"The Near Future of AI is Action-Driven: An Interview With Twitter's Forgotten Founder, Noah Glass"

"The Near Future of AI is Action-Driven: An Interview With Twitter's Forgotten Founder, Noah Glass"

In recent years, the field of artificial intelligence (AI) has made remarkable advancements. From natural language processing models to question-answering tasks, AI has proven its capabilities. But what lies ahead for the future of AI? According to the ReAct model introduced by Yao et al. in 2022, the key to unlocking the true potential of AI lies in being action-driven.

The ReAct model takes a three-step approach: Thought, Act, and Observation. It emphasizes the importance of not just thinking about what is needed, but also taking action and observing the outcomes of those actions. This iterative process allows AI models to act as agents, making choices and learning from the results. In fact, an action-driven Language Model (LLM) closely resembles the concept of Artificial General Intelligence (AGI), which is a hot topic of debate among academics.

LLMs have shown impressive performance in question-answering tasks, especially when prompted to "think step by step." However, they can achieve even better results when provided with external cognitive assets. This means that by fetching data from external sources, LLMs can fill in the resource gap and enhance their performance.

OpenAI's 002-text-davinci model has been successful due to a combination of instruction tuning and Reinforcement Learning from Human Feedback (RLHF). With RLHF, human raters evaluate the success of a given prompt, allowing the model to improve over time. But to achieve the best results, actual reinforcement learning, where the system is trained to produce better outcomes based on a specific metric, seems to be the way forward.

Looking ahead, startups in the AI field have an opportunity to create powerful feedback loops. By identifying and solving customer pain points, collecting data on improving solutions, and training their models to be more consistent, these startups can iterate and build a competitive advantage. This iterative process, combined with an action-driven approach, can be seen as the moat in AI for now.

Drawing parallels to the world of tech startups, Noah Glass, forgotten founder of Twitter, highlights the significance of collaboration and group effort in the success of a product. In an exclusive interview, Glass emphasizes that while some individuals may receive credit, the reality is that it takes a collective effort to bring an idea to fruition. Collaboration, in his view, is a necessity that leads to groundbreaking innovations.

Glass reflects on the early days of Twitter, where he had a strong belief in the product's potential. He recognized the social aspect between people as a driving force behind its growth. Furthermore, the idea of using Twitter for updates from police departments or fire departments was present from the beginning. This foresight and vision paved the way for Twitter's widespread adoption and diverse applications.

Ev Williams, co-founder of Twitter, played a crucial role in recognizing the value of the platform. Known for his shrewd business sense, Williams had experienced failures in the past. He saw the potential in Twitter and made it the focus of Obvious, an umbrella company initially intended for multiple projects. The simplicity and engagement of Twitter were highly compelling, even impressing experts in SMS applications.

As we consider the future of AI and the lessons from Twitter's success, there are actionable insights that can guide us:

1. Embrace an action-driven approach: Just as the ReAct model suggests, AI models should not be limited to passive thinking but should actively engage in actions and learn from the outcomes. This iterative process will lead to better results and closer alignment with AGI.

2. Leverage external cognitive assets: By incorporating external data and resources, AI models can bridge the resource gap and enhance their performance. Access to a wider range of information and knowledge will enable models to provide more accurate and comprehensive answers to complex questions.

3. Foster collaboration and group effort: Recognize that great ideas and successful products are often the result of collaboration. Encourage teamwork, value diverse perspectives, and create an environment where collective efforts can thrive.

In conclusion, the near future of AI is action-driven. The ReAct model, combined with the utilization of external cognitive assets, holds the key to unlocking the full potential of AI. As startups explore the AI landscape, they can learn from the collaborative spirit that drove Twitter's success. By embracing an action-driven approach, leveraging external resources, and fostering collaboration, we can shape an AI future that is both powerful and inclusive.