philip lelyveld The world of entertainment technology

(Wired) The AI Detection Arms Race Is On

... Ben Colman, CEO of Reality Defender, says his company’s detection tools are unevadable in part because they’re private. (So far, the company’s clients have mainly been governments and large corporations.) With publicly available tools like GPTZero, anyone can run a piece of text through the detector and then tweak it until it passes muster. Reality Defender, by contrast, vets every person and institution that uses the tool, Colman says. They also watch out for suspicious usage, so if a particular account were to run tests on the same image over and over with the goal of bypassing detection, their system would flag it. ...

The purpose of Writer isn’t to write for you, she says, but rather to make your writing faster, stronger, and more consistent. That could mean suggesting edits to prose and structure, or highlighting what else has been written on the subject and offering counterarguments. The goal, she says, is to help users focus less on sentence-level mechanics and more on the ideas they’re trying to communicate. Ideally, this process yields a piece of text that’s just as “human” as if the person had written it entirely themselves. “If the detector can flag it as AI writing, then you’ve used the tools wrong,” she says. ...

Mollick, who teaches entrepreneurship at Wharton, not only allows his students to use AI tools—he requires it. “Now my syllabus says you have to do at least one impossible thing,” he says. If a student can’t code, maybe they write a working program. If they’ve never done design work, they might put together a visual prototype. “Every paper you turn in has to be critiqued by at least four famous entrepreneurs you simulate,” he says. ...

“I think that came out of Covid,” Amiton continued. “People reevaluated what the purpose of work even is, and if you can use ChatGPT to make your life easier, and therefore have a better quality of life or work-life balance, then why not use the shortcut?” ...

I noticed that his framing of GPTZero/Origin was shifting slightly. Now, he said, AI-detection would be only one part of the humanity-proving toolkit. Just as important would be an emphasis on provenance, or “content credentials.” The idea is to attach a cryptographic tag to a piece of content that verifies it was created by a human, as determined by its process of creation—a sort of captcha for digital files. Adobe Photoshop already attaches a tag to photos that harness its new AI generation tool, Firefly. Anyone looking at an image can right-click it and see who made it, where, and how. Tian says he wants to do the same thing for text and that he has been talking to the Content Authenticity Initiative—a consortium dedicated to creating a provenance standard across media—as well as Microsoft about working together. ...

It’s a subtle but meaningful distinction: Human writing may not be better, or more creative, or even more original. But it will be human, which will matter to other humans. ...

See the full article here: https://www.wired.com/story/ai-detection-chat-gpt-college-students/