A few months have passed since the inception of the AI Summaries project. I'd like to share an update on its progression, what's been achieved so far, and the roadmap ahead.
Longform Podcast Summaries
The project's core initiative was to summarize community podcasts, including InLeo AMA, Lion's Den, CTT Podcast, Cryptoholics, and Hive Town Hall, translating this valuable audio content into accessible written formats that can be translated and generally be made accessible to more people. To date, 34 full-length summaries have been published as regular blog posts thus far.
Expanding into video Content on 3Speak
After doing podcast-summaries for a while, I was approached by a channel owner looking to summarize the videos in their 3Speak-channels, first and foremost to add it to the database for AI indexing and SEO optimization purposes. After some development work and a significant amount of computing power, the project now includes summarizing content from some of the most influential 3Speak video channels. Summaries are systematically posted in the comments sections of each video. So far we've summarized
- Approximately 3000 videos on @taskmaster4450’s two channels,
- 637 videos from @jongolson’s channels, including all the Cryptomaniacs episodes,
- Over 900 videos from @bradleyarrow’s channel.
Currently being processed is:
- The entire channel of @theycallmedan and @cttpodcast
- In addition we're doing the approximately 100 videos from @selfhelp4trolls' channel.
*Note: All channels mentioned are being processed in agreement with the channel owners.
Access to Newer 3Speak Videos
To be able to transcribe videos it's necessary to be able to download them, but because of a restriction on downloading newer video content from 3Speak, I had to reach out to the team and request access. The team proved to be very positive towards the project and provided access to download the content from the highlighted channels.
Request From the 3Speak Team
After my initial request, a representative of the 3Speak team reached out to me, expressing excitement for the project and with a request to process the CTT Podcast episodes and the videos in the TheyCallMeDan channel. This with the intention of both summarization and making it more accessible, but also for potential future use-cases.
I've been a listener to the videos/livestreams in these two channels for years, andis very grateful to get the chance to collaborate with them on such a big and important job. The potential use-cases of the enormous amounts of data contained within these episodes in written form, are huge.
Hive Dictionary Initiative
A side project emerged as a consequence of the generation of the approximately 15 million words transcribed so far; the Hive ASR Dictionary. A simple program to enhance the accuracy of Automated Speech Recognition (ASR) systems in processing Hive-specific terminology. I will do a separate write-up on this neat little program. Here's a teaser though:
https://inleo.io/threads/view/mightpossibly/re-leothreads-33apr98ea
The Power of Quality Transcripts
All transcripts are time-coded, ensuring they're ready for future features, such as Closed Captions on 3Speak and potentially for building Hive-specialized datasets for AI training. This not only serves as an invaluable asset for channel owners but also paves the way for future technological enhancements, like more advanced LLMs and increased computing power. Not to mention increasing the value of Hive as a whole.
Value Proposition
This initiative significantly contributes to the Hive Database, feeding projects like LeoAI with structured text data, which is crucial for content analysis. It also makes the ecosystem's rich content more accessible to individuals facing language barriers or disabilities, embodying the spirit of inclusivity and accessibility that Hive stands for.
Moving Forward
The AI Summaries project was born from dialogue and idea exchanges with community members, and is a testament to the synergy between individual initiative and collective benefit. It underscores a commitment to enhancing content accessibility and inclusivity within our ecosystem, while at the same time adding large amounts of valuable data to the database which is Hive, thus increasing its value.
I'm very excited about continuing this journey to see where it will lead me (and us). Your thoughts, suggestions, and engagement is – as always – most welcome.
Thank you for reading!
What is Hive?
To learn more about Hive, this article is a good place to start: What is Hive?. If you don't already own a Hive account, go here to get one.
@leoglossary links added using LeoLinker.
Posted Using InLeo Alpha
It is important for community members to get involved with different things to bring more value to the platform.
Something like this adds enormous value especially in the era of learning models and other AI related algos.
We need as much data possible.
I completely agree. Thank you for your support and encouragement to keep pursuing this.
I am a huge advocate of more data.
The basic premise is data wrapped in algos combined with compute.
Hence the infrastructure is data.
So AI will just do everything then o.o we could probably have ai among the users and we won't know it soon lol