Adobe's New Video Editing Features: Integrating Third-Party AI Models and the Ethical Implications

San Francisco, California United States of America
Adobe is integrating third-party AI models into Premiere Pro for Object Addition & Removal and Generative Extend features.
Adobe's own Firefly model and third-party models like Pika and Sora from OpenAI are being considered.
Data security and privacy concerns arise with the integration of third-party models.
Integration of third-party models could increase accessibility to creative professionals.
Use of generative AI for object removal may result in unwanted artifacts or inconsistencies.
Adobe's New Video Editing Features: Integrating Third-Party AI Models and the Ethical Implications

Adobe, a leading software company, is exploring partnerships with artificial intelligence (AI) companies such as OpenAI and others to enhance its video editing software Premiere Pro. The new features include Object Addition & Removal and Generative Extend, which will allow users to add or remove objects in videos using text prompts and extend the length of clips respectively. Adobe's own generative AI model, Firefly, is being used for these features. However, the company is also considering integrations with third-party models like Pika and Sora from OpenAI. These tools could provide users with more choice and flexibility in their video editing projects.

Adobe's move to integrate third-party AI models follows a trend in the industry as both free and paid training data for AI companies is running out, leading to a potential crisis by 2026. Some companies have even resorted to generating synthetic data for training large language models (LLMs). However, this approach raises legal concerns as lawsuits have been filed against chatbot makers for using unlicensed material and shadow libraries.

Despite these challenges, Adobe's primary advantage lies in its extensive use of Firefly models within its software ecosystem. The integration of third-party AI models could benefit both Adobe and the AI companies by increasing accessibility to creative professionals who might not otherwise use separate dedicated apps.

It is important to note that while these new features offer exciting possibilities, they also come with potential risks. For instance, the removal of objects using generative AI may result in unwanted artifacts or inconsistencies. Additionally, the integration of third-party models could lead to issues related to data security and privacy.

Adobe's history of paying artists for content used to train its AI models raises questions about the ethical implications of using such data. The company has not disclosed whether it will continue this practice for its new video editing features. It is crucial that Adobe and other companies in the industry address these concerns transparently and responsibly as they develop and deploy new AI-powered tools.



Confidence

91%

Doubts
  • Is there a risk of unwanted artifacts or inconsistencies when using generative AI for object removal?
  • What are the potential data security and privacy concerns with integrating third-party models?

Sources

99%

  • Unique Points
    • Adobe is developing generative AI video tools for Premiere Pro called Firefly.
    • Adobe plans to apply Content Credentials labels to generated clips to identify which AI models have been used.
    • Third-party integrations with Runway, Pika Labs, and OpenAI’s Sora are being explored for Premiere Pro.
  • Accuracy
    • , Adobe plans to apply Content Credentials labels to generated clips to identify which AI models have been used.
    • , Adobe is exploring partnership with OpenAI and other third-party generative AI tools for its video editing software Premiere Pro.
    • Adobe plans to add AI-based features to Premiere Pro, such as filling in parts of a scene with AI-generated objects or removing distractions.
  • Deception (100%)
    None Found At Time Of Publication
  • Fallacies (100%)
    None Found At Time Of Publication
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (100%)
    None Found At Time Of Publication
  • Author Conflicts Of Interest (100%)
    None Found At Time Of Publication

66%

  • Unique Points
    • Adobe’s Firefly image-generating software was trained in part on AI-generated content from rivals.
    • Adobe did not disclose in numerous presentations and public posts that its model used images from these same competitors.
  • Accuracy
    • Adobe's Firefly image-generating software was trained in part on AI-generated content from rivals.
    • Competitors like Midjourney train their models by scraping pictures from the internet.
  • Deception (25%)
    The article reveals that Adobe's AI Firefly was trained on images from its rivals without disclosure. This is an example of selective reporting and deception by omission.
    • Adobe said, the artificial intelligence model was trained mainly on Adobe Stock... But behind the scenes, Adobe also was relying in part on AI-generated content to train Firefly...
  • Fallacies (75%)
    Adobe's authors, Rachel Metz and Brody Ford, commit an omission fallacy by failing to disclose that Adobe used AI-generated images from rivals for training Firefly. They also use inflammatory rhetoric by implying that Adobe's competitors are unethical for scraping pictures from the internet.
    • Adobe said, the artificial intelligence model was trained mainly on Adobe Stock, its database of hundreds of millions of licensed images.
    • But behind the scenes, Adobe also was relying in part on AI-generated content to train Firefly.
    • In numerous presentations and public posts about how Firefly is safer than the competition due to its training data, Adobe never made clear that its model actually used images from some of these same competitors.
  • Bias (50%)
    The authors of the article imply that Adobe misrepresented the training data for their AI model Firefly by stating that it was mainly trained on Adobe Stock, while in reality they also used images from their competitors. This is an example of monetary bias as Adobe may have had a financial incentive to downplay or hide the use of competitor's data to make their product appear more superior.
    • Adobe never made clear that its model actually used images from some of these same competitors.
      • But behind the scenes, Adobe also was relying in part on AI-generated content to train Firefly, including from those same AI rivals.
      • Site Conflicts Of Interest (100%)
        None Found At Time Of Publication
      • Author Conflicts Of Interest (100%)
        None Found At Time Of Publication

      99%

      • Unique Points
        • Adobe is exploring partnership with OpenAI and other third-party generative AI tools for its video editing software Premiere Pro.
        • Adobe plans to add AI-based features to Premiere Pro, such as filling in parts of a scene with AI-generated objects or removing distractions.
        • Firefly, an AI model already deployed in Adobe’s Photoshop software, will be used for these features.
        • Adobe is developing a way to let users tap third-party tools from OpenAI, Runway and Pika Labs to generate and use video within Premiere Pro.
        • OpenAI has demonstrated its Sora model generating realistic videos based on text prompts but has not made the technology public or given a timeline for availability.
        • Adobe released a demonstration of Sora being used to generate video in Premiere Pro, describing it as an ‘experiment’ with no timeline for availability.
        • Revenue generated by third-party AI tools used on Adobe’s software platform is yet to be settled between Adobe and outside developers.
      • Accuracy
        • , Adobe plans to add AI-based features to Premiere Pro, such as filling in parts of a scene with AI-generated objects or removing distractions.
        • , Adobe is developing a way to let users tap third-party tools from OpenAI, Runway and Pika Labs to generate and use video within Premiere Pro.
        • , No release date has been set for Adobe’s new video generation tools.
        • , OpenAI’s Sora has superior generational capabilities compared to other offerings.
      • Deception (100%)
        None Found At Time Of Publication
      • Fallacies (100%)
        None Found At Time Of Publication
      • Bias (100%)
        None Found At Time Of Publication
      • Site Conflicts Of Interest (100%)
        None Found At Time Of Publication
      • Author Conflicts Of Interest (100%)
        None Found At Time Of Publication

      93%

      • Unique Points
        • Adobe is paying artists $120 to send videos for training its AI models
        • Adobe’s goal is to use these videos to train its AI tools for text-to-vector video generation and video editing capabilities
        • AI companies are facing a data crisis as both free and paid training data runs out, with some predicting it could be depleted by 2026
        • Lawsuits have been filed against chatbot makers for training their models on unlicensed material and shadow libraries, including against OpenAI by several authors and news outlets
        • French regulators fined Google over a similar issue last week
        • Congress held a hearing in January to discuss whether authors and artists should be paid for use of their content
      • Accuracy
        • Adobe's goal is to use these videos to train its AI tools for text-to-vector video generation and video editing capabilities
      • Deception (80%)
        The article reports on Adobe's practice of paying artists for content to train its AI models. While this information is factual, the author uses emotional manipulation by implying a sense of urgency and precariousness in the industry due to the depletion of training data. The author also selectively reports details that support her position, such as Adobe's goal to use videos for text-to-vector video generation and video editing capabilities, without mentioning any potential drawbacks or opposing viewpoints. Additionally, there are references to lawsuits against chatbot makers for using unlicensed material without consent, but the author does not provide any context or details about these lawsuits.
        • The company is now paying artists $120 to send them videos.
        • Adobe is one of a few companies that pay contributors to submit content to train their AI models.
        • Business Insider reports that Open AI has even considered generating its own synthetic data to continue training large language models, or LLMs, which power AI chatbots.
      • Fallacies (95%)
        The article mentions several facts about Adobe's efforts to obtain training data for its AI models and the challenges faced by the industry. There are no explicit fallacies found in the author's statements. However, there is an appeal to authority when Bloomberg is cited as a source of information about Adobe paying artists $120 for videos and Adobe's goal to use those videos for training its AI tools.
        • Adobe is one of a few companies that pay contributors to submit content to train their AI models. Bloomberg, citing documents, reports that the company is now paying artists $120 to send them videos. Adobe’s goal is to use those videos to train its AI tools, Bloomberg reports.
      • Bias (100%)
        None Found At Time Of Publication
      • Site Conflicts Of Interest (100%)
        None Found At Time Of Publication
      • Author Conflicts Of Interest (100%)
        None Found At Time Of Publication

      99%

      • Unique Points
        • Adobe is introducing new features to Premiere Pro using its generative AI model, Firefly.
        • Object Addition & Removal feature allows users to select and track objects, then replace or remove them using Firefly’s generative AI.
      • Accuracy
        • Adobe is developing generative AI video tools for Premiere Pro called Firefly.
        • , Generative Extend feature in Premiere Pro uses AI to create extra media and make clips longer for easier editing.
        • , Object Addition & Removal feature allows users to select and track objects, then replace or remove them using Firefly’s generative AI.
        • , Adobe plans to apply Content Credentials labels to generated clips to identify which AI models have been used.
        • , OpenAI’s Sora has superior generational capabilities compared to other offerings.
      • Deception (100%)
        None Found At Time Of Publication
      • Fallacies (100%)
        None Found At Time Of Publication
      • Bias (100%)
        None Found At Time Of Publication
      • Site Conflicts Of Interest (100%)
        None Found At Time Of Publication
      • Author Conflicts Of Interest (100%)
        None Found At Time Of Publication