The popular artificial Intelligence platforms may return results which plagiarize content that is protected by copyright. This problem can be difficult to solve.

Christopher Alexander, chief analytics officer at Pioneer Development Group told Fox News Digital that this is an incredibly difficult problem to solve. Current AI can’t look at a white canvas and come up with something new. Instead, generative AI uses existing imagery to create a new image and then follows the prompts of a human. To combat this, the only way to be effective is to use AI to detect plagiarism and to automate [copyright] strikes on accounts that repeat offenses.

Alexander’s comments follow an IEEE Spectrum report that outlined the continuing problems with AI plagiarism in popular language-learning model (LLM), platforms like ChatGPT , and imagery platforms such Midjourney V6.

EXPERTS PREDICT AI DEVELOPMENT TO “EXPLODE” IN 2024

The New York Times has filed a suit against OpenAI for allegedly using works that are protected by copyright.

The New York Times uses a screenshot from a ChatGPT story that almost verbatim copies an article in the New York Times. This raises legal questions regarding how platforms could be held accountable for violating copyright laws.

OpenAI had not responded to a Fox News Digital comment request by the time of publication.

Aiden Buzzetti told Fox News Digital that the core problem is that the LLMs scraped the data without care of ownership. It’s an outrage that people could be infringing on the copyrights of works they have never heard of. It is the responsibility of those companies who scraped information without paying royalties or lip service, and have not developed any safeguards to protect themselves from legal liability.

ECONOMIST WARNS THAT NEW TECH WILL MAKE HIGHLY SKILLED JOBS OBSOLETE

IEEE Spectrum points out that Midjourney V6 and other image- generation platforms are also guilty of plagiarism.

The authors of the report used Midjourney V6 to reproduce nearly identical images from popular movies like “The Avengers.” The authors also used the platform to produce nearly identical images from the TV series “The Simpsons.” They did this by asking the AI to “popular ’90s cartoon with yellow skin — v 6.0 –ar16:9 –style Raw.”

The authors state that “in light of these results it is almost certain that Midjourney V6 was trained using copyrighted material (whether they were licensed or not, we don’t know) and their tools could have been used to produce outputs that violate.”

Fox News Digital’s request for a comment from Midjourney did not receive a response immediately.

NEW TECH PROMISES TO IMPROVE TRAFFIC FLOW IN MAJOR CITIES, EXPERTS SAY

Phil Siegel, founder of the Center for Advanced Preparedness and Threat Response Simulation at Fox News Digital, said that more detailed prompts make plagiarism more likely, as the AI has fewer words from which to choose for a reply.

It is more likely for a targeted, crisp prompt to cause plagiarism than a simple one. A question such as ‘Find a fun vacation for me’ would have a lot of training data and be less likely to cause plagiarism,” Siegel explained. Asking a question such as “Find me the most fun water sports to do on vacation in Aruba” might make it more likely that someone will copy the answer because there are fewer words to choose from.

Samuel Mangold Lenett, an editor at The Federalist said that there was a way to solve the problem, but that it would likely hinder development.

Click Here to Download the FOX News App

“Generative artificial intelligence and LLMs work within a secret ecosystem called ‘black box’. Mangold-Lenett, Fox News Digital reported that developers often ignore intellectual property when they immerse their systems into these environments to saturate them and strengthen them with data. This can be “solved” by forcing transparency. However, if done, this could slow AI developments.