- The educated guard seeks to maintain courteous chatbots and less prone to exploitation
- With NLP in its nucleus, it works by classifying the text on a four -point courtesy scale
- The data set and the source code are available in Github and hugging
Intel has presented courtly courtes, an open source AI tool aimed at assessing the courtesy of a text and allowing chatbots to remain consistently educated for customers.
In a publication in the Blog of the Intel community, the latest incorporation into the Intel’s portfolio hopes to provide a standardized framework to evaluate linguistic nuances in the communication promoted by AI.
Taking advantage of the processing of natural language (NLP), Intel affirms that the educated guard, by classifying the text into four different categories of courtes, somewhat educated, neutral and impolite, helps mitigate the vulnerabilities of AI to “provide a mechanism of Defense against adverse attacks. “
Intel’s polonada guard role for SMEs
According to Intel, the educated guard reinforces the resilience of the system by guaranteeing consistent educated production even when the potentially harmful text is handled.
The company expects this approach to do it “[improve] Customer satisfaction and loyalty “for companies that implement it.
Launched under the MIT license, Polite Guard gives developers the flexibility of modifying and integrating it into their own projects.
Its data set and source code are available in Github and Hugging Face, with more developments that will be published through the Intel community blog.
