AI-Summaries Development Update #2

This is the second edition of the @ai-summaries development logs. Here you will get the latest updates about the AI-summaries project and various statistics and information about the activities since the last update

What is AI-Summaries?

The overarching goal of the project is to add as much high quality data as possible to our common database, which is Hive. It is an effort to add a ton of valuable text-data to the chain that would otherwise not be on it. The philosophy of the project in the form of a quote:

If data is the new oil, why give it all away to big tech? @anderssinho

The way we've been going about this was simple initially; Transcribe and summarize HIVE and LEO-related podcasts that was only available in audio format and post it to chain. It was then expanded to summarize entire (select) 3Speak channels. At this point we already had thousands of pages worth of content added.

But as cool as all that may sound, the latest and most exciting development, is the youtube summarization service:

The Summarizer

This is an AI agent that can be called upon by issuing the command !summarize together with any youtube video link in anything you post to chain. The agent will then recognize your command, summarize the video for you and post the summary as a series of replies to your comment/post.

Here is an example of how it comes out:

Link to summary: /@taskmaster4450le/re-taskmaster4450le-2a9qd5mst?referral=taskmaster4450le

How to get access

The summarizer is FREE TO USE! No need to subscribe anymore

~~To subscribe you can either~~
~~- subscribe through the INLEO front-end, or~~
~~- transfer 5 HBD to @leosubscriptions with memo subscribe:mightpossibly.~~

~~You should get access immediately after subscribing - if not, please tag me. After subscribing you will have unlimited access to the tool for 31 days.~~

~~After subscribing,~~ you can use it anywhere on Hive by following these simple instructions: (TL;DR: Post anything containing the !summarize command + a youtube video link)~

/@mightpossibly/re-leothreads-cjhbc6ka?referral=mightpossibly

It's convenient, it's fun and it benefits the ecosystem as a whole. Even if you don't intend to summarize a bunch of videos, your support helps further the development of the project.

Democratization of Data

In short, the idea is to provide an easy to use and effective way to democratize data by putting it on the blockchain. If this is the first you're hearing about the democratization of data and the decentralization of AI, I recommend giving this excellent article by @taskmaster4450le a read, where he also discusses the significance of this tool in that context.

https://inleo.io/@taskmaster4450le/aisummaries-agent-helping-to-move-the-database-ahead-6p3?referral=taskmaster4450le

Development Updates

Recently implemented changes

Retry videos that failed to summarize initially
Fixed bug that causes the posting bot to stall when parent comment has been deleted

Planned Features

Leaderboard
A front-end to prepare whole channels for posting (insert channel name > get all video urls with proper tagging ready for copy/pasting)
Highlighting of unposted videos and greying out videos that have already been summarized

Ideas (not on roadmap)

A front-end to search for and view summaries, view entire channels, request channel summarization
Dynamic, true NFTs utilizing the @darkcloaks framework for Layer 1 NFTs on Hive, where you can store your summarization achievements in an immutable badge of honor (and possibly more)
API for developers to be able to fetch various summary data, like summary links, who summarized it etc.

Stats and Activities

Here is an overview of various project activities since the last update

Hive/LEO Livestream Summaries

Youtube Summarizer Stats

Total number of yt-videos processed: 25,000+
Total number of comments posted: 195,000
Total Output Tokens posted to chain: 20,000,000

For your reference, here's what 20 million output tokens approximately equates to:

A typical novel contains roughly 70,000-100,000 words. Let's use 85,000 words as an average. In terms of tokens, English text typically converts to about 1.3 tokens per word (this varies based on the specific text, but it's a reasonable approximation). So:

85,000 words × 1.3 tokens/word = ~110,500 tokens per novel

Therefore, 20 million tokens would be approximately equivalent to:

20,000,000 ÷ 110,500 = ~181 full-length novels

So 20 million output tokens would be roughly equivalent to 180-185 typical novels. This is a significant amount of text - comparable to a small library's worth of fiction.

180 full length novels in 1.5 months 🎤 💧

Learn More

https://inleo.io/@mightpossibly/introducing-ai-summaries-for-hive-and-leo-related-livestreams?referral=mightpossibly

https://inleo.io/@mightpossibly/aisummaries-weekly-report-1-bgn?referral=mightpossibly

Want to contribute? The best way to support it is to subscribe and put the Summarizer to work. There is a near infinite source of information on youtube, and now there is an easy way to tap into it to benefit the value of the network as a whole.