The startup claims it can reliably detect AI-generated content | Jobs Recent

The worlds of media, academia and business swoon over the sophistication of OpenAI LLC’s ChatGPT natural language generator, but content marketers have reason to be concerned.

This industry thrives on search engine optimization and is hypersensitive to intellectual property issues such as plagiarism, especially since Google LLC has made it clear that it looks disapproval of publishers of stolen and auto-generated content. Google isn’t saying whether it has cracked the code for reliable machine-generated text detection, but the Canadian content retailer believes it’s pretty close, even if it’s not entirely sure how the solution it sells actually works. launched earlier this month, claiming it can detect content generated by popular natural language processing engines such as Generative Pre-trained Transformer-3, GPT-J, and GPT-Neo with accuracy well above 90%.

Saving term work founder Gilham is “very confident” in the company’s ability to detect AI-generated content. Photo:

This can be a big deal for those in content marketing and academic fields. The prospect that AI could soon create lengthy content that could rival the quality of human writers even prompted The Atlantic last week to question whether university essays are dead.

“I think there’s a monstrous wave of AI-generated content coming to universities, and they’re going to have a hard time dealing with that because it’s not like plagiarism,” said Jonathan Gilham, founder of

Content vendors use an assortment of plagiarism checkers to protect themselves from posting stolen intellectual property. They don’t fear being accused as much as rap from Google, which can cause their SEO scores to plummet.

Gilham knows this firsthand. He spent seven years starting and running content marketing and advertising companies where he became frustrated with the quality of plagiarism checking in the market.

“They weren’t built for teams that were publishing thousands of content, so we decided to build a plagiarism checker that had both team and enterprise level features,” he said. “And we thought it could detect AI as well.”

Gilham and co-founder Conor Watt commissioned the project to developers who created a machine learning algorithm trained on the most popular natural language processing models. Unlike Hugging Face Inc.’s AI detector, which uses probabilistic techniques to guess the words an AI content generator is most likely to use, Gilham said, “our AI is much heavier on the computational side and looks at the article holistically, not using a linear function.” He noted that a recent test on a small set of articles generated by ChatGPT gave accuracy rates of over 98%.

Black box?

How it’s working? Gilham admits he doesn’t know exactly. However, he knows that in a test of 10,000 samples of GPT-generated content, the algorithm detected machine-written content more than 94% of the time. makes its service available for a penny per 100 words scanned and has acquired over 1,000 paying customers in just two weeks, including users from 15 universities. The founders do not actively raise money. “It looks like our development team will make us cash flow neutral,” said Gilham.

Google hasn’t said whether it has solved the AI ​​content detection problem, but given its deep NLP experience, it’s probably on the right track. Nevertheless, “they’re not going to go out and share their solution, so we think publishers will need their own tool,” Gilham said. “It’s our niche.”

Doesn’t it bother him a bit that he can’t explain how works? Yes, he admitted, but half of trading decisions on exchanges are made by machines without direct human supervision. “I know how the tests work and I’m confident in their accuracy,” he said.

Image: Pixabay

Show your support for our mission by joining our Cube Club and Cube Event expert community. Join the community that includes Amazon Web Services and CEO Andy Jassy, ​​Dell Technologies Founder and CEO Michael Dell, Intel CEO Pat Gelsinger, and many other luminaries and experts.

Source link