20VC: Benchmark's Sarah Tavel on Are Foundation Models Commoditising | Why Frontier Models Will Be Closed Source | Why the Value is in the Application Layer | The Future of AI is "Selling the Work" Not the Tools
May 6, 2024
auto_awesome
Sarah Tavel from Benchmark discusses the future of foundation models, application layers, and AI. She talks about the shift towards selling work outcomes facilitated by AI, challenges in AI training costs, and the importance of strong partnerships in venture capital investments.
The future of AI lies in closed-source frontier models to protect technology and drive innovation.
Value creation in AI is concentrated in the application layer through continuous user-focused improvements.
Benchmark's partnership model prioritizes direct involvement with founders, fostering deep engagement and alignment for success.
Deep dives
The Increasing Cost of Training AI Models
Training advanced AI models is becoming increasingly expensive due to the need for significant investment in compute power. Each successive model requires more resources to train, leading to a potential oligopoly scenario. Despite expected cost reductions in underlying hardware, the demand for compute continues to rise, resulting in a costly race for advancements in AI technology.
Application Layer vs. Infrastructure Layer in AI
The discussion delves into the value generated at the application layer in contrast to the infrastructure layer in AI technology. Emphasis is placed on the belief that the application layer drives significant value creation by owning the end user relationship, allowing for continuous value delivery. The differentiation lies in the ability to capture and create value for users over time, with a focus on offering enduring value propositions.
Balancing Open vs. Closed Source Models in AI
The podcast explores the debate between open and closed source models in AI development. While the trend suggests a move towards closed source for cutting-edge models, there are ongoing considerations around the long-term value of open sourcing innovations. The decision to keep models close source is driven by the need to protect frontier technologies, but emerging approaches like Meta's llama project could potentially disrupt this trend.
Partnership Model and Differentiation in Investing
The episode delves into how Benchmark's partnership model sets it apart in investing. Unlike some firms that delegate aspects like recruiting, Benchmark prioritizes direct involvement with founders. This unique approach creates a deep partnership dynamic, where partners are highly engaged in every aspect of the investment process, ensuring alignment and dedication to the founder's success.
Investment Strategies and Decision Making
Sarah discusses the importance of conviction in investing and the strategy behind when to pay up for a deal. The focus is on supporting founders to build their best companies, not just being a cheerleader but asking critical questions. This approach emphasizes building trust and strong relationships with founders, reflecting a commitment to long-term support and success in their investments.
Sarah Tavel is a General Partner @ Benchmark, one of the most successful and renowned venture firms in the world. At Benchmark, Sarah has led rounds in Chainalysis, Hipcamp, Medely, Rekki, Glide, Cambly and more. Prior to Benchmark, Sarah was a Partner at Greylock Partners. Before Greylock, Sarah was the first 30 employees at Pinterest. Sarah joined Pinterest in 2012 after co-leading the Series A investment while at Bessemer Venture Partners.
In Today's Episode with Sarah Tavel We Discuss:
1. Becoming a GP at The Most Renowned Firm in Venture:
How did the process of Sarah joining Benchmark start? How did it progress? What was it that convinced her to leave Greylock and join Benchmark?
What does Sarah believe makes Peter Fenton the world-class investor that he is?
What does Sarah know now that she wishes she had known when she started in venture?
2. Foundation Models: Is it All Going to Zero:
Will foundation models be commoditised?
Will 99% of the funding going to foundation models go to 0?
How does Sarah view the future of open vs closed source?
Why does Sarah believe that all frontier models of the future will be closed-source?
Why does the business model of foundation models remind Sarah of the food delivery business?
3. Application Layer: Where $BN Companies Will Be Built:
Why does Sarah believe that sustainable value-creating companies will be in the application layer?
How does Sarah determine between a wrapper on top of ChatGPT and true product value?
Are enterprises opening real budgets for AI today or are we still in experimental budgets?
How does Sarah think about how AI companies differentiate when there are so many in the same space of customer service, sales team support etc etc?
Why does Sarah believe that it is rational to pay more for these companies when investing in them?
What does Sarah mean when she says the future is "selling the work and not the tools"?
4. Inside Benchmark: How the Best Do Venture:
What is the one rule that Benchmark is willing to break when doing a deal?
Why do Benchmark aim to be the best recruitment firm in the world?
Why do Benchmark not agree with the concept of reserves?
In a case where Benchmark have lost, why did they lose? How did they change their approach?
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode