After months of anticipation, Gemini Live is finally here. You can start using Gemini Live on your Android phone right away, assuming you have subscribed to Gemini Advanced. I tested Gemini Live on my OnePlus phone, and it did not come across as revolutionary as we saw in the demos during Google I/O 2024. For one, Gemini Live currently does not support other modalities like images or real-time camera input, which was showcased with Project Astra. Right now, it only supports free-flowing audio conversations that work for the most part, but again, there are some fundamental issues related to how the feature has been implemented. But we will come to that later on. Let’s first go through our interactions with Gemini Live.
Testing the Capabilities of Gemini Live
Gemini Live has several features worth exploring, especially how it handles interruptions and multilingual conversations. Users can interrupt Gemini Live, and while it performs well most of the time, it occasionally continues speaking even after being interrupted. Furthermore, it can switch between languages, which makes it versatile in conversation. I tried speaking to Gemini in English, Hindi, and Bengali, and it handled the transitions smoothly, showcasing its multilingual capabilities.
Effective Multilingual Conversations
The ability to converse in multiple languages is a significant plus for Gemini Live. During my testing, I found that it could seamlessly switch between languages, which is invaluable for users who speak more than one language. This feature enhances the user experience and makes Gemini Live accessible to a wider audience. You can check out some demos online to see this in action!
Prepping for a Job Interview
In one of my interactions, I asked Gemini Live to assist me in preparing for a job interview in the AI field. It asked insightful questions about whether I would focus on research or application development, and based on my responses, it provided a list of relevant programming languages and frameworks. This adaptability shows that Gemini Live can be a helpful tool for job seekers and professionals alike.
Accuracy and Missteps: The Hallucination Issue
While Gemini Live performed admirably in many areas, there were instances where it displayed inaccuracies, often referred to as "hallucinations." For example, when I inquired about the latest updates on popular topics like Minecraft, the responses were sometimes outdated or incorrect. This inconsistency raises concerns about the reliability of the information it provides, especially for users relying on it for accurate data.
Analyzing the Hallucination Phenomenon
In several tests, Gemini Live struggled with certain subjects, particularly when discussing recent developments in popular media or technology. For instance, when I asked about the latest Minecraft update, it incorrectly stated the release dates and version numbers. This indicates that while Gemini Live can be an engaging conversational partner, it still has some gaps in its knowledge that need addressing.
Role-Playing Abilities
Another interesting feature is its role-playing capability. I asked Gemini Live to act as an English butler, and while it initially responded with a formal tone, it soon reverted to its normal conversational style. This inconsistency shows that Gemini Live may need further refinement in maintaining character roles over extended conversations.
Finding Information and Recommendations
One of the standout features of Gemini Live is its ability to provide recommendations and information. For example, when I asked for the best biryani places in Kolkata, it suggested popular restaurants like Arsalan and Karim’s, which are well-known among locals. Additionally, it provided useful suggestions for laptop repair services, demonstrating its capacity to assist users in finding reliable information.
Comparing Gemini Live with ChatGPT Advanced Voice Mode
When comparing Gemini Live with ChatGPT's Advanced Voice Mode, there are notable differences in performance. For instance, Gemini Live struggled to count quickly or repeat tongue twisters without pausing, unlike ChatGPT, which managed to maintain a natural rhythm. These differences highlight the need for further development in Gemini Live's voice processing and conversational abilities.
Exploring Emotional Recognition Capabilities
Gemini Live was marketed as a native multimodal model capable of understanding emotional tones in speech. However, during my testing, it failed to identify basic emotional cues and could not process sound beyond verbal communication. This limitation suggests that while the technology shows promise, it currently lacks the depth needed for a truly interactive and emotionally aware conversation.
Final Thoughts on Gemini Live's Implementation
To wrap up my experience, I found that Gemini Live, while impressive in some aspects, feels more like a sophisticated text-to-speech engine rather than a fully realized conversational AI. As it stands, it offers a glimpse of what could be possible, but it still has a long way to go in delivering the seamless and natural conversations we expect from advanced AI technologies.
Have you had a chance to try out Gemini Live? What has been your experience so far? Share your thoughts in the comments!
