Edward Tian has had a busy few months.
In December, the Princeton student used his vacation to create a tool for educators to help them determine whether or not student essays were written with OpenAI’s ChatGPT called GPTZero. Buoyed by growing concerns about emerging technology and the nascent AI boom, Tian’s tool went viral, amassing more than 6 million users in just a few months.
Since then, he’s received calls and meetings from a slew of investors, created a team and startup named after GPTZero to help refine the bot detector, and secured millions of dollars in funding. for their new product: Origin, a web extension that he says can detect AI-generated text on web pages. It has already caught the eye of media moguls, tech founders and deep-pocketed venture capitalists.
Oh, and on top of all that, he’ll be graduating from college soon.
“The last few months have been crazy,” Tian told The Daily Beast. “It’s been really crazy.”
“Finding the source of the information is critically important in a world of fake news. If you don’t know the source, how do you trust the information you consume?”
— Edward Tian, GPTZero
The same can be said for the tech world more broadly. Since the release of ChatGPT in November 2022, it seems like the whole world immediately forgot the old fashions of crypto and the metaverse, and started betting big on generative AI like chatbots and image generators. While companies as big as Google and Microsoft are betting on AI, there’s a parallel industry of tools to combat and detect these bots quietly growing alongside it – and GPTZero is a big part of that effort.
The startup’s new tool, Origin, represents another weapon in this growing AI arms race. The app currently works as a Chrome extension that allows users to analyze any text they come across online to see if it was AI-generated or not. Origin isn’t just useful for teachers trying to determine if their students have written an essay on the Battle of Hastings; According to Tian, it could also help people like journalists and tech watchdogs identify AI-generated misinformation online.
“Finding the source of the information is critically important in a world of fake news,” Tian said. “If you don’t know the source, how can you trust the information you consume? »
Tian describes it as a way to bolster media and digital literacy at a time when people are increasingly wary of the things they see online, and with good reason. Media companies like News Feed And CNET have already started quietly producing AI-generated content. Last month, Initiated editor Nicholas Carlson announced his intention that his editorial team would begin experimenting with bot-written content.
Having tools that can discern whether or not something was generated by a bot becomes just as important as the bot that generated it, if not more so. That’s why the tastes of the ancients The New York Times CEO Mark Thompson and former Reuters CEO Thomas Glocer started investing in GPTZero. As the world begins to tackle all the ramifications of generative AI, it becomes more important than ever that we can separate things that have been created by a human from content produced by a string of code. .
For Tian, the impact of tools like Origin can be summed up by why he developed GPTZero in the first place: “Humans deserve to know when writing isn’t human.”
“The value of human writing is there, but it’s undermined if people can’t tell the difference in the information they consume,” Tian said. “And I think our eyes are no longer enough to see the difference.”
The AI boom has created some strange bedfellows. At its I/O Developer Conference on Wednesday, Google announced a whole host of AI-injected products to help users write emails, generate images for slideshows, and even come up with fun captions for images.
However, the company also announced that it will launch a watermark feature on all its AI-generated images so that users can know whether or not an image was created using their image generator tool. . The watermark itself is actually in the metadata of those images. So there’s no easy way to immediately know whether or not an image was created by a bot without downloading it and digging into the metadata.
Google also said it will flag AI-generated images on Google Image results, which will allow users to determine whether or not an image was created by a bot. This summer, the company is also launching a new tool for the US called ‘About This Image’, which allows users to essentially reverse image search that shows you where the image first appeared. on the Internet according to Google’s indexing.
The irony here, of course, is that Google is also going all-in on AI. Not only does it appear to be infusing its proprietary PaLM 2 into nearly its entire product line, but the company is also launching its Bard chatbot in over 180 countries. This will put powerful and potentially dangerous robots in the hands of millions, if not billions of people.
Google isn’t the only company that seems to play both sides, so they always end up on top either. Jack Altman, the brother of OpenAI CEO Sam Altman, invested in GPTZero using family investment firm Altman Capital. Tian told The Daily Beast that while Sam was interested, he couldn’t get directly involved due to “conflicts of interest,” likely due to him being the CEO of one of the companies. world’s most important AIs.
“There’s an internal value to having a human on the other side, which I don’t think will ever change.”
— Edward Tian, GPTZero
It is clear that as AI has increased, the tools and services aimed at combating and detecting it, such as Origin, will also increase. But more importantly, they could provide us with an opportunity to confront a kind of digital solipsism that seems to permeate the world as AI becomes ever more powerful and widespread. We need to know if the people we interact with, the content we read, and the media we engage with are real or not, otherwise the internet would be a truly isolated place.
“I feel like human society will lose its momentum for progress and come up with new things if everyone just uses the AI output that takes what’s already on the internet and regurgitates it,” Tian said. . “There’s an internal value to having a human on the other side, which I don’t think will ever change.”