The Rise of AI: New Theories and Insights

The Rise of AI: New Theories and Insights

Just as the internet revolutionized the world by pushing distribution costs to zero, AI is poised to transform industries by driving creation costs towards zero. The economic value generated by AI will not be evenly distributed along the value chain. Instead, it will result in rapid consolidation and power law outcomes among infrastructure players and endpoint applications. In this article, we will explore six new theories about AI and its potential impact on various sectors.

1. Fine-tuned models win battles, foundational models win wars

In traditional markets, the different components of a value chain are often distinct and separate. However, in the AI landscape, each building block is interconnected. This is particularly evident in the competition between foundational models and fine-tuned models. Foundational models aim to excel in broad tasks, while fine-tuned models are tailored for specific use cases.

While fine-tuned models may not always outperform foundational models, they have the advantage of being cost-effective for narrow use cases. Over time, fine-tuned models can gradually improve their performance, while foundational models experience step-changes. The ability to fine-tune models for specific purposes lowers the cost of prompt completion and enables more efficient AI applications.

2. Long-term model differentiation comes from data-generating use cases

Data loops play a crucial role in AI development. These loops occur when an AI provider incorporates feedback mechanisms into their product and utilizes that feedback to retrain the model. To fully leverage data loops, AI providers may need to own the endpoint solution. Startups that can capture the output of a model to retrain it have the potential to build specialized winners.

The hierarchy of model types can be ranked as follows: small, specialized models; large general models; and large specialized models. Companies that can harness the power of data-generating use cases and establish feedback loops will have a competitive edge in the AI landscape.

3. Open source makes AI startups into consulting shops, not SaaS companies

The availability of open-source AI models has resulted in a shift for AI startups. Instead of operating as software-as-a-service (SaaS) companies, many have transformed into consulting shops. Open-source AI projects, such as DALL-E 2, have led to rapidly eroding market power for proprietary models.

To combat downward pricing pressure, model providers selling access to their models via API often need to compromise on pricing. However, some companies have found alternative solutions. For example, OpenAI has created a venture fund and taken equity stakes in promising startups to navigate pricing challenges.

4. Most endpoints compete on go-to-market strategies, not AI

In the AI services market, the purchasing decision is often driven by go-to-market (GTM) strategies rather than a direct comparison of AI capabilities. Companies that sell AI services either need to fully own fine-tuned models or compete with the attributes of a typical SaaS startup.

For startups competing on the basis of SaaS, the advantage lies with companies that already possess inherent distribution or product capabilities. The integration of generative AI into existing products is expected to become a norm for major software providers in the coming months.

5. AI will amplify existing power law dynamics in the creator economy

The creator economy, which is already characterized by a concentration of revenue among a small percentage of creators, will be further amplified by AI. With the democratization of content creation, AI tools will play a crucial role in enabling creators to produce better content at a faster pace.

While content creation will become more accessible, distribution will remain the key determinant of success. Creators who effectively utilize AI tools to enhance their content and build a critical mass of fans will thrive. The winner-takes-all dynamic will be further exaggerated in the digital media landscape.

6. Invisible AI: The most valuable deployment of AI

The true power of AI lies in its ability to seamlessly integrate into various industries without drawing attention to itself. Invisible AI refers to companies that are powered by AI but never explicitly mention it. AI products that enable new modalities of digital interactions and transform existing use cases have the greatest potential for success.

The combination of search capabilities with generative AI opens up exciting possibilities for companies across different sectors. The ability to leverage AI technologies in a way that enhances user experiences and drives innovation will be a key differentiator in the future.

Actionable Advice:

1. Embrace fine-tuned models: Consider the potential benefits of deploying fine-tuned AI models for specific use cases within your organization. By tailoring models to your unique requirements, you can improve efficiency and reduce costs.

2. Establish feedback loops: If you are an AI provider, explore ways to incorporate feedback mechanisms into your product and utilize them to retrain your models. This will enable continuous improvement and give you a competitive advantage in the market.

3. Focus on distribution and go-to-market strategies: In the increasingly crowded AI landscape, differentiation lies not only in the capabilities of your AI but also in your ability to effectively reach your target audience. Invest in robust distribution channels and develop strong go-to-market strategies to maximize your chances of success.

In conclusion, the rise of AI brings both challenges and opportunities for businesses across industries. By understanding the dynamics of fine-tuned models, data loops, go-to-market strategies, and the amplification of power law dynamics, companies can navigate the AI landscape more effectively. Embracing invisible AI and leveraging its capabilities to transform existing use cases will be a key driver of success in the future.