Interview, Fireside Chat
Sarah Guo: On Her New $101M Fund; How AI Impacts Inequality; AI Startups vs Incumbents | E1007
- Malicious code generation by nation states and hackers is predicted as an inevitable consequence of AI capabilities, requiring immediate attention and investment in defensive measures over the near term.
- Regulatory bodies are expected to face a widening technical knowledge chasm within the next two years, creating a risk of ineffective regulation despite society's likely resistance to halting productivity-enhancing technologies.
- The technology industry is forecast to produce abundance similar to previous revolutions, though initial wealth distribution is anticipated to be highly unequal before broad adoption occurs.
- The vast majority of companies will need to apply existing models rather than train new ones, with fewer than ten instances deemed viable for training from scratch, while the industry shifts toward leveraging APIs, open source models, and fine-tuning.
- The Conviction fund plans to focus on capital-efficient seed and Series A investments ranging from one million to ten million dollars, anticipating that the bulk of opportunities lie in companies avoiding hundreds of millions in upfront capital.
- Success metrics for AI adoption are predicted to shift from auto-complete statistics to a model where humans perform planning and AI executes iterative changes via natural language instructions within five years.
- The market is expected to move toward extreme unbundling and micro-serviced offerings driven by personalization, a trend where large companies will struggle to tailor to divergent user preferences.
- Startups are expected to gain an advantage over incumbents through speed in a "warp speed" environment, as data moats become less significant barriers due to creative data collection methods.
- Microsoft is currently viewed as the incumbent best navigating the AI transition, with competitors like Google, Amazon, and Apple expected to make significant investments to catch up, particularly as the latter two currently lack cutting-edge labs.
- Future software development is anticipated to be led by general product engineers rather than specialized AI researchers, significantly expanding the number of people capable of leveraging these models.
- The market will likely digest the era of over-capitalized "fat startups" over a "gloomy" period, requiring many companies to improve efficiency to become durable businesses.
- Significant AI opportunities are identified in services markets, such as legal, where replacing or enabling labor is predicted to generate a larger economic impact than direct software sales.
- Subscale seed funds without differentiated strategies are expected to struggle to persist as Limited Partners become more selective due to the current turn in the market cycle.
- Early-stage company outcomes are described as unknowable, with market structure determined by actor agency rather than static structural advantages, rendering traditional reserves or "complete bullshit" as a metric.
- Success for Conviction is defined over a twenty-year horizon by achieving best-in-class venture multiples, building a partnership with important entrepreneurs, and nudging AI use toward productivity and alignment.