Why this CEO thinks video games make better training data than the internet

Why Gaming Data Might Be the Secret Sauce for Artificial General Intelligence

The pursuit of artificial general intelligence (AGI) has hit a roadblock, and it’s not just about the limitations of large language models like ChatGPT and Claude. The real challenge lies in their inability to understand the physical world, where objects move through space and time. This is where gaming data comes in, and General Intuition, a Bezos-backed startup, is betting big on it. With a recent $320 million funding round and a valuation of $2.3 billion, the company is making a bold claim: world models trained on gaming data might be the next big leap in physical AI.

This mirrors the early days of computer vision, where researchers used video games to train AI models to recognize objects in images. Games like GTA V and World of Warcraft provided a wealth of data on how objects interact with each other and their environment. Similarly, General Intuition is leveraging gaming data to train its models to understand the physical world. This approach has the potential to overcome the limitations of traditional data sources, which often lack the depth and complexity of real-world interactions.

The implications of this shift are significant, and it’s not just about the potential breakthroughs in AGI. The use of gaming data also raises questions about the ownership and control of this data. As AI models become increasingly dependent on gaming data, the companies that own this data will have significant influence over the development of AGI. This is a concern that General Intuition’s CEO, Pim de Witte, acknowledges, highlighting the need for transparency and accountability in the use of gaming data.

General Intuition’s Decision Logic: A Deep Dive into the Mechanics

So, what’s driving General Intuition’s decision to bet on gaming data? According to Pim de Witte, it’s all about the quality and quantity of the data. Gaming data provides a level of depth and complexity that traditional data sources can’t match. It’s not just about the number of data points; it’s about the richness and diversity of the interactions that games provide. This is particularly important for training models to understand the physical world, where objects interact with each other in complex ways.

But what about the operational mechanics? How does General Intuition plan to leverage gaming data to train its models? The company is using a combination of machine learning algorithms and data annotation techniques to extract insights from gaming data. This involves annotating data points with labels that indicate the type of interaction, the objects involved, and the outcome. This annotated data is then used to train models to predict the behavior of objects in different scenarios.

The tradeoffs here are significant. General Intuition is investing heavily in data annotation and machine learning infrastructure, which requires significant resources and expertise. However, the potential payoff is substantial, as the company believes that its approach can lead to breakthroughs in AGI.

Winners, Losers, and Disrupted Parties: The Impact of Gaming Data on AGI

So, who benefits from General Intuition’s approach? The obvious winners are the gaming companies that own the data, such as Epic Games and Activision Blizzard. These companies can monetize their data by licensing it to AI startups like General Intuition. The losers, on the other hand, are the traditional data providers, such as research institutions and government agencies, which may see their data sources become less relevant.

But what about the adjacent markets? The use of gaming data in AGI has significant implications for industries like robotics and autonomous vehicles. These industries rely heavily on AI models that can understand the physical world, and the use of gaming data can provide a significant boost to their development. The downstream effect of this development is substantial, as it can lead to breakthroughs in areas like healthcare and transportation.

The impact on job categories is also significant. As AI models become increasingly dependent on gaming data, the demand for data annotators and machine learning engineers will rise. However, this may also lead to job displacement in areas like data collection and annotation, where automation can replace human labor.

The Skeptical Case: Why Gaming Data May Not Be the Silver Bullet

But what about the skeptical case? Why might gaming data not be the silver bullet that General Intuition claims it to be? One reason is that gaming data may not be representative of the real world. Games are designed to be engaging and entertaining, not to accurately simulate the physical world. This can lead to biases in the data that may not be representative of real-world interactions.

Another reason is that the use of gaming data may not be scalable. As AI models become increasingly complex, they require larger and larger amounts of data to train. Gaming data may not be able to provide the scale and diversity of data that these models require. This is a concern that General Intuition acknowledges, highlighting the need for continued investment in data annotation and machine learning infrastructure.

The Signal to Watch Next: A Concrete Reason to Return to This Topic

So, what’s the next verifiable event that will confirm or disprove the thesis of this article? One signal to watch is General Intuition’s next funding round. If the company can secure additional funding, it will be a strong indication that investors believe in the potential of gaming data to drive breakthroughs in AGI. Another signal is the company’s next product launch. If General Intuition can demonstrate a working product that leverages gaming data to achieve significant breakthroughs in AGI, it will be a strong indication that the company’s approach is viable.

Pick one tactic from this post and apply it today. Which one will you start with?

By Daniel Cross, Digital Growth Strategist at TrendFlashy