ISCC Foundation

We're pleased to announce the release of ISCC-SCT. It generates semantic, language-independent codes for text, capable of matching translations, paraphrases, and granular text chunks. Explore the demo here: 🔗 huggingface.co/spaces/iscc/is… @mixedbreadai @JinaAI_ @nomic_ai

Today marks a pivotal moment in the digital media landscape with ISO's release of the #ISCC under #ISO 24138, a significant leap forward for digital transparency and trust in the dawning AI era! /1

There's big news on the content provenance front. The @iscc_foundation has a new standard out - ISO 24138 - which provides a link between an identifier for a work (like an ISBN or DOI) and the various forms that work can take online.

The Semantic Code is engineered to be robust against a broader range of variations that cannot be matched with the perceptual Image-Code. github.com/iscc/iscc-sci

On the ISCC Playground you can test the experimental Semantic Code of the ISCC. It is a compact binary code based on deep learning vector embeddings, which allows matching and comparison of semantic characteristics of content files. huggingface.co/spaces/iscc/is…

New development, new ISO stage (code 50.00), the Draft International Standard (DIS) was approved, the final text is now at ISO for further processing. – ISCC is now ISO/PRF 24138 – International Standard Content Code (ISCC). Follow the updates, here --> iso.org/standard/77899…

International Standard Content Code (ISCC), a content-derived identifier for digital content identification, is currently under development as ISO/DIS 24138 (Draft International Standard). On the ISO platform you can get a glimpse on the upcoming standard: #iso:std:iso:24138:dis:ed-1:v1:en" target="_blank" rel="nofollow noopener">iso.org/obp/ui/en/#iso…

@alexjc Totally agree, including the use of iscc codes to identify content as you say later. Add an ingestion time stamp for comparison with a possible tdm opt-out activation by the provider of the content (see w3c tdm reservation protocol).

𝙷 𝙾 𝚆 ? I'm not talking about providing per-item URLs like a dataset, but content codes or hashes for training data — along with some metadata such as the source (e.g. domain), size and mime-type, and the legal basis for accessing the file (e.g. contract, license, fair use).

I'm calling it! 🤙 It's inevitable that item-level training data disclosures will become a standard for AI companies that want to exploit or license models commercially. If you're operating a commercial service, you'll be expected to provide data manifests... What? Why? ⤵️

Interesting column by @PPLUK's CIO @mpdouglas, suggesting to further investigate the ISCC, which could support ‘clustering’ of same or near-duplicate content to help solving problems ranging from accuracy to fraud. musicbusinessworldwide.com/clustering-red…

Thank you! Iscc standard is on the radar btw.

Create ISCC codes for your creative works and digital media files: iscc.io Now with support for metadata embedding for PDF and EPUB files, and thumbnail generation for PDF, EPUB and audio files with cover images.

ISCC encryption & webtoon content, an upcoming revolution for data's reports? One "unconventional" use I am really looking forward to, for the digital comic / webtoon industry, is the possibility to track all the data's related to the use / reads / downl…lnkd.in/erX7t27r