Rival student-developed tools try to hide and uncover something crucial.
People of a certain advanced age may remember this catchy slogan about authenticity: Is it alive or is it Memorex?
In the quaint days of cassette recordings, signal-to-noise ratios, and tape hiss, the phrase encapsulated the company’s goal of recording sound indistinguishable from the real thing.
Oh, that life was still so simple.
The onslaught of artificial intelligence tools is accelerating, bringing with it a radical reworking of what is authentic, trustworthy, and even human.
Microsoft co-founder Bill Gates has said that technology is “as fundamental as the creation of the microprocessor, the personal computer, the Internet and the mobile phone.”
And as fast as AI develops, people are trying to come up with tools to detect its use.
We need to know
Part of the impetus comes from one of the earliest use cases for AI: academic cheating.
That? Forgot to research and write that term paper? No problem, just give Chat GPT a message about what it’s about and how long it should last and you’ll be up and running in no time.
That was great, for about two seconds, until a savvy Princeton college student devised an app to detect the telltale signs of an AI-written document. The tool, called GPT Zero, examines text for signs of automatic writing.
Edward Tian released a preliminary version in January arguing in a series of tweets that “we as humans deserve to know!”
The folks at OpenAI, who created Chat GPT, followed just a few weeks later with their own “sorter to distinguish between text written by a human and text written by AI from a variety of providers.”
Among the reasons why they see it as necessary are people who “run automated disinformation campaignsusing AI tools for academic dishonesty and positioning an AI chatbot as a human being.”
Challenge accepted
But since every action has an equal and opposite reaction, another clever student, this time from Stanford, has devised a tool to hide the use of artificial intelligence in writing.
Soham Govande just released it this month.
“HideGPT Launched! Completely masks the GPT fingerprint in any text (up to 5000 words!). Write AI essays without getting caught,” Govande wrote in a tweet.
Of course, it’s not just heavy-drinking college students who might use the tool. Govande noted in a tweet that “Google takes down blog posts written by AI,” suggesting a potentially lucrative, real-world use case.
In the meantime, students may want to consider asking Chat GPT to add a few extra misspellings and apostrophes to give their “work: a more human touch.”
As for which student tool will win out in the end, time will tell. But in a potentially crucial test, the two apparently met in person without creating a mater-antimater cancellation.
