Tools for detecting jobs written by artificial intelligence

The development of A.I. raises the Turing question again and again: Is this information of human origin? Am I dealing with a human or a robot?

To better understand the logic behind detection tools, I invite you to read this interesting article by Olivier Duffez with two AI and SEO specialists: Vincent Terrasi and Sylvain Peyronnet - Le point sur la détection de contenus générés par AI - presents different methods for identifying the origin of texts and their limits. A real race is on.

Findings

• Artificial intelligence doesn't make spelling or grammar mistakes.
  
  
• Artificial intelligence proceeds by probabilistic judgment; it takes the most likely and most frequently encountered explanations. It rarely uses rare words, because rare words are rarely used. Its language is generally made up of common words.
  
  
• The algorithms used each have their own signature, their own imprint, which can eventually be recognized.
  
  
• Artificial intelligence ensures the coherence of its texts by taking context into account. There are no mixtures of terms without context.
  
  
• Algorithms are characterized by their results, such as the frequency of certain words, the density of keywords and the language rules they follow "to the letter".
  
  
• An A.I. can write implausibilities without any restraint.

Based on these kinds of observations, here are services that calculate the degree of probability that a text is written or not by an artificial intelligence, a new kind of Turing Test.

1. AI Detector Pro - by Westbourne International... so you don't have to pay a lot for machine-generated content.
   
   
2. Compilatio AI Detector by Compilatio - multilingual
   
   
3. Copyleaks AI Content Detector - by Copyleaks, 99.12% efficiency... claims to be verified.
   
   
4. Crossplag AI Content Detector - Artificial intelligence is learning to detect artificial intelligence... as long as we don't get a DNA test to check we're not a robot.
   
   
5. Detecting-AI.com detects texts generated by artificial intelligence, used in over 100 countries. Thanks to models trained on several languages, it offers high accuracy in identifying content produced by models such as GPT, while also offering complementary functions such as plagiarism detection.
   
   
6. Draft&Goal - Specifically calibrated for Chat GPT, but its approach will enable it to detect other A.I. signatures - In English and French.
   
   
7. GPT-2 Output Detector Demo based on RoBERTa (A Robustly Optimized BERT Pretraining Approach)
   
   
8. Giant Language model Test Room by MIT-IBM Watson AI lab and Harvard NLP (in English)
   
   
9. GPT Radar by Neuraltext
   
   
10. GPT Zero - by Edward Tian of Princeton University
    
    
11. Moxby - supported by MailChimp
    
    
12. Open AI Text Classifier - by Open AI, the creator of Chat GPT. With reliable data on the detector's real effectiveness.
    
    
13. Originality - Originally a plagiarism checker, now augmented by an AI checker.
    
    
14. Pangram - Claims an almost 0.01% false positive rate. In other words, when it says it's not an AI, it is.
    
    
15. Sapling - extends its AI offering with this service. What's human and what isn't is a curious ethical question indeed.
    
    
16. Winston AI - Covers all known AIs with a 99.98% efficiency rate.
    
    
17. Writer AI Content Detector - by Writer - Free version limited to 1500 characters.
    
    
18. ZeroGPT - Scalable detection model. Free up to 15,000 characters.
    
    
    

Several of us have submitted this text to them, and in every case it has been certified 99% A.I.-free.

It's safe to assume that similar tools will soon follow, eventually creating A.I.s each with their own identifiable writing style.

Will they even imitate the writing style of schoolchildren or children? There's no reason to believe they couldn't, just as they've already done for art, with Deep Art Effect.

Illustration : Deposit Photos - paulfleet

Related articles

Artificial intelligence in a frenzy - try out dozens of public apps
https://cursus.edu/fr/25952/intelligence-artificielle-en-frenesie-essayez-plus-de-30-applications-publiques

Knowledge as a public good to be seized
https://cursus.edu/fr/26201/la-connaissance-comme-bien-public-a-semparer

How to easily detect whether a text has been generated by artificial intelligence
https://cursus.edu/fr/32163/comment-facilement-detecter-si-un-texte-a-ete-genere-par-une-intelligence-artificielle