Update: Our Multilingual AI Detector Has Gotten An Upgrade

We’re excited to announce that we are launching a new and improved version of our Multilingual AI detector, which now also includes the new OpenAI model GPT-4-turbo.

Our Multilingual AI Detector version 1.3 demonstrates superior performance over version 1.2, with outstanding prediction quality across all standard metrics and for all AI models.

Especially the accuracy of the new model GPT-4-Turbo is approximately 99%, and improved accuracy across all languages tested.

How We Conducted Our Version Upgrade Testing

Of course, we don’t expect you to just take our word that version 1.3 is better than 1.2! We needed to conduct a detailed and thorough test to be sure that 1.3 was actually better than 1.2 and laid out some simple parameters to follow.

We categorized each piece of text we ran through the 1.2 and 1.3 versions of the multilingual AI detector into four different categories. 

Categories

1. True Positive - AI detector correctly identified AI content as AI.
2. False Negative - AI detector incorrectly identified AI content as Human.
3. False Positive - AI detector incorrectly identified Human content as AI.‍
4. True Negative - AI detector correctly identified Human content as Human.

The Results

From our tests, here are the standout results: 

• The achieved results across all languages consistently exceed 93%.
• The accuracy rate for 14 out of 15 languages exceeds 95%.
• The accuracy for English significantly improved, rising from 90% to 98%.
• Overall, model version 1.3 demonstrated better performance across languages.

Results Per Source Data

When we look at those results based on the source data, version 1.3 also continues to impress:

• The version 1.3 model dominates the results over the 1.2 model across all sources.
• The standout feature of this release is the accuracy of GPT4turbo, which reaches 99%.

In Summary

In summary, the Multilingual AI Detector version 1.3 represents a significant improvement over its predecessor, version 1.2.

The findings highlighted above demonstrate significant promise in addressing the challenges of multilingual AI detection, with strong indications of further enhancements and breakthroughs shortly.

Future Works

But our work doesn’t stop here. 

Just as leading AI companies continue to try and develop their AI models, we will continue to research more academic papers to apply new ideas to our AI detector product.

Our role within this growing industry has not changed, and our focus will always be on leading the conversation in AI content detection, its importance, and the value of openness and transparency.

With the rapid evolution of LLMs, we need to collect more data from new LLMs like Mistral, OpenHermes, Gemma, Gemini, and so on, to continue expanding our product’s capabilities.

So watch this space!