Technology is everywhere and can be misused. The GPT-3, which is an AI tool that is used for language generation landed up in the same grounds of misuse which was unethical and also unbelievable. Let’s take a look at the whole incident which is in trend on the top of the Hacker’s News.
GPT-3: What is it?
Designed by San Francisco based OpenAI GPt-3 is the latest version of a series of AI auto-complete tools. It has been in the development mode for several years. Simply, it is a tool that auto-completes your text based on prompts from the human writer.
GPT-3: What did exactly happen?
A college student named Liam Porr used the AI tool for writing fake blogs. Yes, if you can’t believe your eyes read that again. She knew the basics of the tool and by the end of college; Porr used the AI model to produce an entirely fake blog under a fake name. Guess what? Technology simplifies the human task and work-load, the same happened here.
It was done as a fun experiment but turned out to be a major one. One of her blogs reached the top of Hacker’s news. And, as we humans are keen readers of things at the top some of them noticed that the blog was completely AI-generated instead of being human written. And, the most interesting part of it is that some of them subscribed it too.
GPT-3: How did it happen?
Well, if we go back in time then we can recall the incident from February last year when OpenAI made headlines with GPT-2. It is an earlier version of the algorithm, and it was announced that it would be withheld as it can be abused. Though, the decision received immediate and major backlashes by the researchers. And, by November the lab had reversed positioning and released the model stating that it had detected no strong evidence of misuse so far.
Though, a different approach was applied in the case of GPT-3. It was neither withheld nor was granted for public access. Instead, the algorithm was provided to particular researchers who applied for the private beta, with the goal of gathering their feedback and commercializing the technology by the end of the year.
In this case, Porr too submitted an application. He filled out a questionnaire form that consisted of simple questions. But, his patience was limited. After reaching out to several researchers of Berkeley AI community, he found out a Ph.D. student who had access to it. The student agreed to collaborate and Porr wrote a small script for him to run.
GPT-3: How did he do it?
Porr picked up a popular blog category that doesn’t require much rigours logic. He wrote headlines following a simple formula. Scrolling around on Medium and Hacker News to see what was performing in those categories and put together something that was relatively similar. He explains that he wanted to prove that GPT-3 could be passed off as a human writer.