Attribution For Artificially Produced Content
Now, we return to our main theme of attribution for social media posts. Already, you may have been tempted to install one of the ChatGPT plugins such as AI Engine on your blog or website or Compose AI, a Chrome browser extension, or AI Connector usable on Facebook, or WhatsGPT or Kidix, or BuddyGPT for Whatsapp. Such tools can cut down on the time necessary to produce content, but they cannot eliminate, given current capabilities, unintended hallucinations (fiction,) and copyright infringement risk, the need for human supervision. Likely, you have already been automatically producing content. But have you been fact-checking and attributing your content?
Why attribution?
There are several reasons why attribution is important for AI-generated content. First, the interpretation of copyright and database rights is not entirely finalized concerning AI. Then, it helps to ensure that the original creators of the data and/or algorithms used to train the LLM are given credit for their work. There may be some residual rights accruing to the LLM prompter due to the skill, effort, and creativity applicable to in-context learning (ICL) and producing prompts. Fourth, attribution is important for both ethical and legal reasons as the law now stands. Further, attribution can assist in building confidence and transparency around AI-generated content. When users are cognizant of how the content was created they can better understand its limitations and potential biases. Lastly, attribution, as a cautionary sign, can help to deter plagiarism and copyright infringement. Note, the European Parliament is the first to propose AI legislation. It’s draft legislation on Generative AI requires attribution or disclosure to combat disinformation as described in Title IV, Transparency Obligation For Certain AI Systems, Article 52:Generative AI, like ChatGPT, would have to comply with transparency requirements:
· Disclosing that the content was generated by AI
· Designing the model to prevent it from generating illegal content
· Publishing summaries of copyrighted data used for training
–“EU AI Act: first regulation on artificial intelligence.” European Parliament President’s Website, News European Parliament, https://www.europarl.europa.eu/news/en/headlines/society/20230601STO93804/eu-ai-act-first-regulation-on-artificial-intelligence
When is attribution required?
In general, attribution is essential whenever AI-generated content is used in a way that could be considered original, creative, or a derived work. This is of special importance when the content is published or shared, particularly on social media, or when it is used to make decisions that could have a significant impact on society or someone.Attributing AI-generated content
- [Generated/Translated/Created] by [LLM name] from/powered (for translations) by [provider company]
- [Created/Generated (for images)] with the assistance of [LLM name]
When attributing AI-generated content, it is important to be clear, concise, and accurate. It is also important to provide enough information so that users can understand the limitations and potential biases of the AI, with the aim of improving confidence in the technology. Above all, the laws regarding intellectual property should be kept in mind. Next time we look at prompt engineering.
–Richard Thomas
Previous, Part III/V
Next, Part V/V
