Google has come up with AI accelerated spam detection feature for Gmail users to make their Gmail account safer. The AI update is termed as RETVec – (Resilient and Efficient Text Vectorizer). RETVec is an important upgrade in the text classification technology system created to counter adversarial text manipulations, make text classifiers more robust and efficient, and assist Gmail in fighting against spam.
What is RETVec and How it will Work?
According to Google, platforms such as YouTube, Gmail, and Google Play use text classification models to recognize maleficent content, such as negative comments or phishing attacks. However, for machine learning such texts become difficult to classify because bad actors fiddle with the text to escape detection, using keyword stuffing and invisible characters.
To deal with such a challenge, google has introduced RETVec for Gmail. This feature is capable of managing the challenge of adversarial text manipulations, which are emails unrecognized with special characters, typos, or emojis. These tactics often evade traditional spam filters, which allow malevolent mail and potentially create harm to users.
According to Google, RETVec has undergone substantial testing over the past years and has come up with effectual security. RETVec has shown a 38% enhancement in spam detection rate and depletion of 19.4% in false positive rate contrasted to text vectorizer in Gmail’s spam classifier.
Further, deploying RETVec led to an 83% reduction in TPU (Tensor Processing Unit) usage for the model, making it a substantial upgrade in defense capabilities. Additionally, the implementation of RETVec resulted in a significant decrease of 83% in the utilization of TPU (Tensor Processing Unit) for the model. This enhancement greatly enhances our defensive capabilities.
Google added – RETVec is an open-source text vectorizer that allows you to develop more efficient on-device and server-side text classifiers. The Gmail spam filter utilizes this feature to safeguard Gmail inboxes from potentially harmful emails.
Bottom Line
The remarkable adaptability of RETVec allows it to be used effectively on on-device, mobile applications, web, and large-scale text classification deployments. The compact format of RETVec-trained models enables quicker inference speeds, while the smaller models produced by RETVec lead to decreased computational expenses and latency.