Experimenting with AI: GPT-2 Model Implemented in Excel

Intrigued by the mechanics of artificial intelligence, software developer Ishan Anand has taken a creative dive into the world of large language models (LLMs) by embedding the GPT-2 algorithm into a Microsoft Excel spreadsheet. This endeavor is more than a mere fascination with spreadsheets; it presents a unique gateway into comprehending the intricacies of AI for those who are familiar with the gridlines of Excel.

Anand’s experiment has resulted in a hefty 1.25GB Excel file, attainable on GitHub, that demonstrates the workings of GPT-2, the once-revolutionary language model. Remarkable for its time in 2019, GPT-2 is notable for its smart “next-token prediction” capabilities, powered by the Transformer architecture which considerably transformed AI’s approach towards predictive text generation.

This particular GPT-2 manifestation utilizes the smaller variant of the original model, packing 124 million parameters into an Excel format compared to the full version’s 1.5 billion parameters. Though dwarfed by the colossal up to 175 billion parameter GPT-3 and even larger GPT-4, Anand’s model serves as an enlightening low-code introduction to AI principles.

A prime educational tool, Anand’s spreadsheet model demonstrates AI’s foundation—applicable to numerous LLMs including OpenAI’s ChatGPT and Google’s Bard. It is particularly aimed at those unversed in AI, such as tech executives or AI policy-makers. Despite its limitations, like supporting only 10 tokens of input, the model runs entirely offline on a PC, providing a sandbox for AI enthusiasts without involving cloud-based constraints.

The developer cautions against running this hefty spreadsheet on Mac platforms or cloud services but recommends it as an intellectually stimulating engagement for those on the latest version of Excel. Through this initiative, Anand underscores the enduring relevance of the Transformer architecture in AI development, offering a tangible demonstration for all to explore.

The Integration of AI in Spreadsheets: An Educational Leap

Ishan Anand’s integration of GPT-2 into a Microsoft Excel spreadsheet represents a significant stride in making artificial intelligence concepts accessible to a broader audience. By leveraging the familiar environment of Excel, Anand has opened a gateway that allows those outside traditional tech circles, such as tech executives and policy-makers, to experience the mechanics of AI firsthand.

The industry surrounding large language models like GPT-2, GPT-3, and GPT-4 has witnessed substantial growth in recent years. These AI systems have not only fueled advancements in natural language processing (NLP) but have also been instrumental in shaping the landscape of AI as a whole. The Transformer architecture, which underpins these models, has become a cornerstone in the field, enabling a leap in the AI’s ability to process and generate human-like text.

Market Forecasts and Industry Growth

Market forecasts for AI and, by extension, NLP technologies are bullish, with the AI sector expected to grow significantly in the next decade. Analysis from firms such as Grand View Research and MarketsandMarkets predicts that the AI market will be worth hundreds of billions of dollars, with NLP forming a key component due to its myriad applications in customer service, content generation, and more.

The implementation of AI within popular software like Excel serves not just as a convenient educational tool but also as a potential trendsetter, hinting at the future of AI democratization. Ensuring that AI tools can be engaged without needing sophisticated coding skills may drive the future of software development and increase AI adoption across various industries.

Issues and Challenges

Despite its promise, the integration of AI technologies such as LLMs into widespread applications does not come without its complexities. There are concerns over AI ethics, including data privacy, algorithmic bias, and the impact of automation on the job market. Moreover, the technical challenge of embedding such sophisticated models into user-friendly formats without loss of functionality or performance demands attention and innovation.

Anand’s heavy Excel-based GPT-2 file is also reminiscent of the need for robust hardware that can handle advanced computations. As AI models grow in size and complexity, there is a parallel increase in the computational power required to run them, which poses a barrier for widespread access and use.

Apart from individual experiments such as Anand’s, there are whole industries and companies dedicated to AI research and development. The AI landscape includes giants such as OpenAI, Google, Microsoft, and Amazon, all competing to advance AI technology and its applications.

For users intrigued by the applicability of AI in an everyday software context, Anand’s project serves not just as an enlightenment tool but also as a testament to the evolving intersection of AI and consumer software. It exemplifies the potential for complex algorithms to be repackaged into more approachable and practical forms, allowing AI education and experimentation to permeate into new domains and reach a wider audience.



Oliwier Głogulski is a distinguished author and expert in the field of new technology equipment and services. His work is characterized by in-depth analyses and reviews of the latest tech innovations. Głogulski's articles and publications are valued for their comprehensive coverage and insightful perspectives on emerging trends and technologies. His contributions significantly influence consumer and professional understanding of the rapidly evolving tech landscape.