Wikipedia

Artificial intelligence content detection

(Redirected from Artificial intelligence content detector)

Artificial intelligence detection software aims to determine whether some content (text, image, video or audio) was generated using artificial intelligence (AI).

However, the reliability of such software is a topic of debate,[1] and there are concerns about the potential misapplication of AI detection software by educators.

Accuracy issues

edit

Multiple AI detection tools have been demonstrated to be unreliable in terms of accurately and comprehensively detecting AI-generated text. In a study conducted by Weber-Wulff et al., and published in 2023, researchers evaluated 14 detection tools including Turnitin and GPT Zero, and found that "all scored below 80% of accuracy and only 5 over 70%."[2]

Text detection

edit

For text, this is usually done to prevent alleged plagiarism, often by detecting repetition of words as telltale signs that a text was AI-generated (including AI hallucinations). They are often used by teachers marking their students, usually on an ad hoc basis. Following the release of ChatGPT and similar AI text generative software, many educational establishments have issued policies against the use of AI by students.[3] AI text detection software is also used by those assessing job applicants, as well as online search engines.[4]

Current detectors may sometimes be unreliable and have incorrectly marked work by humans as originating from AI[5][6][7] while failing to detect AI-generated work in other instances.[8] MIT Technology Review said that the technology "struggled to pick up ChatGPT-generated text that had been slightly rearranged by humans and obfuscated by a paraphrasing tool".[9] AI text detection software has also been shown to discriminate against non-native speakers of English.[4]

Two students from the University of California, Davis, nearly faced expulsion after their professors scanned their essays with a text detection tool called Turnitin, which flagged the essays as having been generated by AI. However, following media coverage,[10] and a thorough investigation, the students were cleared of any wrongdoing.[11][12]

In April 2023, Cambridge University and other members of the Russell Group of universities in the United Kingdom opted out of Turnitin's AI text detection tool, after expressing concerns it was unreliable.[13] The University of Texas at Austin opted out of the system six months later.[14]

In May 2023, a professor at Texas A&M University–Commerce used ChatGPT to detect whether his students' content was written by it, which ChatGPT said was the case. As such, he threatened to fail the class despite ChatGPT not being able to detect AI-generated writing.[15] No students were prevented from graduating because of the issue, and all but one student (who admitted to using the software) were exonerated from accusations of having used ChatGPT in their content.[16]

In May 2024, students from the University of Pennsylvania published RAID, a benchmark for evaluating AI text detectors.[17] RAID includes over 6 million generations across 11 models and 11 adversarial attacks, revealing that current detectors struggle with adversarial strategies and unseen models. According to RAID benchmark leaderboard, Desklib AI Detector with an accuracy score of 92.4%, holds the first position.

An article by Thomas Germain, published on Gizmodo in June 2024, reported job losses among freelance writers and journalists due to AI text detection software mistakenly classifying their work as AI-generated.[18]

To improve the reliability of AI text detection, researchers have explored digital watermarking techniques. A 2023 paper titled "A Watermark for Large Language Models" [19] presents a method to embed imperceptible watermarks into text generated by large language models (LLMs). This watermarking approach allows content to be flagged as AI-generated with a high level of accuracy, even when text is slightly paraphrased or modified. The technique is designed to be subtle and hard to detect for casual readers, thereby preserving readability, while providing a detectable signal for those employing specialized tools. However, while promising, watermarking faces challenges in remaining robust under adversarial transformations and ensuring compatibility across different LLMs.

Anti text detection

edit

There is software available designed to bypass AI text detection.[20]

A study published in August 2023 analyzed 20 abstracts from papers published in the Eye Journal, which were then paraphrased using GPT-4.0. The AI-paraphrased abstracts were examined for plagiarism using QueText and for AI-generated content using Originality.AI. The texts were then re-processed through an adversarial software called Undetectable.ai in order to reduce the AI-detection scores. The study found that the AI detection tool, Originality.AI, identified text generated by GPT-4 with a mean accuracy of 91.3%. However, after reprocessing by Undetectable.ai, the detection accuracy of Originality.ai dropped to a mean accuracy of 27.8%.[21][22]

Some experts also believe that techniques like digital watermarking are ineffective because they can be removed or added to trigger false positives.[23] "A Watermark for Large Language Models" paper by Kirchenbauer et al. [24] also addresses potential vulnerabilities of watermarking techniques. The authors outline a range of adversarial tactics, including text insertion, deletion, and substitution attacks, that could be used to bypass watermark detection. These attacks vary in complexity, from simple paraphrasing to more sophisticated approaches involving tokenization and homoglyph alterations. The study highlights the challenge of maintaining watermark robustness against attackers who may employ automated paraphrasing tools or even specific language model replacements to alter text spans iteratively while retaining semantic similarity. Experimental results show that although such attacks can degrade watermark strength, they also come at the cost of text quality and increased computational resources.

Multilingual text detection

edit

One shortcoming of most AI content detection software is their inability to identify AI-generated text in any language. Large language models (LLMs) like ChatGPT, Claude, and Gemini can write in different languages, but traditional AI text detection tools have primarily been trained in English and a few other widely spoken languages, such as French and Spanish. Fewer AI detection solutions can detect AI-generated text in languages like Farsi, Arabic, or Hindi.[citation needed]

Image, video, and audio detection

edit

Several purported AI image detection software exist, to detect AI-generated images (for example, those originating from Midjourney or DALL-E). They are not completely reliable.[25][26]

Others claim to identify video and audio deepfakes, but this technology is also not fully reliable yet either.[27]

Despite debate around the efficacy of watermarking, Google DeepMind is actively developing a detection software called SynthID, which works by inserting a digital watermark that is invisible to the human eye into the pixels of an image.[28][29]

See also

edit

References

edit
  1. ^ "'Don't use AI detectors for anything important,' says the author of the definitive 'AI Weirdness' blog. Her own book failed the test". Fortune. Retrieved 2023-10-21.
  2. ^ Weber-Wulff, Debora; Anohina-Naumeca, Alla; Bjelobaba, Sonja; Foltýnek, Tomáš; Guerrero-Dib, Jean; Popoola, Olumide; Šigut, Petr; Waddington, Lorna (2023-12-25). "Testing of detection tools for AI-generated text". International Journal for Educational Integrity. 19 (1): 26. arXiv:2306.15666. doi:10.1007/s40979-023-00146-z. ISSN 1833-2595.
  3. ^ Hern, Alex (31 December 2022). "AI-assisted plagiarism? ChatGPT bot says it has an answer for that". The Guardian. Retrieved 11 July 2023.
  4. ^ a b Sample, Ian (10 July 2023). "Programs to detect AI discriminate against non-native English speakers, shows study". The Guardian. Retrieved 10 July 2023.
  5. ^ Fowler, Geoffrey A. (2 June 2023). "Detecting AI may be impossible. That's a big problem for teachers". The Washington Post. Retrieved 10 July 2023.
  6. ^ Tangermann, Victor (9 January 2023). "There's a Problem With That App That Detects GPT-Written Text: It's Not Very Accurate". Futurism. Retrieved 10 July 2023.
  7. ^ "We tested a new ChatGPT-detector for teachers. It flagged an innocent student". The Washington Post. 1 April 2023. Retrieved 10 July 2023.
  8. ^ Taylor, Josh (1 February 2023). "ChatGPT maker OpenAI releases 'not fully reliable' tool to detect AI generated content". The Guardian. Retrieved 11 July 2023.
  9. ^ Williams, Rhiannon (7 July 2023). "AI-text detection tools are really easy to fool". MIT Technology Review. Retrieved 10 July 2023.
  10. ^ "AI Detection Apps Keep Falsely Accusing Students of Cheating". Futurism. 6 June 2023. Retrieved 2023-10-21.
  11. ^ Jimenez, Kayla. "Professors are using ChatGPT detector tools to accuse students of cheating. But what if the software is wrong?". USA TODAY. Retrieved 2023-10-21.
  12. ^ Klee, Miles (2023-06-06). "She Was Falsely Accused of Cheating With AI -- And She Won't Be the Last". Rolling Stone. Retrieved 2023-10-21.
  13. ^ Staton, Bethan (3 April 2023). "Universities express doubt over tool to detect AI-powered plagiarism". Financial Times. Retrieved 10 July 2023.
  14. ^ Carter, Tom. "Some universities are ditching AI detection software amid fears students could be falsely accused of cheating by using ChatGPT". Business Insider. Retrieved 2023-10-21.
  15. ^ Verma, Prashnu (18 May 2023). "A professor accused his class of using ChatGPT, putting diplomas in jeopardy". The Washington Post. Retrieved 10 July 2023.
  16. ^ "College instructor put on blast for accusing students of using ChatGPT". NBC News. 18 May 2023. Retrieved 10 July 2023.
  17. ^ Dugan, Liam; Hwang, Alyssa; Trhlik, Filip; Ludan, Josh Magnus; Zhu, Andrew; Xu, Hainiu; Ippolito, Daphne; Callison-Burch, Chris (2024-06-10), RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectors, doi:10.48550/arXiv.2405.07940, retrieved 2024-10-17
  18. ^ Germain, Thomas (2024-06-12). "AI Detectors Get It Wrong. Writers Are Being Fired Anyway". Gizmodo. Retrieved 2024-06-18.
  19. ^ Kirchenbauer, Johannes; Geiping, Jonas; Wen, Yuxin; Katz, Jonathan; Miers, Ian; Goldstein, Tom (2023-06-24). "A Watermark for Large Language Model". arXiv:2301.10226 [cs.LG].
  20. ^ Beam, Christopher. "The AI Detection Arms Race Is On—and College Students Are Building the Weapons". Wired. ISSN 1059-1028. Retrieved 2023-12-02.
  21. ^ Taloni, Andrea; Scorcia, Vincenzo; Giannaccare, Giuseppe (2023-08-02). "Modern threats in academia: evaluating plagiarism and artificial intelligence detection scores of ChatGPT". Eye. 38 (2): 397–400. doi:10.1038/s41433-023-02678-7. ISSN 1476-5454. PMC 10810838. PMID 37532832.
  22. ^ Thompson, David (November 30, 2023). "Researchers Say Undetectable AI May Be a 'Modern Threat to Academia'". The Science Times.
  23. ^ Knibbs, Kate. "Researchers Tested AI Watermarks—and Broke All of Them". Wired. ISSN 1059-1028. Retrieved 2023-10-21.
  24. ^ Kirchenbauer, Johannes; Geiping, Jonas; Wen, Yuxin; Katz, Jonathan; Miers, Ian; Goldstein, Tom (2023-06-24). "A Watermark for Large Language Model". arXiv:2301.10226 [cs.LG].
  25. ^ Thompson, Stuart A.; Hsu, Tiffany (28 June 2023). "How Easy Is It to Fool A.I.-Detection Tools?". The New York Times. Retrieved 10 July 2023.
  26. ^ Rizwan, Choudhury (October 15, 2023). "Expert debunks AI tool's claim that Israel's photo is fake". Interesting Engineering. Retrieved October 22, 2023.
  27. ^ Hsu, Tiffany; Myers, Steven Lee (18 May 2023). "Another Side of the A.I. Boom: Detecting What A.I. Makes". The New York Times. Retrieved 10 July 2023.
  28. ^ Pierce, David (2023-08-29). "Google made a watermark for AI images that you can't edit out". The Verge. Retrieved 2023-10-21.
  29. ^ Wiggers, Kyle (2023-08-29). "DeepMind partners with Google Cloud to watermark AI-generated images". TechCrunch. Retrieved 2023-10-21.
Retrieved from "https://en.wikipedia.org/w/index.php?title=Artificial_intelligence_content_detection&oldid=1255851589"