Automation vs Personalization: What SmartReply on YouTube Means for Content Creators

essidsolutions

Google launches SmartReply on YouTube. We look at how this changes the game for content creators on the world’s largest video streaming platform.

Google announced the rollout of its SmartReply feature for YouTube creators yesterday. The AI-based technology that taps into a deep retrieval system has been designed to help YouTube creators, particularly the ones with large followers, engage with their viewers more efficiently. It automates comment replies by offering smart suggestions, similar to prompts we’ve grown accustomed to using on Gmail, Android Messages, Android Wear, and the Play Developer Console.

Putting the Smarts in the SmartReply on YouTube Creator Studio

The company in its blog postOpens a new window , revealed that it had to significantly reprogram this feature for YouTube as language semantics on the platform is more dynamic, often including abbreviations, slangs, inconsistent punctuations, and emojis.

YouTube creators can now respond to viewer comments in English and Spanish. We can expect Google to expand this feature to include other languages in the near future as well.

Also Read: YouTube Marketing 101Opens a new window

The beauty of SmartReply on YouTube is that unlike earlier versions of the AI feature seen on Gmail and Android, Google decided to build a single cross-lingual model for all supported languages instead of developing language-specific models. This enables the AI program to deliver mixed-language, contextual prompts, thereby reducing the complexity involved in switching between languages, numbers or emojis.
The development team at Google was able to cluster the suggestions produced by the model, giving it its cross-linguistic capabilities. So, when users probe the model with an input comment, the AI suggests the most relevant response based on the proximity of the input comment to applicable clusters. The image below demonstrates how SmartReply leverages clusters to provide seamless multi-lingual support.

Also Read: 5 Keys for Building Your Brand’s YouTube FollowingOpens a new window

A 2D projection of the model encodings when presented with a hypothetical comment and a small list of potential replies. The neighborhood surrounding English comments (black color) consists of appropriate replies in English and their counterparts in Spanish and Arabic. Note that the network learned to align English replies with their translations without access to any parallel corpus.

Source: Google AI BlogOpens a new window

Google states that the SmartReply feature only makes suggestions when it is very likely to be useful. It uses contextual signals to discern whether the creator would reply to a comment and only then displays suggestions that have a high chance of providing a sensible and specific response. “To accomplish this, we trained auxiliary models to identify which comments should trigger the SmartReply feature,” says Rami Al-Rfou, Research Scientist at Google Research.

Does SmartReply on YouTube Signal the End of Personalization for Content Creators?

We’ve all been there – lazily accepting Google suggestions when responding to emails. “Thanks, that sounds great,” or “Sure, I’m interested,” seem all too familiar for Gmail users. While it saves users valuable time, it also camouflages a user’s own voice, their language personality, reducing replies to chatbot-esque homogeneity. While that may not be a big concern for people using SmartReply for work emails or responding to long email threads, it is a bit of a concern for YouTube creators looking to establish a personal connection with their followers or viewers.

Also Read: What the Worldwide Rollout of YouTube’s Newest Ad Format Means for MarketersOpens a new window

At a time when personalizationOpens a new window is a top-of-mind priority for customer experience leaders, features like SmartReply can take away from the very experience that brands and content creators seek to deliver.
There’s no doubt that AI systems are getting better at mimicking human language every day. However, you’ll still be able to tell auto-responses from human responses. This begs the question – is there a trade-off between automation and personalization? Yes, in the context of SmartReply, there will be an inevitable trade-off that brands and content creators will have to contend with.

The idea is to apply human discretion to using SmartReply selectively when responding to comments.
As YouTube outlined in its release, the system will only offer suggestions when it deems them to be useful. This leaves much room for marketers and content creators to personalize all high-impact responses.
The silver lining is, creators could use the SmartReply feature to engage with viewers and followers in multiple languages, adding personalization back into the mix. Like with all things AI, human intelligence, or rather discretion, in this case, should guide how marketers and creators engage with this feature on the platform.