Ai Task using knowledge base gives wrong informati...
# 🤝help
q
I am using a text document and a table. Both of them contain 46 property listings. When I ask questions like "How many properties are in Pestalozzistr?" , it gives me answers like 10 or 12. Its a very simple question, the ai just needs to count all rows. How can I fix this problem?

https://cdn.discordapp.com/attachments/1146363129563983922/1146363129723355206/image.png

https://cdn.discordapp.com/attachments/1146363129563983922/1146363130025365514/image.png

https://cdn.discordapp.com/attachments/1146363129563983922/1146363131602415656/image.png

https://cdn.discordapp.com/attachments/1146363129563983922/1146363131828895784/image.png

https://cdn.discordapp.com/attachments/1146363129563983922/1146363132072185896/image.png

a
Hey @quaint-plastic-78691 have you tried using a selector for this instead? A natural language query like this could be used to filter table data, and then you can pass this filtered data to the AI task and that might give a more accurate answer. If you're not afraid of code, there's a tutorial here that shows how you can do this process automatically based on the conversation https://discord.com/channels/1108396290624213082/1142011151874543747/1142050700646416544

https://cdn.discordapp.com/attachments/1146363129563983922/1146396274250240061/image.png

g
ChatGPT can't count. It's a language model. v4 is better but it's still can't be relied on.
2 Views