FindYourNextBook.AI

[amber dagger]

Our goal is to use AI to inspire readers to read. Tell our book recommendation bot what you want to read and we'll find you the perfect match. It's the best book recommendation service on the internet today!

https://Findyournextbook.ai?utm_source=openai_discord

[fierce moss]

Works really well; an amazon or audible link would be a convenient feature to add.

Your fastest route to SEO dominance would probably be acquiring or merging with the current leader for searching "book recommendations," which is whatshouldireadnext(.)com. Looks like they gained rank through quotes and dedicated pages/lists for each author, as well as name recognition.

[amber dagger]

Thank you this is really helpful. Adding Amazon links / more books is the next thing we're going to do. My free dataset to build this with did not include the right identifier for Amazon affiliate links but we're upgrading the dataset this weekend.

@fierce moss how do the quotes/dedicated pages help with SEO? I was essentially planning to reach out to book bloggers and reports to try to get back links.

[supple charm]

Just curious, what does AI permit in this project ?

I mean you use "embeddings", does it means that you send the book summary to AI and then you match it with a description of what someone likes to read ?

[amber dagger]

I store embeddings of all of the summaries. I convert the user query to an embedding and find the most similar summary, that's the recommendation. That information is piped into a prompt to write out the recommendation

[fierce moss]

Having quotes and dedicated pages for each author can help with SEO because it provides more content for search engines to index on your website. This can help increase the relevance of your website for specific search queries related to those authors, which can in turn improve your ranking.

In addition to reaching out to book bloggers and reporters, there are a few other strategies you can try to improve your SEO:

• Optimize your website's on-page elements, such as the page titles and headings, to include relevant keywords and phrases.
• Use social media or targeted campains (producthunt, hackernews, reddit, twitter) to promote your website and create inbound links.
• Create high-quality, original content that provides value to users and improves the ranking strength with good site analytics.

I'll look forward to trying out the larger database, I ran into some dead ends and irrelevant recommendations for a few prompts - but that should be improved with 1m books. Davinci or Curie for the response text?

[olive jay]

Where can I find the link for your project? It sounds great.

[amber dagger]

I don't know why it's not showing as a hyperlink but it's at www.findyournextbook.ai

[amber dagger]

It's DaVinci. I'd love to use curie as it's faster and cheaper but I couldn't get it to sound very human with the few prompts I tried

[fierce moss]

Okay, yeah I bet Davinci will be worth it once you give it more context. It reads well but struggled a bit with interpreting the right subjects and giving a response.

[amber dagger]

The recommendations come from my database. I feed them to DaVinci to communicate. My database of books is very small today. I'm increasing it 15x by tomorrow

[languid dew]

Great job on this 🤘🏾

[frigid lily]

in terms of UI - "get another recommendation" suggests that I'm gonna get another recommendation based on the prompt I already entered, not a restart.

[amber dagger]

You're probably right

[amber dagger]

We've 10Xed our library and improved our response prompt today. I've observed much better recommendations but I'm curious if others have thoughts!!

[faint eagle]

hi , can i ask you how you feed large dataset to open ai ?

im working on a project i need to know how put large content to gpt without limitation

[amber dagger]

I wrote this blog to show how it works: https://medium.com/findyournextbook/find-your-next-book-building-a-book-recommendation-engine-with-openai-apis-4cf5a84ebe8a

[amber dagger]

Made some slight tweaks to the UI and added thumbs up /thumbs down feedback on the recommendation. Adding another 200k books tomorrow!

[vital steeple]

I really like the simplicity of it. Can't wait to check back tomorrow and see what it gives me!

[olive jay]

Thank you this is great, well done. How did you generate the UI?

@amber dagger do you have a repo for this code?

[amber dagger]

The UI is in a private repo. The front end is written in the Svelte framework. The back end is a python server that runs almost the exact code in the tutorial I posted but it is connected to databases instead of storing everything in memory, and I have a lot more books now.

[amber dagger]

https://medium.com/findyournextbook/thriving-in-the-ai-era-how-to-take-advantage-of-opportunities-created-by-gpt-51a87002995b

[olive jay]

Thank you!

[amber dagger]

We are up to 435K books now!

[simple tapir]

That's what I was thinking, too, when I first tried it out. I quickly got over it as the restart forced me to rethink my original prompt and I got something totally new as a result. Maybe have two buttons, one to restart the recommendation process and one to keep it going based on the previous recommendation?

[amber dagger]

Yep. We're a couple of programmers we have an actual designer putting together V2 now. You will be able to continue to scroll additional books for your original prompt and go back to reprompt.

Feedback is appreciated tho. Thank you!

Were thinking like Tinder style where you like or dislike the recommendation and it goes to the next one potentially

[simple tapir]

Really great job! I'm excited to see what books it recommends. I noticed that it recommends the same book if I put in the same prompt. Will this be randomized in the future or do I need to just change my prompt in order to change the recommendation?

Also, I like the idea of adding in the Amazon and Audible links. Looking forward to that feature add

[amber dagger]

Thank you! I want to add some randomness but I need figure out the right way to do it to make sure the recommendations remain this relevant.

[simple tapir]

For sure. I totally understand how that can be a challenge. Looking forward to what you come up

[amber dagger]

Happy Friday everyone! Here's the algorithms recommendation for a book about a few friends getting into weekend shenanigans. Hopefully your weekend is less creepy lol

[amber dagger]

We've added a buy on Amazon button below the recommendation. We have a big design change coming so we put it in the quickest way we could. If we make three sales it will open up the Amazon Products Api to us and help us make the experience much better!

[olive jay]

Just out of interest how did you not hit the rate limit for this with the thousands of summaries you've generated embeddings for?

[amber dagger]

What do you mean?

I've been working on writing automated blogs recommending some books on timely topics. I got it to a pretty awesome place: https://medium.com/findyournextbook/unlock-the-potential-of-ai-5-books-to-explore-the-implications-on-society-95fbba29d9d. This first blog is books about AI. The entire article was written with the prompt: openAI GPT is making breakthroughs and everyone is excited. This technology is so incredible!

[olive jay]

oh just the embeddings model is way cheaper than text completion but OpenAI has still established a monthly limit for API usage, it seems like your scaling your dataset and I was just wondering how this hadn't reached that rate limit yet (I think they have put it at $120 per month?)

[amber dagger]

I'm pretty sure you can change that $120 limit but yeah the embeddings are pretty cheap

[amber dagger]

Yesterday I found this example in my logs:

User: "Motherhood and Entrepreneurship"

Recommendation: A book about being a stay at home mother.

Any ideas how I can prevent issues like this? I understand how the AI could make that leap but it's concerning!

[olive jay]

I don't think it's feasible you could? ....

Is there a different way to construct the query that would weight different embeddings? Is it overly weighting 'motherhood' as opposed to 'entrepreneurship'?

[fierce moss]

I’ve been the target of an twitter advertisement by one of your competitors at bookabout.io 👀

[amber dagger]

Oh. How'd it compare?!

Ah I posted on Reddit on Jan 3rd and this group bought the domain on Jan 4th

Fun

[fierce moss]

When did you first ship it?

He lists multiple recommendations and on-page read more sections. The summaries are similar, on a confidence-score level I’m not sure - haven’t really used enough

[amber dagger]

I like their design alot. They don't respond with words like we do

[fierce moss]

Do you know how to build IOS apps? You might have more success driving engagement/retention with a dedicated app.

[fierce moss]

Looks like a well-planned out product and works well, so agreed I give them credit for their implementation.

[amber dagger]

Nice to have some competition/validation

[amber dagger]

It's been awesome to see the global community in the AI space. With no paid advertising we've gotten traffic from a majority of the globe:

[amber dagger]

I just made a Twitter Account for the site. I'd really appreciate some follows from the community ! https://twitter.com/FYNBook. Thank you!

[amber dagger]

We are officially over ONE MILLION books in the database. We should have a great recommendation for any request. 🥳

[cedar bear]

The question is, did you get rights from authors for this job

[amber dagger]

Is that the question? These books are in Public domain

[cedar bear]

wait how did you found 1 million public domain books. Let me know the source.

[amber dagger]

I mean that anyone can write a review about any book, goodreads can list any book, you don't need author permission because the book is a public item.

[cedar bear]

Goodreads list books because authors uploading.

my friend made a project like this he only used title isbn numbers and book covers, no pirating etc he is still in prison its been 8 years, he got 498 years or something for copyright violation

[amber dagger]

Please link to the news article where this was reported I'd be super interested to learn about it

[amber dagger]

The ToS for the company I get the data from approves this use case. I wonder if your friend stole the data.

[amber dagger]

Thank you AI for Beginners for the great explainer video!

[golden stratus]

@amber dagger

gitzalytics's message blocked by AutoMod

We are committed to fostering a positive atmosphere and have implemented additional filters to block any inappropriate or disruptive language and topics.

Reminder: Refrain from bypassing the AutoMod filter as per server Rule #10.

[amber dagger]

https://twitter.com/FYNBook/status/1619881142320242689?t=b3j8wS8KB1gL5Pa3k5dgSA&s=19

FindYourNextBook.ai (@FYNBook)

Hey #Niners fans, we've got just the book for you!

#NFCChampionship

Twitter · Jan 30, 2023, 2:12 AM

[cedar bear]

its not about ToS its about laws in usa

FBI doesn't care about ToS of a site.

[amber dagger]

I'm getting very close to being able to mass produce articles like this: https://twitter.com/FYNBook/status/1621576656266858497?t=oha_LRbH9iKF7BfnsuURCg&s=19

FindYourNextBook.ai (@FYNBook)

Just posted our AI's three recommended reads for Black History Month:

https://t.co/wm36RmG1bp

#BlackHistoryMonth #Bookish #booktwt

Twitter · Feb 3, 2023, 6:29 PM

[unreal nest]

https://ai-librarian.teleporthq.app

I made this design, what do you guys think? (No mobile responsive yet)

[amber dagger]

This is really awesome. Thank you! I'm going to DM you later today.

[magic python]

I learn how to use embeddings reading your article! very nice, thanks!

as a user, I maybe prefer to list the books i like about some topic

like: I read and enjoy the books A,B and C, what should i read now...

[quiet kelp]

Can I ask what you’re really expecting to be recommended when your input is books you’ve already read? Books with similar genre? Plot? Author? Feeling? Length? All or something else?

[magic python]

Like, when I read about Shakleton, I would like to read another real adventure. A real person suggested me Papillon, and was the kind of book I was searching for next reading exactly

[twilit galleon]

The Use and Misuse of Language

[twilit galleon]

This was a huge help for my current understanding of lexical communication and analysis.

[amber dagger]

ANNOUNCEMENT: We have just released our GA Product.

• Improved site design
• Ability to keep receiving recommendations for your initial query (a frequent ask)

Try it out and let me know what you think!

[wooden ivy]

1111

qqq

[minor moth]

@amber dagger Can we use semantic search on the books summary to find the book nearest to the user query?

[minor moth]

@amber dagger let's say if users wanted to find out in which book some particular part of the story is written, then how can we do that?

[amber dagger]

This is how it works

[amber dagger]

I don't totally follow this

[minor moth]

I was talking about a feature where users can ask some type of questions and we can get the answers from the database that we have.

[frosty pebble]

Hey! I like this a lot. I think you're onto something adapting the summaries to fit the user's query.
I've been using a similar approach to combine user queries + ads to run relevant ads on AI apps.
Could be a good way to supplement the amazon affiliates links, since davinci's expensive!
Example:

[amber dagger]

Omg that's awesome! @frosty pebble

[frosty pebble]

Thanks! This is my project: https://openads.ai/
Even if you're not ready to run ads, it has a "google analytics for prompts" that'll let you explore what your users are searching for by topic:

[minor moth]

@amber dagger one question: what's the purpose of creating embeddings? Can't we simply give the entire database as a prompt and then GPT will take care of rest?