.
• AngelList also caters to large venture funds by offering automated software and customer service.

"The startup world is just buzing about how fast they've been shipping products that meaningfully improve the lives of both startups and fund managers and their investors."

This quote praises AngelList for its rapid product development and the positive impact it has on various stakeholders in the venture ecosystem.

• Emad suggests that incumbents are well-positioned, but acknowledges successful startups reaching billion-dollar valuations.
• He references ITA Software and Kayak as examples of startups that achieved significant value as a layer on top of existing infrastructure.
• Emad points out that innovation isn't the only factor in creating value and establishing moats.

"I think it's incumbents, but there's a lot of startups that be billion dollars. And even on the thin layer thing, ITA software sold for 700 million and kayak sold for 2 billion. And that was a layer on top of ITA."

The quote acknowledges the success of startups in the tech industry while emphasizing that incumbents still hold a strong position due to their existing infrastructure.

• Tom Tungus's analysis is mentioned, comparing infrastructure and application layers in AI, both worth approximately $2 trillion.
• A stark contrast is noted between the number of companies in each layer: three in infrastructure versus fifty in the application layer.
• This results in a significant difference in average enterprise value between the two layers.

"Infrastructure versus application layer and both actually were about $2 trillion town. The difference is in the infrastructure layer there was three companies and in the application layer there was 50."

The quote summarizes Tom Tungus's analysis, highlighting the market value concentration in the AI industry and the disparity in enterprise value between the infrastructure and application layers.

• Emad compares the business models of different AI companies, including Google and Anthropic.
• He points out the vast resources that Google allocates to AI, including DeepMind's salary budget.
• OpenAI's mission to build AGI (Artificial General Intelligence) is mentioned, with a focus on creating a utopia rather than a sustainable business.

"Google spend $20 billion a year on AI. DeepMind salary budget is 1.2 billion a year." "The objective function of OpenAI is to build AGI, and they reckon they need $10 billion to do it."

These quotes reveal the significant financial commitment from companies like Google towards AI development and OpenAI's ambitious goal of building AGI with substantial funding requirements.## AI and Enterprise Transformation

• AI is a transformative force for enterprises, with startups identifying and offering solutions.
• Enterprises are under pressure to develop AI strategies, similar to the urgency once seen for COVID strategies.
• The potential for AI to impact enterprises is massive, with predictions of substantial spending on AI technologies.

"I have a solution and I'm going to start with you. And I might go bigger, but I'm going to help you through this period by doing this and this. And they will appreciate that."

This quote emphasizes the strategic approach a startup can take by offering tailored AI solutions to an enterprise and potentially scaling up later. It highlights the value of starting with a focused, dedicated partnership.

"When this starts going in enterprise, it's going to be a freaking train."

This quote conveys the anticipated rapid and unstoppable adoption of AI within the enterprise sector, likening it to a powerful, fast-moving train.

"My Tam analysis is that 1000 companies will spend 10 million in the next year. 100 companies spend 110 companies spend a billion."

This quote provides a quantitative forecast of enterprise spending on AI, indicating a significant investment from a large number of companies.

• The AI sector is seeing unprecedented investment levels, with companies receiving large funding rounds based on potential rather than proven business models.
• The economic impact of AI could surpass that of COVID, with potential job losses in some sectors balanced by new opportunities in technology and entrepreneurship.

"The amount of capacity versus the amount and whale and wall of money into something that's growing faster than anything we've ever seen is completely mismatched."

This quote discusses the imbalance between the capacity to absorb investment and the immense amount of money flowing into the rapidly growing AI sector.

"This will be a bigger economic impact than Covid."

This quote suggests that the economic repercussions of AI adoption could be more significant than those experienced during the COVID pandemic, though the direction of the impact is uncertain.

• The best AI talent is concentrated within major tech companies like OpenAI, Google, Facebook, Microsoft, and Amazon.
• Despite organizational challenges, these companies have the resources and distribution networks to innovate and integrate AI rapidly into their products.

"The talent is so clustered in these big tech companies."

This quote highlights the concentration of AI expertise within large tech corporations, suggesting they have a competitive advantage due to their access to top talent.

"Microsoft 200,000 person company, they've shipped AI in Bing AI and PowerPoint."

This quote exemplifies how even large companies are quickly incorporating AI into their products, demonstrating their ability to innovate despite their size.

• Startups have traditionally had the advantage of speed and agility, but big companies are now shipping AI products rapidly.
• The challenge for startups is to create unique user experiences that incumbents do not offer, as this is where they can still outcompete.

"I'd bet on the startup ten times out of ten."

This quote expresses confidence in startups that differentiate themselves by offering novel user experiences not yet provided by incumbents.

"What you're describing is a great opportunity for AI native startups to disrupt the incumbent solutions, even if it's not sustainable in the long run."

This quote suggests that while AI native startups have the potential to disrupt incumbents, their advantage may not be sustainable over time.

• There is a spectrum of outcomes for AI integration, from incumbents effectively leveraging their distribution channels to startups winning with flexibility and innovation.
• The pace of AI integration by companies like Microsoft and Adobe is significant, placing startups in an unusual position of being behind in the market.

"The pace with which Microsoft is injecting its products with llms is astounding."

This quote underscores the rapid integration of language models into Microsoft's product suite, highlighting the company's commitment to AI.

"Startups are in this unusual position where they have negative time to launch."

This quote captures the pressure on startups to innovate quickly, as they are competing against incumbents who are already bringing AI products to market.## Execution as the Moat

• The key to success for startups, even against large incumbents, is superior execution.
• Notion and Snowflake are examples of startups that succeeded in markets with strong incumbents by executing well.
• Venture capitalists and startup founders need to believe that great execution can lead to significant wins.
• Foundational model layers require substantial capital due to the costs of training and GPU access.

"Which is better execution is the moat. If you can build a better CRM and get it into market, you can win."

This quote emphasizes that the primary competitive advantage or 'moat' for startups is their ability to execute their business plans effectively.

• The proprietary nature of data models and sets is a concern for startups.
• The rapid evolution of models suggests that today's models may be outdated within a year.
• The need for better data to feed models is highlighted, with an emphasis on moving away from web scraped data to high-quality national data sets.
• Biases in models, such as those seen in Dali 2, indicate the importance of diverse and culturally sensitive data sets.
• There is a call for a six-month pause to prepare for the challenges posed by the rapid development of AI technologies.

"The reality is, no models that are out today will be used in a year."

This quote suggests that the landscape of data models is changing so rapidly that current models are likely to become obsolete in a short time span.

• The conversation shifts from the business side to the research aspect of AI.
• Jan Lacoon is mentioned as a leading expert in AI research.
• There's a belief that the value in AI will accrue to incumbents, but this may not necessarily be the case.
• An open platform for base LLMs (Large Language Models) is proposed as the future infrastructure for AI, similar to TCP/IP or Linux.
• Proprietary approaches to AI are predicted to fall behind due to the collaborative nature required for factual and correct systems.

"There's no one better than this than Jan Lacoon."

This quote introduces Jan Lacoon as an authority in AI research, setting the stage for his perspective on the future of AI development.

• The future AI ecosystem is envisioned as an open platform with an economy of companies building specialized applications on top of it.
• This ecosystem is expected to create jobs rather than eliminate them.
• The proprietary model is expected to fall behind because open models benefit from a larger pool of contributions, ensuring accuracy and correctness.
• The scenario of an open AI platform is compared to Wikipedia in terms of collaborative contributions.

"So the scenario I think will happen, and I'm certainly rooting for, is the scenario I described earlier, where you have some sort of open platform for base LLMs."

This quote describes the speaker's belief that the future of AI will involve an open platform for base LLMs, fostering an ecosystem of specialized applications.

• There is a question about whether the value in AI will go to incumbents or if startups can also succeed.
• The example of OpenAI's success with ChatGPT, despite the presence of larger companies like Google and Meta, is discussed.
• The reason for OpenAI's success is attributed to the lack of pressure on larger companies to innovate and take risks.
• The case of Galactica, a large language model designed to assist scientists, demonstrates the risks and backlash that can occur with new AI products.

"Will the value accrue to the incumbents? Or do you believe that given what you just said about size not being everything in terms of models, it could be startups as well?"

This quote raises the question of whether startups can compete with incumbents in the AI space, considering the dynamic nature of AI model development.## AI Doomers and Risk Analysis

• AI doomers are individuals who are overly concerned about the risks of AI without considering the benefits.
• Scientific publications go through a vetting process, reducing the risk of flooding literature with nonsense.
• The reaction to new AI technologies can vary significantly depending on the company's size and reputation.
• Large companies face greater scrutiny and potential stock impacts when their AI makes mistakes, as seen with Google's Bard.
• There is a misconception that intelligence inherently leads to a desire for domination and uncontrollability.

"AI doomers, people who just did not think about the risk benefit analysis, the risk of flooding the literature with nonsense is ridiculous because scientific publications are vetted and things like that."

This quote addresses the criticism that AI skeptics, or "doomers," often fail to consider the rigorous vetting process that scientific publications undergo, which mitigates the risk of disseminating inaccurate information.

"And then a few months later, Google came out with Bard. And in the demo, Bard made a tiny, minor factual mistake about some astronomical fact, and Google's stock went down by 8%."

This quote illustrates the high stakes for large companies like Google when their AI technology makes public errors, even minor ones, and the significant financial repercussions that can result.

"The companies that have the best technology basically can't have difficulties putting it out because of those legal issues and public image, people are kind of extrapolating."

Here, the speaker points out the paradox where companies with advanced technology face challenges in deployment due to legal concerns and the fear of damaging their public image.

"There is this idea. Somehow the desire and the ability to dominate is linked with intelligence."

The speaker challenges the common belief that intelligence is inherently linked to a desire for domination, which is a significant topic in the debate about AI's impact on society.

• Intelligence does not equate to a desire to dominate or become uncontrollable.
• The desire to influence others is an evolutionary trait in social species, not a byproduct of intelligence.
• Non-social species like orangutans do not exhibit a desire to dominate, separating the concepts of will to dominate and intelligence.
• Superintelligent machines could empower individuals, paralleling having a team of more intelligent staff.

"Even within the human species. It is not the smartest among us that want to dominate the others, to dominate other entities, you don't necessarily need to be smarter than them, but you need to want to dominate them."

This quote emphasizes that the desire to dominate is not necessarily linked to intelligence but rather to the individual's intentions and personality traits.

"The desire to influence others was built into us by evolution, because we are a social species, same as baboons and chimpanzees and wolves and dogs and et cetera."

The speaker explains that the drive to influence others is an evolutionary trait found in social species, suggesting that it is not solely a function of intelligence.

"We need to separate those two concepts, the will, the desire, and the ability to dominate on one hand, and intelligence on the other hand."

By distinguishing between the will to dominate and intelligence, the speaker argues that possessing intelligence does not automatically lead to a desire for control.

"The fact that we're going to have superintelligent machines at our disposal means that every one of us is going to be like a business leader, a politician, or an academic with a staff of people working for them that are more intelligent than themselves."

The speaker suggests that superintelligent machines could serve as highly capable assistants, enhancing our capabilities rather than posing a threat to control.

• Startups' primary advantage is speed, which is crucial in a rapidly changing environment.
• Large organizations struggle to match the agility of startups.
• Data moats are less significant as entrepreneurs find creative ways to collect and generate data.
• The competition between startups and incumbents is not clear-cut; both have unique challenges and advantages.

"Classically, the only real advantage startups have is speed."

This quote highlights the traditional view that startups' main competitive edge is their ability to move quickly in the market.

"What's the quote? Some decades nothing happens, and some years a decade happens. I feel like that is happening right now."

The speaker reflects on the current pace of change, implying that rapid developments are giving startups an opportunity to leverage their agility.

"On the incumbent advantage side, much ado has been made about this idea of a data moat, but honestly, there's a lot of data out there, and entrepreneurs are incredibly creative about collecting it and increasingly about generating it."

Here, the speaker challenges the notion that incumbents have an insurmountable advantage due to their data reserves, pointing out that startups are finding innovative ways to access and create data.

• Coder provides a unified workspace for teams to collaborate and manage content for projects.
• Brex offers a financial stack with global capabilities, essential for startups looking to scale.
• AngelList is a comprehensive platform for cap table management, banking, and fundraising for startups and venture funds.

"Coder allows your team to operate on the same information and collaborate in one place by putting data in one centralized location regardless of format, eliminating roadblocks that can really stop your team in their tracks."

This quote explains how Coder helps teams by centralizing data and collaboration, which enhances productivity and project management.

"With Brex, you get a high limit corporate card, a high yield business account with up to $6 million in FDIC protection and bill pay, all billed with a global first mindset."

The speaker describes the financial services provided by Brex, emphasizing its high limits and international focus, which are beneficial for growing startups.

"Thousands of startups have moved their camp tables to Angellist in the past year."

The quote indicates the popularity and trust that startups place in AngelList for managing various aspects of their business operations.