AI detector shut down by OpenAI due to low accuracy

OpenAI Shuts Down AI-Detection Software Citing Low Accuracy

Artificial intelligence pioneer OpenAI has quietly discontinued its AI-detection software due to its low level of accuracy. The AI classifier, developed by OpenAI, was initially launched on January 31 with the objective of assisting educators, such as professors and teachers, in distinguishing between human-written and AI-generated text.

However, according to OpenAI’s original blog post introducing the tool, the AI classifier has been deactivated as of July 20, 2023, due to its insufficient accuracy:

“As of July 20, 2023, the AI classifier is no longer available due to its low rate of accuracy.”

The tool’s link is no longer functional, and OpenAI’s explanation for its discontinuation was brief. Nonetheless, the company clarified that it is actively exploring more effective methods of identifying AI-generated content and integrating user feedback.

OpenAI’s former AI classifier in action. Source: OriginalityAI

From the outset, OpenAI stressed that the detection tool had limitations and could not be considered entirely reliable. The company acknowledged that the AI detection tool was “very inaccurate” when verifying text with fewer than 1,000 characters, and it often misidentified human-written text as AI-generated.

OpenAI’s Product Scrutiny: ChatGPT

This AI classifier is the latest addition to a series of OpenAI products facing scrutiny. On July 18, researchers from Stanford and UC Berkeley published a study revealing a decline in the performance of OpenAI’s flagship product, ChatGPT, over time.

We evaluated #ChatGPT‘s behavior over time and found substantial diffs in its responses to the *same questions* between the June version of GPT4 and GPT3.5 and the March versions. The newer versions got worse on some tasks. w/ Lingjiao Chen @matei_zaharia https://t.co/TGeN4T18Fd https://t.co/36mjnejERy pic.twitter.com/FEiqrUVbg6

— James Zou (@james_y_zou) July 19, 2023

Researchers discovered that ChatGPT-4’s ability to accurately identify prime numbers plummeted from 97.6% to a mere 2.4% over the past few months. Furthermore, both ChatGPT-3.5 and ChatGPT-4 experienced a significant decline in their capacity to generate new lines of code.

Opportunities for Enhanced AI-Generated Content Identification

OpenAI acknowledges the need for improved techniques to identify AI-generated content and aims to explore more effective provenance methods for text. The company is dedicated to developing and implementing mechanisms that enable users to discern between AI-generated and human-generated audio or visual content.

While OpenAI’s AI-detection software has been discontinued, the company’s commitment to refining and advancing the field of artificial intelligence remains steadfast.

Editor’s Notes

OpenAI’s decision to discontinue the AI-detection software emphasizes the challenges involved in accurately distinguishing between human-written and AI-generated content. As AI technology continues to progress, the development of more effective detection methods is crucial in maintaining transparency and authenticity in various domains, including education, journalism, and entertainment.

For the latest news and updates in the world of artificial intelligence, visit the GPT News Room.

Source link

from GPT News Room https://ift.tt/JPzT6qF