The actual work is not taking notes, but using them and building experience. While taking notes helps us accumulate knowledge, it is not enough if we don't actively engage with the information. We need to rephrase and self-explain what we consume, understand it deeply, and make connections to what we already know. Only through this balance of building knowledge and experience can we truly start learning.

The actual work is not taking notes, but using them and building experience. While taking notes helps us accumulate knowledge, it is not enough if we don't actively engage with the information. We need to rephrase and self-explain what we consume, understand it deeply, and make connections to what we already know. Only through this balance of building knowledge and experience can we truly start learning.

To facilitate thinking and playing with our notes, we need a space where we can freely explore and muddle through them without judgment. One approach is to use a note-taking app like Obsidian, which provides two spaces for this purpose: vague-notes and thinking-notes. Vague-notes are more random and allow us to write whatever comes to mind regarding a particular note. Thinking-notes, on the other hand, are more specific and focused, where we can shape our thoughts, pull up related notes, and create links.

Building knowledge and experience through note-taking is valuable, but it's also important to consider the bigger picture of the AI value chain. When it comes to AI threats, there are two types: the doomsday scenario where a super-intelligent AI wipes out humanity, and the scenario where a small group of people make significant profits from AI without reaching artificial general intelligence.

In the AI value chain, there are several key components. The compute layer provides the raw power required to run AI algorithms, but it's not as simple as running them on a specific type of chip. AI algorithms often require running multiple GPUs simultaneously. Data is another crucial component, as AI models are trained on datasets. Previously, labeled datasets were thought to be necessary for training AI, but new developments have shown that it's possible to generate models without labeled data using techniques like Stable Diffusion and fine-tuning them for specific use cases.

Companies can choose to create their own AI processes or utilize someone else's offering. Some companies, like OpenAI and Copy.ai, have built their products on top of GPT-3, but have likely done some fine-tuning to customize the models. Integrated AI is another approach, where AI capabilities are integrated into existing products without displacing incumbents. Microsoft, for example, has begun integrating AI models like Dall-E into its Office suite.

Infrastructure as a Service is also a significant aspect of the AI value chain, where cloud providers like AWS, Oracle, and Azure build their own custom AI workload chips, networking software, and in-house models for reference. The intelligence layer focuses on the improvement of fundamental models, with fine-tuning becoming less about output quality and more about output cost and speed. Companies competing in this layer will depend on attracting top talent and achieving extraordinary feats.

Finally, the concept of invisible AI suggests that the most successful AI companies are those that seamlessly integrate AI into their products. Bytedance, the parent company of TikTok, is considered a prime example of this. While AI can provide delight and magic, the staying power of a product lies in its ability to solve a job-to-be-done, with AI being just one component.

To make the most of AI and note-taking, here are three actionable pieces of advice:

1. Emphasize active engagement with your notes. Instead of simply transcribing information, rephrase and self-explain what you consume. Take the time to understand and make connections to what you already know.

2. Use a note-taking app like Obsidian to create spaces for thinking and playing with your notes. Vague-notes allow for random thoughts and connections, while thinking-notes provide a more focused environment for shaping ideas and exploring related concepts.

3. Consider the broader context of the AI value chain. Understand the different components involved, such as compute power, data, fine-tuning, integrated AI, infrastructure as a service, the intelligence layer, and the concept of invisible AI. This understanding will help you navigate and make informed decisions in the AI landscape.

In conclusion, note-taking is not just about accumulating knowledge but actively engaging with it. By using notes, exploring connections, and building experience, we can deepen our understanding and spark new curiosities. In the realm of AI, understanding the value chain and the various components involved can provide insights into the potential impact and opportunities in this rapidly evolving field. By combining effective note-taking practices with a broader understanding of AI, we can navigate this landscape with greater clarity and intention.