August 6, 2024 • 1hr 13min

Sarah Guo - The Power of Conviction - [Invest Like the Best, EP.383]

Invest Like the Best with Patrick O'Shaughnessy

In this episode of Invest Like the Best, Patrick O'Shaughnessy interviews Sarah Guo, founder and CEO of Conviction, an early-stage venture capital firm focused on AI companies. Sarah previously spent 9 years as a partner at Greylock before starting Conviction in 2022. The conversation covers Sarah's perspective on the AI landscape, her approach to evaluating AI startups, and her predictions for the future of the technology.

0:00

-0:00

Key Takeaways

AI is the most important technological advancement of our lifetime, with potential for massive productivity gains and economic value creation
Application companies need to avoid building things that foundation models will soon replicate, and instead focus on the "last mile" of bringing AI capabilities to specific use cases and workflows
Efficiency and cost management will become increasingly important as AI models and infrastructure mature, contrary to the current narrative that "efficiency doesn't matter"
There are still many "uncanny valleys" to cross in AI capabilities, particularly around multi-step reasoning, hallucination management, and domain-specific applications
The AI ecosystem is likely to remain competitive, with multiple foundation model providers and opportunities for startups to create value on top of models
Key challenges for AI adoption in enterprises include change management, risk/compliance concerns, and achieving "minimum viable quality" for specific use cases
Areas ripe for AI disruption include enterprise software configuration/maintenance, material science, healthcare operations, and government services

Introduction

In this episode of Invest Like the Best, Patrick O'Shaughnessy interviews Sarah Guo, founder and CEO of Conviction, an early-stage venture capital firm focused on AI companies. Sarah previously spent 9 years as a partner at Greylock before starting Conviction in 2022. The conversation covers Sarah's perspective on the AI landscape, her approach to evaluating AI startups, and her predictions for the future of the technology.

Topics Discussed

Recruiting and Key Traits for Early-Stage VCs (2:39)

Sarah discusses the challenges of recruiting for early-stage venture capital roles:

Small sample sizes make it difficult to predict success based on paper qualifications
Look for baseline understanding of technology, competitiveness, team orientation, and judgment
Early career judgment is hard to assess, so look at understanding of businesses and product thinking as proxies

Lessons from Early Investments (4:02)

Sarah shares two formative early investments:

Awake Security - Incubated at Greylock, taught her about company building from scratch
Figma - Showed her what an inexperienced but great entrepreneur looks like early on

She emphasizes the value of patience and focusing on substance early in a VC career rather than rushing to "level up".

The Decision to Start Conviction (7:04)

Sarah explains her motivation for starting Conviction:

Recognized the massive potential of AI after years of following the field
Saw an opportunity too big to ignore as foundation models became increasingly capable
Wanted to focus purely on early-stage AI investing and be an entrepreneur herself

Launching Conviction and Initial Steps (9:41)

Sarah describes the early days of starting Conviction:

Managed the transition from Greylock responsibly
Started talking to LPs and raising a fund
Developed the firm's strategy and thesis
Launched just before ChatGPT, validating the timing

She advises others considering starting a fund: "If anybody who is really convinced that they should do this, but they're like, oh, it'll be better if I just optimize like another month or three months or something, no, don't do that. The market is moving, it's always going to be hard."

First Investment at Conviction and Evaluating AI Application Companies (11:39)

Sarah discusses Conviction's first investment, Harvey AI, and her approach to evaluating AI application companies:

Look for the same fundamentals as other software companies: distribution, quality of people, size of opportunity
Be wary of popular narratives like "everything is GPT wrappers"
Focus on the value of workflows, customer relationships, specific data, and change management
Avoid building things foundation models will soon replicate
Look for teams with research backgrounds who deeply understand the models

Challenges and Opportunities in AI Applications (16:02)

Sarah elaborates on the challenges and opportunities in building AI applications:

Conservative industries: Selling probabilistic AI to risk-aware, non-technical audiences (e.g. legal industry)
Economic benefits: Demonstrating massive productivity gains (e.g. Harvey AI automating junior lawyer tasks)
New capabilities: Enabling tasks that were previously infeasible at scale (e.g. analyzing 20,000 contracts)

Minimum Viable Quality in AI Products (21:47)

Sarah introduces the concept of "minimum viable quality" for AI products:

Defining the quality threshold where an AI product becomes useful for a specific use case
Example: Heygen reaching the quality bar for commercial video generation
Importance of creative product design to improve perceived quality (e.g. verification, ranking of model outputs)

She notes: "Trying to define within a company what's good enough quality as the customer experiences it is like this really important new thing as a moving target."

Future of AI and Frontier Models (31:10)

Sarah shares her hopes for the next generation of AI models:

Better calibration: Models that can assess their own confidence
Hallucination management: Reducing false or inconsistent outputs
Multi-step reasoning: Improving logical thinking capabilities

She emphasizes that improvements in these areas could unlock significant enterprise adoption by addressing key concerns around reliability and risk.

The Unpredictable Future of AI (32:26)

Sarah cautions against overconfident predictions about AI's future:

The field is evolving rapidly and unpredictably
Even large labs have been surprised by developments like the success of open-source models

She hopes for a diverse ecosystem with "a million flowers blooming" rather than dominance by a few players.

The Importance of Efficiency in AI Models (39:09)

Sarah discusses the growing importance of efficiency in AI:

Contrary to some narratives, efficiency has always mattered in computing
As models scale, power and infrastructure constraints become more significant
Efficiency innovations will be crucial for continued progress

She notes: "There is no such thing as cheap enough, because we've been working a compute for a long time as an industry, and it is not so cheap that nobody cares because we just keep doing more with it."

The Business of AI: Costs and Margins (42:40)

Sarah addresses concerns about the high costs and potentially lower margins of AI businesses compared to traditional software:

The infrastructure ecosystem for AI is still immature, with room for significant efficiency gains
Expects improvements in hardware utilization, chip design, and systems management
Doesn't believe AI applications are structurally less profitable than traditional software in the long run

Infrastructure and Hardware Challenges (46:29)

Sarah explains some of the challenges in developing new AI hardware:

Difficulty in testing large-scale training workloads without access to massive clusters
High barriers to entry for new chip designs due to the need for real-world validation at scale
Advantages of incumbents like Google who can test on their own large workloads

The Competitive Landscape of AI Chips (48:54)

Sarah discusses the competitive landscape for AI chips:

NVIDIA's strong position with CUDA and years of optimization
Efforts by AMD, cloud providers, and startups to challenge NVIDIA
Structural advantages for incumbents, but potential for innovation given the massive opportunity

The Future of AI and Society (50:26)

Sarah shares her perspective on AI's societal impact:

Potential for significant productivity gains and "abundance"
Concerns about job displacement and societal adaptation to rapid change
Transformative effects on learning, creativity, and access to knowledge

Opportunities and Innovations in AI (54:17)

Sarah highlights two areas of opportunity in AI:

Enterprise software configuration and maintenance: Potential to disrupt billions spent annually on consulting services
Material science: Applying foundation models to accelerate discovery and innovation

She emphasizes the potential for AI to tackle inefficiencies in large, complex domains like healthcare and government services.

Concerns and Ethical Considerations (59:41)

Sarah addresses concerns about AI:

Believes existential risks are a distraction from more immediate concerns
Focuses on near-term abuses like misinformation and fraud
Emphasizes the need for solutions to address the amplification of existing problems

Debates and Research in AI (1:01:04)

Sarah discusses interesting debates in AI research:

How to improve multi-step reasoning in a general way
The role of math and computer science in advancing AI capabilities
Challenges in evaluating increasingly capable models that surpass human experts

Personal Reflections and Closing Thoughts (1:06:38)

Sarah shares her motivation for working in AI:

Believes AI is the most important technological change in our lifetimes
Excited by the potential for productivity gains and new economic platforms
Driven by the opportunity to work with and support the most interesting and motivated people

She reflects on key people who have supported her career, including her husband, Anil Busri (former Greylock partner), and Aseem Chandna (former Greylock partner).

Conclusion

Sarah Guo provides a comprehensive and nuanced perspective on the current state and future potential of artificial intelligence. She emphasizes the massive opportunity AI presents while acknowledging the challenges in development, adoption, and societal impact. Her insights on evaluating AI startups, the importance of efficiency, and the evolving competitive landscape offer valuable guidance for investors and entrepreneurs in this rapidly changing field. Sarah's passion for the transformative power of AI and her commitment to supporting talented founders shine through, highlighting why she's a respected voice in the AI venture capital ecosystem.