The moment I told ChatGPT it needed a history lesson—and it agreed

The Limitations of AI: A Personal Experience

Artificial intelligence has made remarkable strides in recent years, with tools like ChatGPT and Google’s Gemini leading the charge. These platforms have shown incredible potential in various tasks, from generating text to analyzing complex data. However, my recent experience with these technologies has highlighted that they are not infallible. While I remain optimistic about the long-term future of AI, it's crucial to approach its capabilities with a balanced perspective.

A Specific Project That Highlighted AI's Shortcomings

This week, I encountered a situation that served as a stark reminder of AI's limitations. My daughter was working on her doctoral dissertation and needed a comparative map showing the borders of the Byzantine Empire in 379 AD and 457 AD. I decided to test the capabilities of AI by using a specific prompt:

"Create a detailed map that overlays the borders of the Byzantine empire in 379AD at the start of the reign of Theodosius the Great versus the borders in 457AD at the end of the reign of Marcian. I need both borders shown clearly on a single map. Use a historical map style and highlight major cities."

The Deep Research option, which is known for its thoroughness, provided an excellent textual analysis of the changing borders, major cities, and historical events. However, when I asked the AI to generate a visual representation of this information, things quickly went awry.

The Map Generation Fiasco

The generated map contained several inaccuracies. It invented names, misspelled existing ones, and placed cities in random locations. For instance, Rome appeared twice on the Italian peninsula, which was particularly frustrating since the correct names and locations were present in the text. I attempted to correct these errors, but the AI failed to respond appropriately. After some frustration, I told the AI that its results were unacceptable, and it responded with a polite acknowledgment. Unfortunately, this did not lead to any improvement in the output.

I then tried Google Gemini with the same query. The results were even more concerning. The map displayed "Rome" in the middle of the Iberian Peninsula and featured multiple instances of Antioch across Europe. Other names seemed to be pulled from fantasy novels, making the map almost unusable.

Broader Implications of AI Errors

This experience is not isolated. A friend shared a similar story where he used ChatGPT to add the text "Mahalo from Hawaii 2025" to a photo. Instead of simply adding the text, the AI altered the image significantly, changing people's appearances and even gender. Another friend recounted how an AI-generated biography incorrectly mentioned his twin children, which he does not have, and included a link to a non-existent source.

The Importance of Verification

Ronald Reagan famously said, "Trust but verify." This advice is especially relevant when dealing with AI. While these tools can be incredibly useful, they should not be treated as infallible. It's essential to review their outputs carefully, ensuring that the information is accurate and reliable.

Conclusion

AI is undoubtedly a powerful tool that can assist in many areas. However, it's important to recognize its limitations and not place blind trust in its outputs. As we continue to integrate AI into our daily lives, we must remain vigilant and critical of its results. The rise of AI is inevitable, but today is not the day to relinquish our responsibility for verification.