Foundational AI models are a general technology like electricity or the printing press

Entrepreneurial possibilities lie in using these technologies to enable new businesses

Foundational AI models are a kind of general technology, akin to electricity or the printing press. These are large language models like GPT-4, Claude, etc. Most of the developments here will either remain in large legacy tech companies or open source initiatives. These models all require vast amounts of data, and very expensive training to function properly. While it is true that there has been research into smaller models, these models require very high quality data, and still require significant training expenses. Unless data quality requirements and training costs decline significantly, I don’t see foundational models being a fruitful area for entrepreneurs and venture capitalists.

Foundational AI models are broadly applicable. They can be used to enhance almost any human activity. The most significant entrepreneurial opportunities in AI lie not in the creation of alternative or superior foundational models, but in leveraging these existing technologies to enable new business models and ventures that were previously unfeasible without advanced AI capabilities.

Historical Parallels: Electricity and the Printing Press

When electricity and the printing press were first developed, no one could foresee the thousands of enterprises that arose as a result of their invention. However, over time, electricity and the printing press fundamentally reshaped industries, economies, and societies. Electricity did not just bring light; it revolutionized manufacturing, transportation, and domestic life. Similarly, the printing press not only made books more accessible but also democratized knowledge, fostering an era of englightenment and scientific revolution. I think that artificial intelligence will be similarly transformative: it will allow for the creation of tens of thousands of new businesses that were not previously possible.

Like electricity and the printing press, foundational AI models such as GPT-3 and GPT-4 are general technologies with wide-ranging applications. They are not confined to a single domain but have the potential to impact numerous industries and aspects of human life. These models, with their ability to understand and generate human-like text, can perform tasks ranging from writing and translation to more complex problem-solving and creative endeavors.

Given the broad applicability and advanced capabilities of these AI models, the primary entrepreneurial opportunity lies in application, not creation. I see three main opportunities that foundational AI models give rise to:

1. New Business Models: Foundational AI models can power new types of businesses that were previously impossible. For example, they can enable automated, personalized content creation at scale, revolutionize customer service with intelligence virtual assistants, or enhance decision-making in complex scenarios like financial analysis or medical diagnostics.

2. Traditional Industry Transformation: These AI models can also transform traditional industries. AI-driven analysis could lead to better healthcare diagnostic tools. Personalized learning experiences can be created at scale. In entertainment, AI can be used to generate novel content, from music to interactive storytelling.

3. Democratization of Expertise: Foundational AI models democratize access to skills and knowledge. They can provide expert-level guidance in areas like legal advice, business consulting, or technical support, making these services more accessible and affordable.