by 0x1b35f4d22b2c8bc706bcfcc383a1ca77fb4fc600 (punkpink)
I have been working on a ChatGPT specialized in Decentraland documentation. Although it’s useful for quick queries, it lacks much other information from Decentraland that could enrich it, such as DAO proposals, forum posts, GitHub data, and other sites.
Two possible paths are presented:
Option 1
Improve ChatGPT DCL-Docs with proper instructions to extract information through URLs related to the platform. This can be relatively quick and easy, but while the information extraction from docs.decentraland.org was successful, it couldn’t properly extract it from sites like Notion. This could lead to the failure to extract information from other places.
Advantage: Quick and easy improvement.
Challenge: Issues extracting data from sites like Notion.
Disadvantage: Incomplete information.
Estimated Cost: $480.
Option 2
This would be an appropriate way for ChatGPT DCL-Docs to function properly and provide precise answers from all the currently available information. However, it requires a more complex process that, although it can be automated at times, will not always be so and will require manual intervention.
Advantage: Complete and precise answers from all documentation.
Challenge: Collection, extraction, and processing of information, a complex process with possible manual intervention.
Disadvantage: Higher cost.
Estimated Cost: $3,200.
Additional Considerations
Currently, only ChatGPT Plus users have access to special functions such as Advanced Data Analysis and Browsing. This limits the current chat’s reach to these users.
Hey @PunkPink, love that you made this. I do have a question however before voting. What precisely are the fees for? I don’t quite understand. Is that for the work put into the improvements, is to cover the cost of Plus subscription, or ??? Thank you.
I have carried out this poll basically to know the opinion of the community regarding this, so if the feeling of the community is good, I would present a grant to implement the improvement indicated in the result of this proposal.
$480 would be the cost of improving it in a simpler way and $3,200 would be the cost of improving it in a more complete way.
To define these budgets I have taken into account a salary of $20/hour * 24 hours (3 days of work) and 160 hours (20 days of work)
I assume the cost of my ChatGPT subscription and I do not include it here nor do I plan to include it in the grant (if it were published), but you should know that the indicated amounts do not reflect the cost of the subscription that each member may have. , each user who wants to access non-free ChatGPT functions and use DCL-Docs must subscribe to an OpenAI subscription on their own.
I don’t know if I have answered your question correctly, if you have any more questions let me know.
I’m voting yes to get this to the next round… could you elaborate on your next proposal the difference between the two costs? I see that one will be simpler and the second will be more complete, but can you break it down further as to what that actually means? What can a $3,200 product give us that a $480 cannot? Thank you!
Hello @Canessa, yes, in the future proposal everything will be much more detailed, but basically the difference lies in how ChatGPT will be instructed to extract information. In the first case (the $480 one), all relevant URLs would be provided for ChatGPT to extract information from the websites. This is a simple task, but in some cases, it might not be able to deliver the required information due to the structure of each website. For instance, in some tests I did with sites on Notion, ChatGPT was unable to access the information, making it unable to respond to questions related to those sites.
In the second case ($3,200), there will be a more exhaustive collection of all URLs containing relevant and reliable information about Decentraland (I mean from all official sources), followed by the extraction of texts from these sites, part of this procedure can be done with coding. The next step will be to compile all the information and process it to produce documentation that ChatGPT can easily interpret.
To define these budgets, I have taken into account a salary of $20/hour * 24 hours (3 days of work for the first option) and 160 hours (20 days of work for the second).
I hope I have been able to clarify your doubts as best as possible.
Its working pretty well as far as my testing. I think for coding in SDK7 , learning about the ecosystem , or fetching the latest DAO news or proposals it does the job.
So maybe we can save on costs for this particular grant. @Canessa@Existential14@PunkPink
however there may be more data fed into it, particularly if custom actions are required. But as a general tool it works fine for everyone that needs to interact and learn from the official documentation and resources.
Hello @Eibriel, the reality is that it would be much more accessible if a subscription were not required, however I trust that the accelerated advancement of this industry will make these functions accessible to the general public very soon.
