Skip to main content

Hi,

I’ve created a ‘FB Messenger’ trigger to ‘Conversation with ChatGPT’ zap. I have chosen GPT-4 as a model, however, I do not believe GPT-4 is being used.

Here is the Zap model I chose: 

 

Here is the GPT’s response when I ask it what model it is using

 

In addition, I am finding that the quality of reponses vary greatly - it sometimes responds according to my prompt, but often it does not.

 

Can you please help?

 

Best,

Arnina

Hi @Arnina 

Good question.

I believe the data ChatGPT models are trained on is still limited.

ChatGPT has limited knowledge of events that occurred after September 2021.>35]


Hi Troy,

Thanks for your help. I believe the issue is that when asked, GPT says it’s using the GPT-3 model, not the GPT-4 model even though I specified GPT-4 in the Zap. Do you know why this is or how I can fix it?

 

Thanks,

Arnina 


@Arnina 

That can be due to hallucination.

 

When asked directly in ChatGPT.

 

When asked using a ChatGPT - Conversation Zap action.

 


Interesting! Thanks for checking.

Could you please explain what you mean by 'hallucination'?

In addition, how can we be sure that GPT-4 is actually being used in the ‘ChatGPT - Conversation with Zap' action?

Thanks
Arnina


@Arnina 

How can we be sure that GPT-4 is actually being used in the ‘ChatGPT - Conversation with Zap' action?

It’s based on the model selected in the ChatGPT action.

 

Could you please explain what you mean by 'hallucination'?

Try asking ChatGPT to explain the concept of hallucination.