Two weeks ago, Google began the rollout of its latest Gemini voice assistant for its smart speakers, and in a blog entry commemorating the milestone, the company appears to be quite positive. They even quoted a user stating, “My Google Home just updated with Gemini. It’s actually useful now beyond just checking today’s weather.” That’s rather subtle shade being thrown.
However, if you dig into the FAQ of the post, the situation looks less favorable. Among the most common inquiries from users of the new assistant are:
- The Gemini for Home voice assistant is having difficulty comprehending my request… (For instance: I attempted to set an alarm for 5 PM, but it registered for 5 AM.)
- The Gemini for Home voice assistant is not effectively managing my home devices. What actions should I take?
That’s not ideal.
Managing connected devices such as lights and locks and executing commands accurately are fundamental requirements for a home voice assistant. Since Gemini is set to become the primary voice assistant on all Google Home smart speakers and displays, it’s crucial that it performs these tasks effectively.
The launch of Gemini is part of a comprehensive redesign of Google Home. This update brings Gemini’s AI-infused intelligence to the Google Home app via a new Ask Home chatbot interface, allowing access to the capabilities of the Google Home app through natural language text commands. Additionally, it enhances Google’s Nest security cameras, enabling them to describe what they observe and relay this information into your smart home as part of a daily Home Brief.
The new app and enhanced camera functions were introduced in early October for all users, though certain features necessitate a subscription. To access the Gemini voice assistant, you’ll need to undergo a complicated Early Access sign-up process that commenced on October 28th, which is completely distinct from its Public Preview program. Then the waiting begins.
In a manner similar to Amazon with its Alexa Plus, Gemini is being rolled out very gradually to smart speakers. None at The Verge have gained access yet, and from the Google Home subreddit, it hasn’t been available to many other users either. Google has indicated that the general launch will not occur until next spring at the earliest, which coincides with the anticipated release of their new Google Home smart speaker.
Once Gemini becomes available on your devices, it promises to deliver a completely new voice assistant experience for Google’s smart displays and speakers. The Verge previewed a demo last month and encountered a more conversational assistant capable of comprehending natural language and context.
Google asserts that Gemini will excel at managing your smart home, in part by eliminating the need for stringent phrasing to execute commands and allowing multiple requests simultaneously. Thus, ideally, a command like “Hey Google, turn on the lights, except in the bedroom, turn on the TV, lock the front door, and make it warmer in here” should function seamlessly.
However, we haven’t had the opportunity to test this functionality, and some initial responses on Reddit indicated that chaining any commands isn’t yet operational. A more limited variation of this feature was already incorporated into Google Assistant. This suggests that Gemini may have regressed.
If you possess Gemini on your smart speakers, we would love to hear about your experiences so far. Please comment here or reach out to me at [email protected]. If you’re uncertain, just inquire with your speaker: “Ok Google, who are you?”
A smarter home
Experiencing difficulties with smart home control is likely a contributing factor to Gemini’s slow rollout. As evidenced by the rollout of Alexa Plus, merging generative AI with smart home functionalities poses significant challenges. Both the previous Google Assistant and the initial Alexa were constructed upon a command-and-control framework: when this statement is made, perform this action.
The emerging LLM assistants powered by generative AI do not function in the same manner, and the characteristics that enable these systems to engage in more conversational exchanges and grasp context also leave them prone to error. “LLMs excel at creativity, but struggle with providing consistent output with the same repeated command,” asserted Google Home’s Anish Kattukaran in an interview with The Verge.
This issue is evidently a challenge that both Amazon and Google continue to navigate. From testing Alexa Plus and the preliminary feedback on Gemini’s voice assistant capabilities, it is clear that there remains significant progress to be made before we can confidently entrust our smart homes to this new generation of AI assistant.
