OpenAI's AI Classifier Fails 75% of Time

It's a start. And OpenAI's mindset about technology has been patience. In fact, its founders didn't know if its generative AI being developed would work. Although ChatGPT and DALL - E are causing a sensation of interest they are not really ready for primetime.

Now released by OpenAI is the AI classifier. Its function is to determine human from content generated by a number of AI programs, not only that from OpenAI.

However, OpenAI admits it fails about three-fourths of the time, reports The Wall Street Journal. Also it has to be used in conjunction with other tools. 

Educators already possess some of those other tools. Their need is urgent to figure out if student work has been done by AI or by themselves. If students are skillful in setting up the prompt that will produce the report or essay they need they could be submitting "A" work. Even seasoned educators might not be able to "smell" out that it is AI production. When I was a college instructor I usually could detect traditional forms of plagiarism. Since I have already used ChatGPT I am not confident I could pick up on its role in doing the assignment.

The classifier is free and here you can access it. Essentially the determination is made on a five-point scale. Accuracy tends to be better with long form - more than 1,000 words - and in English. 

In time, given OpenAI's track record, it could be a winner.

Human Resources Communications and Intuitive Career Coaching. Award-winning book author and article ghostwriter. Confidential complimentary consultation. Then fees customized for your unique budget.

Please make an appointment with Jane Genova at janegenova374@gmail.com

 


Comments

Popular posts from this blog

Kirkland & Ellis Reported to Be Building Moat Around Firm to Deter Poaching of Stars

Akin Gump Julia Ghahramani's March 2021 Cocaine+ Death - So?

Up-or-Out: McKinsey Raises the Pressure, In Contrast Some Law Firms Ease It through Nonequity Partner Tier