Skip to content

using exclusive nodebb forum data to create a chatgpt chatbot?

Discussion
7 2 2.0k 1
  • It appears that we are entering a new age… I guess soon we will be wearing glasses that can listen us and analyze statements/commands and make reservations, create alarms, send messages etc.

    With the GPT-4 Turbo (more powerful, faster, and most importantly cheaper), assistants (plug-ins), and a GPT store (specialized gpt versions created by users)…

    I wonder if there’s a potential to leverage NodeBB forum data to train GPT-4 Turbo in a way to create a chatbot, and when users write a question, this bot will give a concise information/answer to users and guiding them directly to relevant topic(s) in the forum by sharing the topic links?

  • It appears that we are entering a new age… I guess soon we will be wearing glasses that can listen us and analyze statements/commands and make reservations, create alarms, send messages etc.

    With the GPT-4 Turbo (more powerful, faster, and most importantly cheaper), assistants (plug-ins), and a GPT store (specialized gpt versions created by users)…

    I wonder if there’s a potential to leverage NodeBB forum data to train GPT-4 Turbo in a way to create a chatbot, and when users write a question, this bot will give a concise information/answer to users and guiding them directly to relevant topic(s) in the forum by sharing the topic links?

    @crazycells Interesting concept, and certainly one that warrants further exploration as it could extend the reach of any forum well beyond what search engines can offer.

    I know that someone in the NodeBB community has already begun something along these lines in the form of ChatGPT moderation, which looks good but isn’t free in the sense that it needs “credits” to work

    https://community.nodebb.org/topic/17198/chatgpt-moderation?_=1699722431266

    From recollection, a pro subscription to GPT4 costs $20/month - and is has caps, such as 50 queries max per 4 hours, and many are questioning if this fee make the product worth it or not - it’s not something I’d specifically use in terms of a subscription per month - at least, not enough to warrant it’s cost.

    For the training part, I guess that ChatGPT would have to crawl your website rather like an indexing service does. However, we already have search capabilities in NodeBB and ChatGPT would effectively be another - the question here is do we really need it?

  • @crazycells Interesting concept, and certainly one that warrants further exploration as it could extend the reach of any forum well beyond what search engines can offer.

    I know that someone in the NodeBB community has already begun something along these lines in the form of ChatGPT moderation, which looks good but isn’t free in the sense that it needs “credits” to work

    https://community.nodebb.org/topic/17198/chatgpt-moderation?_=1699722431266

    From recollection, a pro subscription to GPT4 costs $20/month - and is has caps, such as 50 queries max per 4 hours, and many are questioning if this fee make the product worth it or not - it’s not something I’d specifically use in terms of a subscription per month - at least, not enough to warrant it’s cost.

    For the training part, I guess that ChatGPT would have to crawl your website rather like an indexing service does. However, we already have search capabilities in NodeBB and ChatGPT would effectively be another - the question here is do we really need it?

    @phenomlab said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    From recollection, a pro subscription to GPT4 costs $20/month - and is has caps, such as 50 queries max per 4 hours, and many are questioning if this fee make the product worth it or not - it’s not something I’d specifically use in terms of a subscription per month - at least, not enough to warrant it’s cost.

    Maybe I am mistaken, but I thought this is the price for web-based ChatGPT subscription, where you use it for your own questions, right?

    I thought API price is based on the usage? input tokens & output tokens…

    https://help.openai.com/en/articles/7127956-how-much-does-gpt-4-cost

  • @phenomlab said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    From recollection, a pro subscription to GPT4 costs $20/month - and is has caps, such as 50 queries max per 4 hours, and many are questioning if this fee make the product worth it or not - it’s not something I’d specifically use in terms of a subscription per month - at least, not enough to warrant it’s cost.

    Maybe I am mistaken, but I thought this is the price for web-based ChatGPT subscription, where you use it for your own questions, right?

    I thought API price is based on the usage? input tokens & output tokens…

    https://help.openai.com/en/articles/7127956-how-much-does-gpt-4-cost

    @crazycells said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    where you use it for your own questions, right?

    Yes, that’s right. Sorry - I was trying to draw some parallel in terms of the cost.

    @crazycells said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    I thought API price is based on the usage? input tokens & output tokens…

    Yes, it is, although on larger forums it could still land up being quite expensive depending on how often it is used.

  • @crazycells said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    where you use it for your own questions, right?

    Yes, that’s right. Sorry - I was trying to draw some parallel in terms of the cost.

    @crazycells said in using exclusive nodebb forum data to create a chatgpt chatbot?:

    I thought API price is based on the usage? input tokens & output tokens…

    Yes, it is, although on larger forums it could still land up being quite expensive depending on how often it is used.

    @phenomlab yes, you are right. But my assumption is that larger forums would have money to fund this, although it does not necessarily have to be the case. This is the rate for GPT-Turbo…

    Screen Shot 2023-11-11 at 18.32.43.png

    $1 for 100.000 input tokens
    $3 for 100.000 output tokens

  • @phenomlab yes, you are right. But my assumption is that larger forums would have money to fund this, although it does not necessarily have to be the case. This is the rate for GPT-Turbo…

    Screen Shot 2023-11-11 at 18.32.43.png

    $1 for 100.000 input tokens
    $3 for 100.000 output tokens

    @crazycells your figures don’t seem to match the video?

    Don’t you mean $1 for 1,000? 🤔

  • @crazycells your figures don’t seem to match the video?

    Don’t you mean $1 for 1,000? 🤔

    @phenomlab oh no, that is 1 cent on the video, but you are right, symbols are similar… I just converted it to $1 , since it is more intuitive in daily life…


Related Topics
  • 0 Votes
    1 Posts
    451 Views
    No one has replied
  • Which email client do you use?

    Discussion email client
    56
    22 Votes
    56 Posts
    16k Views
    @phenomlab said in Which email client do you use?: @DownPW Isn’t this more of an email server in it’s own right rather than an email client? Oh yes sorry… I didn’t pay attention
  • CSS codes to prevent votes to disappear

    Solved Customisation nodebb
    3
    1 Votes
    3 Posts
    1k Views
    @phenomlab yes, it is thanks a lot…
  • Threaded post support for NodeBB

    Let's Build It threading nodebb
    146
    3
    50 Votes
    146 Posts
    70k Views
    Updated git for above change https://github.com/phenomlab/nodebb-harmony-threading/commit/14a4e277521d83d219065ffb14154fd5f5cfac69
  • 4 Votes
    4 Posts
    1k Views
    @phenomlab said in TikTok fined £12.7m for misusing children’s data: Just another reason not to use TikTok. Zero privacy, Zero respect for privacy, and Zero controls in place. https://news.sky.com/story/tiktok-fined-12-7m-for-data-protection-breaches-12849702 The quote from this article says it all TikTok should have known better. TikTok should have done better They should have, but didn’t. Clearly the same distinct lack of core values as Facebook. Profit first, privacy… well, maybe. Wow, that’s crazy! so glad I stayed away from it, rotten to the core.
  • Error install plugin

    Solved Customisation plugin nodebb
    8
    1
    1 Votes
    8 Posts
    4k Views
    @pobojmoks Not WP plugin but nodeBB but it a known bug
  • NodeBB slow after DB recovery

    Solved Performance performance nodebb
    1
    5 Votes
    1 Posts
    693 Views
    No one has replied
  • [NODEBB] Help for my custom CSS

    Solved Customisation nodebb css bugfix
    237
    49 Votes
    237 Posts
    99k Views
    @baris said: You should change your selectors so it doesn’t look at the entire document. You probably only want to apply fancybox to stuff inside the #content element which is what changes when the user navigates around the page. So use $('#content a').... for your selectors then the forum logo in the header won’t be selected. I modified the JS Fancybox code now and this code and it seem better // --------------------------------------------- // Fancybox Media Reader (Without Website Logo) // --------------------------------------------- if (top.location.pathname !== '/login') { $(window).on('action:posts.loaded', function(data) { console.log("Polling DOM for lazyLoaded images to apply Fancybox"); $(document).ready(function() { $('#content a').not('.forum-logo').not(".avatar").not(".emoji").not(".bmac-noanimate").each(function() { $('#content a[href*=".jpg"], #content a[href*=".jpeg"], #content a[href*=".png"], #content a[href*=".gif"], #content a[href*=".webp"]').addClass("noanimate"); }); }); }); } if (top.location.pathname !== '/login') { $(document).ready(function() { $(window).on('action:ajaxify.end', function(data) { $('#content a').not('.logo').not(".avatar").not(".emoji").not(".bmac-noanimate").each(function() { $('#content a[href*=".jpg"], #content a[href*=".jpeg"], #content a[href*=".png"], #content a[href*=".gif"], #content a[href*=".webp"]').addClass("noanimate"); data.preventDefault() // Strip out the images contained inside blockquotes as this looks nasty :) $('#content blockquote img').remove(); }); Fancybox.bind( '#content a[href*=".jpg"], #content a[href*=".jpeg"], #content a[href*=".png"], #content a[href*=".gif"], #content a[href*=".webp"]', { groupAll: true, } ); }); }); } // Chat fancybox - fires when chat module loaded and AJAX calls new chat $(document).ready(function() { $(window).on('action:chat.loaded', function(data) { // >>> Se limiter au contenu principal uniquement <<< $('#content img').not('.forum-logo').not(".avatar").not(".emoji").not(".bmac-noanimate").each(function() { var newHref = $(this).attr("src"); $(this).wrap("<a class='fancybox' href='" + newHref + "'/>"); $('#content a[href*=".jpg"], #content a[href*=".jpeg"], #content a[href*=".png"], #content a[href*=".gif"], #content a[href*=".webp"]').addClass("noanimate"); data.preventDefault(); // Strip out the images contained inside blockquotes as this looks nasty :) $('#content blockquote img').remove(); }); Fancybox.bind( '#content a[href*=".jpg"], #content a[href*=".jpeg"], #content a[href*=".png"], #content a[href*=".gif"], #content a[href*=".webp"]', { groupAll: true, } ); }); }); For the logo, I must use overflow: visible !important; on [component="brand/logo"] /* --- Logo --- */ [component="brand/logo"] { max-height: 50px; width: auto; height: auto; max-width: 100%; display: block; object-fit: contain; object-position: left center; overflow: visible !important; } Better result !!