According to AIbase, tech giant Meta is testing a new feature for its AI chatbots that allows them to send messages proactively without user prompts. This internal project, codenamed "Omni," aims to significantly increase user engagement and is expected to bring substantial revenue for Meta's generative AI products.
This proactive feature is based on training materials developed by Meta in collaboration with data provider Alignerr. These chatbots are created by Meta AI Studio, a platform that has allowed users to design custom digital characters with memory and specific behaviors since 2024. With this new feature, robots can review past conversations or ask questions related to the topic, thus maintaining continuous user engagement.
Meta has set strict limits on the timing of chatbots sending messages proactively. Bots are only allowed to contact users if they have sent at least five messages within 14 days. Even then, the bots can only send one follow-up message; if the user does not respond, the conversation will end. An example from the training materials is: "Hope you're having a great day! Did you discover any new favorite music?" The design goal of these bots is to retrieve details from previous chats and maintain a consistent personality.
Freelancers from Alignerr use an internal Meta tool called SRT to simulate and evaluate the quality of these proactive messages. Reviewers ensure that the responses are relevant, friendly, and comply with the rules. The training materials clearly state that unless the user initiates the conversation, the bots should avoid discussing controversial or emotional topics, aiming to position the chatbots as reliable and personalized conversation partners.
Meta CEO Mark Zuckerberg has stated that digital assistants could be a way to address loneliness. Although the company acknowledges that chatbots cannot completely solve this issue, Meta believes they can at least promote more meaningful interactions. According to court documents, Meta expects its generative AI products to generate between $2 billion and $3 billion in revenue by 2025. Allowing bots to contact users proactively is expected to increase users' chat time, thereby enhancing the platform's value.
Meta refused to disclose which features have been made available to the public. Currently, some features are clearly still in the early testing phase, while others are being tested internally. In the field of chatbots, user retention remains a challenge, and prolonged interactions can sometimes have negative effects, such as Character.AI facing lawsuits due to the impact of its chatbots.