Photo by Lukas on Pexels

Leading AI models, including GPT, DeepSeek, Gemini, and Copilot, are still struggling with the intricacies of spreadsheet logic, as demonstrated in a recent experiment. The test, shared on Reddit, challenged the AIs to generate a Google Sheets function for calculating installment payments based on factors like payment plans, invoice amounts, and due dates. While none of the models produced a fully functional solution, DeepSeek showed a slight edge by attempting to reference columns rather than fixed cell locations, indicating a better grasp of spreadsheet structure. This highlights a current weakness in AI’s ability to handle complex calculations within dynamic spreadsheet environments, suggesting further development is needed before these tools can reliably automate sophisticated spreadsheet tasks. The original experiment and discussion can be found on Reddit: [https://old.reddit.com/r/artificial/comments/1oundu4/i_tried_this_problem_on_4_ai_gpt_deepseek_gemini/]