Generative AI: A Creative New World

Abstraction: Sequoia's 2022 thesis on generative AI market opportunity and waves

Key points:

• Identifies four waves: small analytical models (pre-2015), transformer scaling race (2015-2022), cheaper/faster diffusion and LLMs (2022+), and killer-app emergence (now)
• Claims generative AI could bring marginal cost of knowledge work "toward zero," creating trillions in economic value across billions of workers
• GitHub Copilot is cited as generating ~40% of code in projects where installed; code generation is seen as the highest near-term impact domain
• App companies win via a data flywheel: user engagement → model fine-tuning → better performance → more users; vertical-specific tuning cited as key differentiator
• Risks flagged: unresolved copyright, trust/safety, and cost questions; first wave apps are "gimmicky and thin" like early iPhone apps
• Co-written with GPT-3; illustrations generated with Midjourney — an early demonstration of the thesis

Connections: Sequoia Capital · Openai · Github Copilot · Generative AI · Large Language Models · Diffusion Models

Source: https://www.sequoiacap.com/article/generative-ai-a-creative-new-world/