Promising new tech has ‘staggeringly difficult’ copyright problem: expert

Common artificial intelligence platforms sometimes return results that plagiarize copyrighted content, and this problem may not be easily resolved.

“This is a surprisingly difficult problem to solve because current AI can't look at a blank canvas and come up with something original,” said Christopher Alexander, chief analytics officer at Pioneer Development Group, on Fox News. told Digital. “Instead, the generative AI utilizes existing images and follows human instructions. The only real possibility to counter this is to use other AI capabilities to counter plagiarism and automate That’s it.” [copyright] Strike accounts of repeat offenders. ”

Alexander's comments come after a report in IEEE Spectrum magazine revealed that AI plagiarism continues to be an issue in both popular language learning model (LLM) platforms such as ChatGPT and imaging platforms such as Midjourney V6. Announced.

Experts say AI development will “explode” in 2024

ChatGPT's apparent plagiarism issues led to a lawsuit from the New York Times against the platform's parent company, OpenAI, accusing the company of using copyrighted works.

As an example of the lawsuit, the New York Times shows a screenshot of a ChatGPT article that copied the New York Times article almost verbatim, raising legal questions about how platforms can be held responsible for violating copyright restrictions. is occurring.

OpenAI did not respond to Fox News Digital's request for comment at the time of publication.

“The core issue here is that these LLMs scraped the data without caring about ownership in the first place,” Aiden Buzzetti, president of the Bull Moose Project, told FOX News Digital. “Yes, it's a travesty that people may be infringing copyrights on works they've never heard of before. The burden is on them to pay royalties, pay lip service, etc. The legal liability rests with the company that collected the data without developing any safeguards as to whether the material could be included in the data.”

Economist warns new technology could make a wide range of high-skilled jobs 'obsolete'

ChatGPT's apparent plagiarism issues led to a lawsuit from the New York Times against the platform's parent company, OpenAI, accusing the company of using copyrighted works. (Jaap Arrians/NurPhoto via Getty Images)

However, IEEE Spectrum points out that LLM is not the only one that has been victimized by plagiarism; image generation platforms such as Midjourney V6 also return plagiarized results.

Using Midjourney V6, the authors of the report were able to replicate near-identical images from popular movies such as “The Avengers.” In another example, an author instructs her AI to play the popular 90's cartoon “yellow skinned – v 6.0 –ar 16:9 -“, giving the platform an approximation of the TV show “The Simpsons”. I was able to get it to return the same image. raw style. ”

“Given these results, we believe that Midjourney V6 was trained on copyrighted material (which may or may not be licensed) and that the tool was trained on copyrighted material to produce infringing output. It seems almost certain that it may be used,” the authors note.

Midjourney V6 did not immediately respond to Fox News Digital's request for comment.

New technology will improve traffic flow in big cities, experts say

ChatGPT app (Olivier Morin/AFP via Getty Images)

Phil Siegel, founder of the Center for Advanced Preparedness and Threat Response Simulation, told Fox News Digital that the more detailed the prompt, the greater the chance of plagiarism because the AI ​​has fewer words to choose from for its response. .

“Simple prompts are less likely to cause plagiarism, but crisp, focused prompts are much more likely to cause plagiarism. A way to think about this is from Find a Fun Vacation.” questions like that, and there's a lot of training data to pull from. It's probably less likely to be plagiarized,” Siegel said. “If you ask a specific question like 'Find the best water sports for your Aruba vacation,' fewer words will be selected in your response, so you're more likely to get your response copied.” .”

Meanwhile, Samuel Mangold Rennett, an editorial staff member at The Federalist, said there are ways to resolve the issue, but doing so would likely hinder development.

CLICK HERE TO GET THE FOX NEWS APP

“Generative AI and LLM operate within a secret ecosystem known as a “black box.” “Developers often ignore intellectual property by immersing systems in these environments in order to saturate and harden them with data,” Mangold-Renett told Fox News Digital. . “This can be ‘solved’ by enforcing transparency, but doing so may slow down the development of AI.”