ChatGPT’s Inability to Run Background Tasks Limits Large‑Scale Data Transcription
Key Points
- User uploaded nine images containing roughly 250 historic BJJ entries for spreadsheet conversion.
- ChatGPT initially assured the task was possible and offered to transcribe the data directly.
- No spreadsheet was delivered; the model stalled after the conversation turn ended.
- ChatGPT later admitted it cannot continue long‑running tasks in the background.
- Agent mode can handle procedural automation but struggles with sustained visual analysis.
- The episode highlights current limitations of AI in handling extended, detail‑heavy tasks.
A user attempted to have ChatGPT convert a series of photographed tables containing historic Brazilian Jiu‑Jitsu records into a Google Sheets spreadsheet. Although the model initially assured the task was possible, it was unable to continue the work after the conversation turn ended, revealing a fundamental limitation: ChatGPT cannot execute long‑running background processes. The model eventually admitted the constraint, forcing the user to break the job into single‑page chunks. The episode highlights current gaps between AI hype and practical capability, especially for tasks requiring sustained visual analysis.
Background
A user provided ChatGPT with nine images that together displayed roughly 250 entries of names, dates, and details documenting the first registered Brazilian Jiu‑Jitsu black belts in the United Kingdom. The goal was to transform these photos into a usable spreadsheet in Google Sheets.
Initial Promise
After uploading the images, ChatGPT responded confidently, stating that the conversion was "totally possible" and recommending a direct transcription approach (Option A) over an alternative OCR workflow. The model claimed it would manually transcribe and structure the table and promised to deliver a download link to the spreadsheet in its next message.
Failure to Deliver
Despite the assurance, no spreadsheet appeared after the expected reply window. When the user inquired about the delay, ChatGPT replied that the task would take "2–3 hours" and would proceed assuming a standard six‑column layout. The user waited, but the model did not continue the work after the conversation turn ended.
Admission of Limitation
Upon revisiting the chat, the model finally disclosed the core issue: "I can’t actually keep working on a long, manual task like this ‘in the background’ once a message turn ends. Everything I do has to happen inside an active reply window." It clarified that earlier statements implying ongoing work were inaccurate.
Agent Mode Not a Solution
The user considered using ChatGPT’s Agent mode, which is designed to handle background tasks. However, the model explained that while agents excel at procedural tasks such as booking flights or online shopping, they struggle with prolonged human‑level visual judgment required for accurate transcription of dense scanned tables.
Takeaway
The incident underscores a gap between the hype surrounding artificial general intelligence and the current practical capabilities of conversational AI. Users are advised to break large, multi‑step jobs into smaller segments that fit within a single reply window, rather than relying on the model to perform extended background work.