[021] ChatGPT Autoresponder [Deleted]

[Starbucks]

@021 maybe it’s an idea to put somewhere in the settings or so an alert to inform users about how they can limit their spendings, so they won’t have unexpected costs.

You can find the link below from my previous post:

For those that are afraid to end up being billed with a crazy high price, you can easily set up a limit in the OpenAI settings:

OpenAI Platform

Explore developer resources, tutorials, API docs, and dynamic examples to get the most out of OpenAI's platform.

platform.openai.com

 

[colcar]

Implemented this addon around 1pm and my members have been peppering it with questions all day long. Here's my usage:

 

[Kevin]

@021 In the next round of testing, can you please make sure that the bot posts are being bumped up against the normal XF spam checks? In particular it seems to be bypassing the XF 'spam phrases' check.

 

[Mr Lucky]

I really like the new feature with user field to tick yes or no for the not answer.

But is there a way to have it ticked no by default instead of yes, or else unticked by default and so be prompted to tick either yes or no?

 

[021]

@021 In the next round of testing, can you please make sure that the bot posts are being bumped up against the normal XF spam checks? In particular it seems to be bypassing the XF 'spam phrases' check.

I do not see the practical benefit of this step. As far as I know ChatGPT does not respond with spam messages. Can you please give an example of a situation where such option would be relevant?

I really like the new feature with user field to tick yes or no for the not answer.

But is there a way to have it ticked no by default instead of yes, or else unticked by default and so be prompted to tick either yes or no?

Such suggestion is already being considered. You can vote for it on the forum

 

[Kevin]

I do not see the practical benefit of this step. As far as I know ChatGPT does not respond with spam messages. Can you please give an example of a situation where such option would be relevant?

Conversely I don't see a benefit of not having the posts go through XF's normal checks but here's a few items...

• ChatGPT is not immune to giving questionable results that can be problematic. (eg: https://www.thedailybeast.com/openais-impressive-chatgpt-chatbot-is-not-immune-to-racism or https://www.vice.com/en/article/n7z...hatgpt-to-make-it-endorse-racism-conspiracies or https://abcnews4.com/news/nation-wo...tificial-intelligence-chat-bot-chatbot-google) It would quickly turn into a huge mess if some users figured out what to ask ChatGPT to get a problematic response knowing that the response won't get moderated.
• Admins can use the XF 'spam phrases' functionality to catch topics that might be questionable for their community, not just pure "spam". By putting certain phrases in the XF 'spam phrases' it puts any caught posts in moderation where they can be reviewed before being approved to be viewable by the public.
• It would help fix this problem: https://xenforo.com/community/threads/021-chatgpt-autoresponder-paid.212720/post-1617693 (when the bot response quotes the original post that got hit with moderation then it'd make sense that the quoted response would also get hit with moderation).

 

[021]

Conversely I don't see a benefit of not having the posts go through XF's normal checks but here's a few items...

• ChatGPT is not immune to giving questionable results that can be problematic. (eg: https://www.thedailybeast.com/openais-impressive-chatgpt-chatbot-is-not-immune-to-racism or https://www.vice.com/en/article/n7z...hatgpt-to-make-it-endorse-racism-conspiracies or https://abcnews4.com/news/nation-wo...tificial-intelligence-chat-bot-chatbot-google) It would quickly turn into a huge mess if some users figured out what to ask ChatGPT to get a problematic response knowing that the response won't get moderated.
• Admins can use the XF 'spam phrases' functionality to catch topics that might be questionable for their community, not just pure "spam". By putting certain phrases in the XF 'spam phrases' it puts any caught posts in moderation where they can be reviewed before being approved to be viewable by the public.
• It would help fix this problem: https://xenforo.com/community/threads/021-chatgpt-autoresponder-paid.212720/post-1617693 (when the bot response quotes the original post that got hit with moderation then it'd make sense that the quoted response would also get hit with moderation).

Will be added as an option)

 

[Kevin]

Will be added as an option)

Thanks

 

[tajhay]

Do users get notifications/alerts when chatbot has responded to them?

 

[ProCom]

Do users get notifications/alerts when chatbot has responded to them?

yes

 

[Mr. Jinx]

Note about the error "ChatGPT Quick Reply Bot error: Empty reply"

I'm getting this error with some specific questions about Pi and Fibonacci.
With other questions, no problems.

 

[ProCom]

I've been doing a lot more research and testing on pre-training and it's amazing how different the responses can be (in quality, tone, etc.) when using different prompts... like Role Prompting, etc.

The more I use these AI addons, the more I realize how important it is to tweak the AI relative to each community's focus, niche, tone, demeanor, etc. Some forums are more serious. Some more light-hearted. It might be imperative to be able to adjust the AI pretraining accordingly.

(I'm also posting what I posted here since this thread is way more active and I'd love to hear what others think of pre-training)

 

[tajhay]

Anyone using language as anything other than auto? Tried "New Zealand English" but it still annoyingly spells things in American I.e. "apologize", "color" etc

 

[colcar]

Anyone using language as anything other than auto? Tried "New Zealand English" but it still annoyingly spells things in American I.e. "apologize", "color" etc

Yes, there are American spellings in British English settings too.

I've discovered the Temperature settings don't actually do anything either.

 

[Mr Lucky]

Anyone using language as anything other than auto? Tried "New Zealand English" but it still annoyingly spells things in American I.e. "apologize", "color" etc

You should ask the bot why this is. I’m sure they will apologize.

 

[tajhay]

You should ask the bot why this is. I’m sure they will apologize.

Certainly apologises a lot. Though I did teach it to talk in kiwi slang a bit. Now it often apologises with "sorry my bad my bro" lol

 

[021]

OMG, OMG, OMG!!! I have something... something COOL...

ChatGPT Role Bots announcement

YO HO HO! Many asked to implement a bot response to mentions in the [021] ChatGPT Autoresponder add-on. But since this add-on was originally created for quick answers in certain sections, this function would be superfluous in it. Instead of spoiling the add-on with unnecessary features, I have...

devsell.io

XenForo add-ons, they're like gold
From UI.X to Resource Manager, they're all bold
Got me hooked like a fisherman, never sold
XenForo add-ons, worth every penny, never old

 

[021]

021 updated [021] ChatGPT Autoresponder with a new update entry:

1.3.0

What's new​

Option to enable checking bot messages for spam
Option to set the default value for the field "Allow AI replies"
Autoresponder permissions (Can be configured individually for sections)
Lightweight refactoring
Don't forget to adjust permissions after this update

Read the rest of this update entry...

 

[Mr Lucky]

Installed update

Nice that you can set the default but the field Allow AI replies when creating a thread has now disappeared.

 

[021]

Nice that you can set the default but the field Allow AI replies when creating a thread has now disappeared.

Don't forget to adjust permissions after this update