PodcastsThe Twenty Minute VC (20VC): Venture Capital | Startup Funding | The Pitch20VC: AI's Biggest Questions: The Commoditisation of LLMs, Open vs Closed: Who Wins, Model Size vs Data Quality, Why Google are Vulnerable and Apple are the Dark Horse
20VC: AI's Biggest Questions: The Commoditisation of LLMs, Open vs Closed: Who Wins, Model Size vs Data Quality, Why Google are Vulnerable and Apple are the Dark Horse
The Twenty Minute VC (20VC): Venture Capital | Startup Funding | The PitchFri Nov 24 2023
Foundational Models and Commoditization:
- Emad from Stability predicts that in the next few years, there will only be five or six foundational model companies globally, including Nvidia, Google, Microsoft OpenAI, Meta, and Apple.
- Dez from Intercom emphasizes that while commoditization hasn't happened yet, they rigorously test all foundational models for quality of conversation and trustworthiness. The comparison between OpenAI and alternative providers revealed a work in progress with constant testing of new providers.
- Jeff Seibert from Digits believes that despite current market dominance by OpenAI, the open source equivalent is inevitable due to market forces pushing towards commoditization.
Importance of Model Size vs. Data Quality:
- Imad highlights the continuous improvement in model size, indicating a shift from 540 billion parameters to 14 over time. He also stresses the need for personalized data sets to interact with base models for customization based on individual stories.
- Yann LeCun argues that large models are not necessary for excellent performance and efficiency in training has improved significantly. He foresees smaller AI systems becoming even more efficient in the future.
Open vs. Closed Ecosystems:
- Yann LeCun advocates for an open ecosystem as it allows recruitment of global intelligence to contribute diverse ideas and innovations which may not be feasible within closed ecosystems.
- Douwe Kiela expresses skepticism about relying solely on open source solutions due to deep understanding and economies of scale possessed by certain organizations like OpenAI.
Value Accrual: Infrastructure vs. Application Layer:
- Tom Tunguz's analysis suggests that value concentration occurs equally in both infrastructure (AWS, GCP, Azure) and application layers (Netflix, ServiceNow), but there are higher success odds at the application layer due to greater diversity of needs.
- Des Traynor anticipates gradual thinning margins and intense price competition among major infrastructure providers such as GCP, AWS, open AI directly, and Azure.
Pricing Model & Business Model Evolution:
- Myles Grimshaw envisions a paradigm shift where businesses sell service level agreements (SLAs) based on work performance rather than traditional uptime guarantees. This signifies a move from co-pilot UX for workers to control center UX for managers.
Apple's Position in AI:
- Speculation on Siri's potential improvement through enhanced conversational abilities and increased usefulness driven by advancements in AI technology.
- Emphasis on Apple's focus on privacy and its capability to run large language models (LLMs) on devices, potentially surpassing open AI systems.
Google's Future Challenges:
- Analysis of Google's vulnerability due to the potential threat posed by chat-based search technology challenging its dominant position in the market.
- Highlighting the necessity for Google to disrupt itself and innovate to avoid a decline in relevance and market share amidst evolving technological landscapes.
Amazon's Strategic Moves:
- Evaluation of Amazon as an engineering-focused organization capable of swift adaptation and innovation while facing challenges in transitioning from research into practical applications.
- Discussion on the strategic implications of potential acquisitions and technological integrations for Amazon, particularly within the EC2 cluster.
Societal Impact of AI:
- Addressing concerns about job displacement and societal impact attributed to AI advancements.
- Expressing optimism about AI bringing forth a new renaissance for humanity through amplified intelligence and creativity while acknowledging the need for political and social changes to ensure equitable distribution of benefits.