Preventing bias while using AI

We have built our models and AI in a manner that ensures there's no bias. Our AI does not use factors like a speaker's gender, name, race or any such aspect that might introduce bias in detecting key moments.

What we have done to minimize and eliminate bias -

• The core Moments AI model has been built without any parameters that utilize a person's identifying characteristics.
• When performing screen analysis, our technology does not identify a person's physical characteristics such as race or gender. It is explicitly programmed to not do that.
• Audio is transcribed and language analysis happens on the transcribed words instead of the actual audio.
• The algorithms do not pay attention to language elements like slang that might inadvertently lead to bias.

The primary focus of our our technology is on analyzing the information content of videos and words.

Note that AI transcription quality depends on a speaker's audio quality and accent. If you are worried about higher error rate for non-native speakers of English, we recommend using our Augmented AI transcription to minimize chances of errors.

Parmonic is a derivative AI and does not create content based on generic, Internet-sourced data like public AI models. This ensures that biases from the internet or someone else's dataset are not introduced into our customer's use of Parmonic.