1 / 3

Translate Plus Owner Publicis Joins Initiative to Authenticate Human-Produced Content

The Coalition for Content Provenance and Authenticity (C2PA), an initiative that aims to identify human-created content in a sea of AI, just gained a new member.

Slator
Download Presentation

Translate Plus Owner Publicis Joins Initiative to Authenticate Human-Produced Content

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Translate Plus Owner Publicis Joins Initiative to Authenticate Human-Produced Content The Coalition for Content Provenance and Authenticity (C2PA), an initiative that aims to identify human-created content in a sea of AI, just gained a new member. Multinational marketing and communications company Publicis Groupe announced in a June 5, 2023 press release that it would join the C2PA’s steering committee, whose ranks include Adobe, BBC, Intel, and Microsoft. Prior to Publicis, the most recent addition to the steering committee was Sony, in March 2022. The roster of more than four dozen members, at different levels, also lists Canon, Nikon, CBC/Radio-Canada, Panasonic, and The New York Times.

  2. Like other participants, Publicis has a vested interest in the C2PA’s goal of a clear and widely used standard for content authentication. Founded in February 2021 by Microsoft and Adobe, the C2PA predates the current hype cycle surrounding AI-powered content generation by ChatGPT and its ilk. Rather, the founding of the C2PA bridged separate but similar projects, the Content Authenticity Initiative (Adobe) and Project Origin (Microsoft and BBC). The C2PA’s membership reflects the range of industries and professions concerned about the effects of ubiquitous AI-generated content and misinformation; certainly, members’ status and size — including Publicis’ 90,000 employees — indicates a high-profile potential impact. While others aim to identify content as created by AI, the C2PA wants to create a “paper trail” that will identify human-made content as such. “Provenance helps people decide what’s trustworthy,” reads CAI’s website, which features a matrix to help readers decide which open-source tool would best suit their needs. The C2PA’s website explains, “Detecting whether or not digital content is fake is currently impossible at internet scale and speed because […] metadata can easily be manipulated and provides no proof of its origins.” In a demo, creator Hallease Narvaez uses the Adobe Firefly beta to create an original image, and explains that CAI content credentials, including her name, the date, tools used, and any edits, will follow her image as metadata wherever the final product goes. Also, log Users can export Content Credentials to the cloud so they can be verified even if someone else uploads or screengrabs the product. Translation: A Square Peg? Released in April 2023, version 1.3 of the C2PA’s technical specification expanded support for a range of media. And according to the C2PA’s FAQ page, support for new file formats, alignment with Generative AI, Live Video & Audio and more” are on the way.

  3. The C2PA expects users of its tools to include content creators, content publishers, content consumers, vendors, and implementers. Where do translators fit in? For that matter, would machine translation (MT) be considered AI-generated content? A May 2023 Harvard Library Innovation Lab blog post flips the C2PA’s take on provenance and explores how the tech behind the C2PA’s existing tools — capable of labeling image, video, and audio files — might “certify” text-based content as AI-generated. “Finding a suitable file format would be the first and probably main hurdle to overcome in order for large language models (LLMs) to label their output. PDF may be a good fit,” wrote Matteo Cargnelutti. He posited that the process would be similar to current workflows for other content: “Creating and signing provenance information at the time of generating the output, serving that information alongside the generated content, and keeping a copy of it on the server side. The end-user would be presented with options to download a copy of this output, which would come with embedded provenance information.” It is currently unclear how, if at all, MT will be treated by the C2PA and its affiliates, but it may become a point of focus now that Publicis is on board: The company acquired language service provider Translate Plus in 2017. “We will lead our industry in setting the standards for content verification. #GenerativeAI is a powerful tool, but also needs guardrails to protect against synthetic content,” Ray Lansigan, an Executive VP at Publicis, posted on LinkedIn.

More Related