20VC: Why OpenAI and Anthropic Won't Win the App Layer | Why...

Full Title

20VC: Why OpenAI and Anthropic Won't Win the App Layer | Why Teams Will Get Bigger Not Smaller in a World of AI | Why AI Removes Incumbents Advantage of Bundling | China vs America: Who Wins the AI War with Arvind Jain, Co-Founder @ Glean

Summary

Arvind Jain, co-founder of Glean, discusses the evolving AI landscape, emphasizing that while frontier model providers are crucial assets, the future of AI lies in enterprise-grade solutions that integrate with existing workflows. The conversation highlights the challenges and opportunities in the AI market, from cost control and data privacy to the competitive dynamics between large tech companies and startups.

Key Points

• Enterprise adoption of AI is hindered by the lack of integration into existing workflows, not by AI's capabilities themselves.
• Frontier model providers like OpenAI and Anthropic should be viewed as assets by other AI companies, not direct competitors, as they enable the development of broader applications.
• Enterprises are increasingly concerned about technological dependence on frontier model providers and the potential for their core IP and data to be compromised.
• The rising cost of AI is accelerating the adoption of open-source models, driven by a desire for cost control and data ownership.
• While open-source models are improving, concerns remain about the use of Chinese-developed models due to potential security risks and lack of trust.
• The bundling strategy of large tech companies like Microsoft, while formidable, may be challenged by the shift towards consumption-based AI models.
• True AI ROI is ultimately a throughput problem, requiring proper contextualization and efficient assembly of information for AI agents to perform effectively.
• The goal of AI should be to augment human capabilities, not to replace individuals entirely, as AI currently lacks the nuanced understanding and intangible qualities of human decision-making.
• The startup ecosystem is seeing an overabundance of capital, which can lead to unsustainable spending habits and hinder the development of truly robust businesses.
• The future of work will involve more composite roles that blend specialized skills, leading to smaller, more efficient teams.
• Data analyst and some HR roles are likely to be significantly impacted by AI, while business owners will gain direct access to data insights.
• The rise of open-source models, particularly from China, presents a competitive challenge to US-based AI companies, sparking discussions about technological sovereignty.

Conclusion

The AI landscape is rapidly evolving, with a critical need for companies to focus on integrating AI into existing workflows for true value realization.

Open-source models present a significant opportunity for cost control and innovation, but concerns around data privacy and geopolitical influences remain crucial considerations.

The future of work will likely involve a shift towards more agile, composite roles, necessitating a re-evaluation of team structures and talent acquisition strategies.

Discussion Topics

• How can enterprises effectively measure the ROI of AI investments beyond simple productivity gains?
• What are the long-term implications of the increasing reliance on AI for data analysis and decision-making within businesses?
• In a world of increasingly powerful AI models, what is the future of human creativity and specialized skill sets in the workforce?

Key Terms

Frontier Model Providers

Companies developing and training the most advanced and powerful large language models (LLMs).

App Layer

The application or software interface that users interact with, as opposed to the underlying model or infrastructure.

Incumbents

Established companies in a market that are challenged by new entrants or technologies.

Bundling Advantage

The benefit a company gains by offering a package of products or services, often creating customer stickiness and reducing perceived cost.

Open Source Models

AI models whose source code is publicly available, allowing for free use, modification, and distribution.

Token Spend

The cost incurred when using large language models, typically measured by the number of "tokens" processed.

Inferencing Workload

The computational process of using a trained AI model to make predictions or generate outputs.

On-Prem

On-premises, referring to software or hardware that is physically located within an organization's own data center.

Vendor Management Problem

The complexities and overhead involved in managing relationships and contracts with multiple technology providers.

Land Grab

A situation where multiple companies are aggressively trying to capture market share in a new and rapidly developing industry.

Composite Roles

Job roles that combine multiple specialized functions, requiring individuals to have a broader skill set.

Sovereignty

The ability of a nation or entity to have independent control over its own affairs, in this context, related to AI models and data.

Regulatory Capture

A situation where a regulatory agency, created to act in the public interest, instead advances the commercial or political concerns of special interest groups.

Timeline