Digital Assistant for Needpedia.org

We made a wiki that lets people share their dreams for a better world, now we need an AI that does some of the work for them.

***We are not talking about a chatbot, it needs to be able to make posts for users. -Including parent posts when necessary.

On Needpedia ideas are listed for problems, and problems are listed by subject. This means it’s not even possible to share ideas until someone’s made the subject and problem [parent] posts.

This is a problem for us because users will lose interest halfway through creating parent posts and leave our site forever, annoyed. Which is where you come in. We need an AI that can generate those parent posts for users. -And idea posts.

Additionally, we need to be able to give the assistant instructions, (as PDFs or something, which we can upload on the backend), to tell it how we want it to do these things, and so that we can edit the instructions. -to improve them over time.

For instance, when a user wants to share an idea, the assistant should be checking the instructions we’ve provided for that scenario. Our instructions might say something like “If a user has an idea, identify what problem it is for, and what subject the problem is related to. Then do a search for similar ideas, and if there are any, provide the user with links to them. Ask: “Do any of these posts match what you had in mind?”.

Later on we might need to focus more on getting users excited, and change the instructions to add questions about what users think their idea’s social impact could be. The ability to edit the instructions is very important to us. It's fine if the instructions have to be structured or phrased a certain way to work best for the AI/tech, as long as the front end remains user friendly.

*This will require APIs to perform searches and create posts, our CTO Murtaza has already made a fair number of them and can make more. You will need to coordinate with him and that can take a day or two, so please take that into account now. This project is our top priority and we promise to get back to you as fast as we possibly can.

Another random development note is that the assistant probably needs to be able to do searches for content. We had a dev who wanted to have the assistant automatically learn and know everything posted to Needpedia, which meant that it would be constantly training on existing data. We don’t think our users would like having their data being used to train an AI, even if it’s for us, so the assistant needs to be able to search up content on Needpedia instead.

Parent posts are generally just placeholders so these posts can be extremely simplistic. The assistant needs to pop up and start the conversation, our CTO can help with that.

We’re also concerned that bots or bad actors will use up our server resources running the bot without creating an account, but we also want visitors to be able to have the assistant’s help figuring out how to use our site, so we want to strongly limit access for people who aren’t signed into an account. -It would also be good to put some kind of limit on people who are signed in too.

Additionally, we didn’t know we needed to say this but apparently there are apis to AI like openai that let you use their service but at a severely reduced capability, (as in, literally doesn’t know what a cat is). We want it to be roughly as capable as gpt4. If you can get better performance from a different AI that’s cool with us, I just need it to be smart enough to help users share ideas. (kinda hard for an AI to help people share ideas if it doesn’t know what “ideas” are). Just to give you an idea of what we’re talking about: https://docs.google.com/document/d/1hMgWzK1xqNtgVCc_MNlI0QiBMArwC0Oto1orCwzZpkc/edit?usp=sharing
^I’m not even kidding, that is the “best” anyone’s produced so far. We understand our budget for this project is modest, but it still has to work better than that).

We’re also very flexible about deadlines, as long as you’re actually working on the project. We don’t want to rush you and end up with something that doesn’t actually work. We hope to have a highly productive but low-stress development process, there’s just one catch: payment happens upon completion.

We’ve had every developer who’s seen this project claim they can do it, some quoting us for as little as $1,200, but then they don’t actually produce anything that can make posts. With people like that our entire development budget would get completely wasted.

***Please only apply if you’re sure you can pull this off.***

Every dev that tried to get their foot in the door and siphon off money “halfway” through has failed. Either plan on delivering what’s described in this document and getting paid in full, or plan on not getting paid at all. The money’s already in Freelancer’s hands, deliver the goods and we will gleefully hand you full payment. This project will complete Needpedia.org which we have been working on for years, believe me when I say we want you to succeed and will make every effort to support you.

And we promise to not add or change features or requirements midway, we’re not gonna make this hard for you. The description here and our mockups are intended to serve as the contract we’re entering into on Freelancer.com, so if you can do what we’re asking here, then I promise you this will be an easy and straightforward process. We will immediately pay in full upon testing 2 abilities:

-Can it create posts (idea, problem, and subject posts, including any parent posts if necessary)?
-Can it follow instructions we give it regarding how to respond to user-requests?

*Our CTO and I will also have to look over the documentation to make sure it’s good enough for us to be able to work with your code, Needpedia aspires to attract a large community of devs and documentation is extremely important to us.

Satisfy these conditions and we will promptly send payment in full, and make you the first person we contact when we start our next AI project. Our ideal budget for this project is $2,000.