OpenAI Vs Google: Gemini Live Rolls Out To Rival ChatGPT Voice Mode

OpenAI Vs Google: The tech giant is rolling out Gemini Live, a voice chat feature for its AI assistant Gemini to rival OpenAI’s new Advanced Voice Mode for ChatGPT. Announced at the 2024 Made by Google event, Gemini Live can be accessed by users of its advanced package.

OpenAI Vs Google: Gemini Live to Rival ChatGPT Voice Mode

In a thread on X, the company has announced the launch of Gemini Live to challenge OpenAI’s recently released Advanced Voice Mode for ChatGPT.

Launched at the 2024 event, the feature has become available for Gemini Advanced users. The feature aims at making the interaction with AI more fluent and less rigid, allowing users to cut off, switch to another topic, or continue the conversation at any time, just like in a phone conversation.

Meet Gemini Live: a new way to have more natural conversations with Gemini. 💬

💡 Brainstorm ideas
❓ Interrupt to ask questions
⏸️ Pause a chat and come back to it

Now rolling out in English to Gemini Advanced subscribers on @Android phones → https://t.co/bTHxAnOeGn… pic.twitter.com/WysHGhxSVe

— Google DeepMind (@GoogleDeepMind) August 13, 2024

The feature is unique in the new speech engine that Google says produces coherent, emotionally intoning, and natural conversational flow in multiple turns. There are 10 natural-sounding voices with the option of the AI’s being able to mimic the user’s speech in real-time. This hands-free mode can be done in the background or even when the phone is locked, this way the user can do other things and not interrupt the conversation.

Move to Enhance AI Interaction

Consequently, the long and complex conversations are made possible through the Gemini 1.5 Pro and Gemini 1.5 Flash models of the AI assistant using a longer context window than in the generative AI models. This enables Gemini Live to engage in longer conversations and also store information more efficiently.

Besides voice commands, the company has confirmed that multimodal input, which was demonstrated for the first time at Google I/O 2024, will be integrated into Gemini Live before the end of the year. These features will help the AI to understand and answer the visual prompts like images and videos making the AI more versatile. At the moment, the update is available only in English, and only for Android devices, however, support for more languages and iOS is planned soon.

As the company rolls out the feature, it is also preparing to introduce additional features and integrations with its services. In the coming weeks, Gemini will gain new extensions for Google apps such as Calendar, Keep, Tasks, and YouTube Music. These updates will allow users to perform tasks like creating playlists, setting reminders, and managing their schedules more efficiently through voice commands.

Furthermore, android users will be able to enable Gemini on top of any app using the power button or voice commands in the near future. This feature will allow users to communicate with Gemini in other applications and ask questions or generate content as simple as an image that can be easily incorporated into the user’s work.

OpenAI Challenges with Advanced Voice Mode

In the OpenAI Vs Google battle, the latter’s Advanced Voice Mode for ChatGPT, despite being a new concept, has faced some problems during its limited alpha testing. The mode, which seeks to provide a more natural conversation, has been criticised for making users dependent on the AI due to the realistic voice interactions.

Consequently, OpenAI released a safety concern that was highlighted recently was the possibility of the formation of social relations between users and AI, which may lead to adverse effects on interpersonal relationships.

We're releasing a new iteration of SWE-bench, in collaboration with the original authors, to more reliably evaluate AI models on their ability to solve real-world software issues. https://t.co/qJuLpCdSWJ

— OpenAI (@OpenAI) August 13, 2024

In addition, the company has been experimenting with improving the software engineering skills of its AI models. To overcome these problems, the firm has recently released a human-evaluated subset of the SWE-bench benchmark that better gauges AI models’ capacity to tackle actual software problems. This move is a continuation of the measures being taken to guarantee that developments in AI are safe and useful in real life.