A Google announced on Wednesday (6/12) its latest advancement in artificial intelligence (AI): Gemini.
This new multimodal major language model (LLM) promises to surpass GPT-4, used in the popular ChatGPT chatbot, and marks a significant milestone in AI development.
Unlike its predecessors, Gemini stands out for its unique ability to “understand, operate and combine different types of information”, covering text, audio, image, video and programming languages.
Demis Hassabis, CEO and co-founder of Google DeepMind, emphasizes that the creation of Gemini seeks to make it less like “intelligent software” and more intuitive and practical.
More about Gemini, Google’s new AI
Flexibility is another notable feature of Gemini, being the most versatile model ever created by Google.
Available in three different sizes — Ultra, Pro and Nano — it can be implemented in data centers or mobile devicesthus providing a variety of uses.
Image: Shutterstock/LookerStudio
Currently, Gemini 1.0 is already available on Google Bard in 170 countries, including Brazil, although restricted to the English language.
The expectation is that, by 2024, the most powerful model will be integrated into the chatbotalso expanding to other languages.
The Google Pixel 8 Pro already incorporates Gemini, and in the future, it will be accessible directly in Google search, Chrome and other company services and products.
In the mobile sphere, the Gemini Nano demonstrates its efficiency when performing tasks in the audio recording application and in Gboard’s Smart Reply function on Whatsapp.
However, the more robust version, Gemini Ultra, scheduled for launch next year, will initially go through a phase restricted to some customers, developers, partners and security and responsibility experts, aiming to collect feedback and improve the product.
Regarding Gemini’s capabilities, its ability to extract insights from extensive documents and its ability to simultaneously understand text, images, audio and more stands out.
This multimodal approach allows for more contextual responses, especially when faced with complex questions.
Google claims that Gemini succeeded in 30 of 32 academic benchmarks, outperforming human experts in diverse areas such as mathematics, physics, history, law, medicine and ethics.
Compared to the GPT-4Gemini Ultra demonstrated superiority in text, reasoning, mathematics and programming, in addition to excelling in multimodal tests involving images, videos and audio.
In summary, Gemini emerges as a promising innovation in AI, challenging the capabilities of ChatGPT and promising a more thoughtful and accurate approach when answering challenging questions.
Support our work ❤️
If you enjoyed this article, consider leaving a tip to help us keep publishing great content.
