The arrival of artificial intelligence software like ChatGPT is a very recent phenomenon, says Nabil Tayeb.
“The first versions of ChatGPT date back to 2019. They were already able to write texts on very specific topics, but you had to place specific commands, with the right keywords.”
The newer version of the software, ChatGPT 3, is much more user-friendly. So much so, says the entrepreneur, that it even surprised companies operating in the field of artificial intelligence.
The number of users of this technology is growing rapidly, and exponentially.
How to detect texts produced by AI?
Draft & Goal is developing a platform capable of producing content using an artificial intelligence module.
“However, while developing our product, says Nabil Tayeb, we realized that the content generated by AI left fingerprints.”
The Montreal startup then developed an algorithm capable of detecting the traces left by an artificial intelligence in its creations.
This tool can detect some ChatGPT flaws, for example:
- misinterpretations;
- exaggerations; and
- the signs that the algorithm does not know the answer.
A trial version of the service for detecting that a text has been written by ChatGPT is available for free.
- For the moment, it only works with texts in English.
- A French version is expected by the end of the month.