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

A report by **TechCrunch** details the core thesis behind **General Intuition**, a company betting that **gaming data** is superior to web text for reaching **artificial general intelligence**. Traditional **large language models** like **ChatGPT** and **Claude** rely primarily on text, but **General Intuition** argues that text datasets alone cannot produce truly generalized intelligence.

Mechanically, models such as **ChatGPT** and **Claude** excel at text processing, yet struggle to comprehend how objects move through **space and time**. **Gaming data** supplies the spatial and temporal information missing from text corpora, allowing models to learn movement and physical relationships. Mastering these spatial-temporal dynamics is essential for producing intelligence that generalizes across environments.

For engineers building advanced systems, this distinction highlights the limits of training exclusively on text. **Gaming data** gives developers a path to teach models physical logic and temporal awareness that text inputs fail to convey. Builders targeting **artificial general intelligence** must look beyond text to environment-based datasets to bridge this operational gap.

In the market context, leading platforms like **ChatGPT** and **Claude** continue to focus on text-heavy models. **General Intuition** differentiates its approach by using **gaming data** as an alternative data source to compete with internet text models. This strategy challenges the assumption that scaling web text alone is sufficient for model development.

The practical takeaway is to evaluate how effectively **General Intuition** converts **gaming data** into measurable spatial reasoning improvements. Technical teams should watch whether training on interactive game environments successfully resolves the spatial and temporal limitations inherent in standard text-based models.