Most AI Detectors are unreliable because they don’t look deep enough into what makes content more Human.
The LanaiLabs AI Detector 'Mattie' calculates a score that is statistically and practically vetted on tens of thousands of training samples both human and AI. In addition to the multi-factor equation we have created that takes into account the ways that humans uniquely express themselves, we also take into account the uncertainty (error) any given single test may have.
What We Analyze
Large Language Models (LLM) like ChatGPT, are great at predicting the most likely next word in any given content it writes, but largely it still lacks human emotion, displaying variety in language. To better identify the lack of humanness our multi-factor equation takes the following into account.
Variety
Prominent researchers in fundamental physics believe that Variety measures “how unique, one from another, the different parts of a system are." Therefore our equation looks at how unique words are in relationship to all text being analyzed. The rarer the word, the greater the weight we apply.
Sentence
General Purpose Transformer models (GPT) apply to language models (LLMs) to predict the next word in a string of words (i.e. paragraph). To predict the next word, the model uses a buffer of previous words. That buffer may be the last 8 words or the last 16 etc., but not all the previous words. Because of that, GPT LLMs tend to be repetitive, and they repeat themselves.
Vocabulary
We've used the Google Billion Word corpus to look at the probability of any other reasonable word choice in any given sentence. The word "the," for example, with the frequency of 23 billion in this corpus would impact the score if used too often.
Better AI means building with your false positives and improving confidence in your results. We also supply detailed reporting for a complete understanding of your results.
Frequently Asked Questions (FAQs)
What is lanai ai text detector?
Lanai's AI text detector is a predictive equation designed using physics principles to maximize the difference in how humans write text and current AI generation text is created.
Does Lanai only work with ChatGPT?
No, Lanai works across any of the AI language models, including ChatGPT, GPT-4, GPT-3, GPT-2, LLaMA, Google's Bard AI, Bing Chat, Claude 2, and AI services based on those models. Developments in LLMs are moving fast, so we are continually making updates to our algorithm.
What languages does the Lanai AI text detection tool support?
Currently, we only support English, but we are working on supporting other languages.
Why doesn’t the extension work on some web content?
Due to how some blogs and older sites structure their HTML, not all of the text may be able to be analyzed. We continue to make improvements to our application to identify these differences in how text is coded and presented on sites. It is our ultimate goal to be able to analyze all online text.
I pasted my own writing, and it says it was AI generated, why?
Lanai uses a different approach than most AI detectors using multi-factor linear regression models. The shorter the text is, the more general it is, and the more essay-like it is, the more likely it is to result in a false positive. We are working on improving the system so that this occurs less frequently.
Sometimes the text in a web article or the text I paste is fully analyze, why?
We are launching with a free version, which is currently truncated to 2000 words. The browser extension version is provided free to users to test on a limited basis. We will be offering Pro and Enterprise subscriptions that will allow much larger bodies of text and batch uploads.
How can I provide feedback or report issues with the AI Text Detection Tool?
We value user feedback and encourage you to report any issues or suggestions you may have. Please email use at contact@lanai.ai