OpenAI tool aims to discover AI-generated text
Open AI launched a text detection tool generated using services like his ChatGPT.
The generative AI models used for services like ChatGPT have raised many societal and ethical questions: Could they be used to generate misinformation on an unprecedented scale? What if students cheat using them? Should an AI be credited when excerpts are used for articles or articles?
A paper (PDF) from the Center on Terrorism, Extremism and Counter-Terrorism at the Middlebury Institute of International Studies even found that GPT-3 is able to generate “influential” text that could radicalize people towards far-right ideologies.
OpenAI was founded to promote and develop “friendly AI” that benefits humanity. As the creator of powerful generative AI models like GPT and DALL-E, OpenAI has a responsibility to ensure that its creations have safeguards in place to prevent – or at least minimize – their use for harmful purposes.
The new tool launched this week claims to be able to distinguish text written by a human from that of an AI (not just GPT). However, OpenAI warns that it is currently unreliable.
On a set of English texts, OpenAI claims that its tool identifies only 26% of AI-written texts as “probably AI-written”. The tool misclassifies human-written text about nine percent of the time.
“While it is impossible to reliably detect all AI-written text, we believe that good classifiers can shed light on mitigations of false claims that AI-generated text was written by a human: for example, running automated disinformation campaigns, using AI tools for academic dishonesty, and positioning an AI chatbot as human,” says OpenAI.
“(The tool) should not be used as a primary decision-making tool, but rather as a supplement to other methods for determining the source of a piece of text.”
The tool’s other main limitations are that it is “very unreliable” on short texts under 1,000 characters and is “significantly worse” when applied to non-English texts.
OpenAI originally provided access to GPT to a small number of trusted researchers and developers. As it developed stronger safeguards, a waiting list was later introduced. The waitlist was removed in November 2021, but work to improve security is an ongoing process.
“To ensure API-based apps are built responsibly, we provide tools and help developers use best practices so they can get their apps into production quickly and securely,” OpenAI wrote. in a blog post.
“As our systems evolve and we work to improve the capabilities of our protections, we plan to continue to streamline the process for developers, refine our usage guidelines, and enable even more breaches. use over time.”
OpenAI recently shared his point of view that there is too much hype around the next major release of GPT, GPT-4, and people are “begging to be disappointed”. He also believes a video generation model will come, but won’t put a deadline on when.
Videos that have been manipulated with AI technology, also known as deepfakes, are already proving problematic. People are easily convinced by what they thought they can see.
We’ve seen deepfakes from figures like disgraced FTX founder Sam Bankman-Fried commit a fraudUkrainian President Volodymyr Zelensky misinformand United States House Speaker Nancy Pelosi defame and make her look drunk.
OpenAI does the right thing by taking its time to minimize risk, keeping expectations in check, and creating tools to uncover content that falsely claims to have been created by a human.
You can test OpenAI’s work-in-progress tool to detect AI-generated text here. An alternative called ‘DetectGPT’ is being created by researchers at Stanford University and is available for demonstration here.
(Picture by Shane Aldendorff to Unsplash)
Related: OpenAI opens waiting list for paid version of ChatGPT
Want to learn more about AI and Big Data from industry leaders? Check AI and Big Data Fair taking place in Amsterdam, California and London.
Check out other upcoming enterprise technology events and webinars powered by TechForge here.
OpenAI tool aims to discover AI-generated text
