A Montreal tool that detects AI can combat “ChatGPT drifts”.

Operation is simple: Open the AI ​​Detection Tool web page in a browser, (new window),We copy and paste a text of at least 200 characters in English (a French version is in preparation) into the box to test, we press to analyzeAnd the software provides a percentage of whether the article was written by a human or artificial intelligence (AI).

If you are testing content that you already know, you can also indicate at the bottom of the screen whether the result obtained by the company is correct or not.

” We did the first analysis on Wednesday based on 2,000 lessons in English. The results show a reliability of 93%. »

For the second version we are working on with an improved algorithm, this percentage is expected to be higherSays CEO of Radio-Canada Company.

Draft & Goal uses machine learning to accurately determine whether a text has been generated by a chatbot.

Generative AI models are often probabilistic, that is, models that ensure that what they write is not completely random.Nabil Tayeb explained.

This is how Draft & Goal approaches Identifying fingerprints left by AI, especially by looking at various aspects of writing style.

While the tool is currently only offered in English, a French version of the service will be online by the end of January, according to Nabil Tayeb.

There are many more English texts on the web, so the starter model is easier and faster to train with English than with FrenchDefends Draft & Goal’s CEO plans to expand the service to other languages ​​in the future.

one A combination of circumstances

Draft & Goal has been working for two years to create an automated content generation tool. She wants to help other companies be more productive, whether it’s writing blog articles, publishing static web pages, or writing product descriptions, while still maintaining some human intervention.

By generating texts, one must know whether it is perceived as a text written by an artificial intelligenceNabil Tayeb, CEO of Draft & Goal, said the aim was not to mislead readers.

” So we created a tool to help us, and we quickly realized that we’re pretty good at spotting AI-generated content. »

The team saw this as an opportunity to share with others, as there are many dangers associated with the democratization of artificial intelligence software that mimics human input in an ultra-realistic manner, similar to ChatGPT d’Open AI.

We see AI as a tool to improve a typewriter, but not to become the center of creation.The CEO of the young shoot believes.

Many areas of action

Teachers are afraid For example these conversational bots infiltrate the essays of their students at all levels.

Professors have been using plagiarism detection software for years, but these are designed to analyze texts that have already been published on the web.

Since AI generates original text, it does not copy text from the webHe says.

A tool like Draft & Goal can also be useful in press rooms to ensure the authenticity of a press release.

” [Une personne malintentionnée] Completely bogus press releases can be sent out without being suspicious, and it can be done on a massive scale. »

During the test phase

Currently, the Draft & Goal tool is still under development. Each request improves its own artificial intelligence.

The company wants to make the software free and available to everyone As long as possible. It still plans to offer paid plans with additional features, among other things, in the form of an extension on the Chrome browser.

A new version of the detector will be put online In the next few days And it’s possible to target more precise parts of text that are suspected to have been generated by artificial intelligence.