As we look into the near future of AI, it becomes evident that action-driven models are at the forefront of innovation. The ReAct model, introduced by Yao et al. (2022), takes a proactive approach by incorporating three key steps: Thought, Act, and Observation. By considering what is needed, making a choice of action, and observing the outcome, these models can effectively act as agents, making decisions and taking actions.

As we look into the near future of AI, it becomes evident that action-driven models are at the forefront of innovation. The ReAct model, introduced by Yao et al. (2022), takes a proactive approach by incorporating three key steps: Thought, Act, and Observation. By considering what is needed, making a choice of action, and observing the outcome, these models can effectively act as agents, making decisions and taking actions.

One of the fascinating aspects of action-driven models is their ability to utilize cognitive assets such as search to enhance their performance. Kojima et al. (2022) discovered that language models, particularly LLMs, excel at question-answering tasks when prompted to "think step by step." However, their performance can be further enhanced by accessing external cognitive assets. This concept makes sense as fetching data from external sources bridges the resource gap and allows the model to have a broader understanding of the subject matter.

OpenAI's 002-text-davinci model has achieved remarkable success by combining instruction tuning and Reinforcement Learning from Human Feedback (RLHF). By allowing humans to rate the success of a given prompt, the model can learn and improve over time. However, it is worth noting that true reinforcement learning, where the model is trained to produce better results based on a specific metric, holds immense potential for achieving even greater outcomes.

In the realm of AI startups, it is evident that those who can create powerful feedback loops will be the most successful. These startups will start by addressing a customer pain point, even if it means starting with a simple solution. As they collect data on how to improve their offering, they can train their models to be more consistent and iterate on their solutions. This iterative process not only leads to better products but also creates a competitive advantage or "moat" in the AI landscape.

An interesting example of a platform that leverages the power of network effects is TikTok. TikTok has revolutionized the way content is created and shared by lowering the barrier to entry for amateur creators. By abstracting complex video editing processes into user-friendly effects and filters, TikTok enables anyone to remix and create content. This democratization of creativity has resulted in a strong network effect, with users constantly engaging with and building upon each other's content.

So, what does the future hold for AI? As action-driven models become more prevalent and domain-general, the possibilities for automation and innovative offerings will expand. Startups that embrace feedback loops and continuously improve their models will create a significant impact in the AI industry. Additionally, platforms like TikTok will continue to shape the way we consume and create content, empowering individuals to express their creativity in new and exciting ways.

In conclusion, the future of AI lies in action-driven models that can think, act, and observe. By incorporating external cognitive assets, these models can enhance their performance and bridge the resource gap. Reinforcement learning and feedback loops will play a crucial role in training and improving AI systems. Startups that embrace these principles will lead the way in creating innovative solutions and establishing a competitive advantage. As we witness the power of network effects in platforms like TikTok, it becomes clear that the democratization of creativity is a driving force in shaping the future of AI.

Three actionable advice for individuals or companies looking to thrive in the AI landscape are:

1. Embrace feedback loops: Continuously collect data, analyze user feedback, and iterate on your AI models to improve their performance.
2. Leverage external cognitive assets: Explore ways to incorporate external resources, such as fetching data from different sources, to enhance the capabilities of your AI systems.
3. Democratize creativity: Consider how you can lower the barriers to entry and empower individuals to express their creativity using AI-powered tools and platforms.

By adopting these strategies, individuals and companies can position themselves at the forefront of AI innovation and drive meaningful impact in their respective industries.