The race for artificial intelligence (AI) continues, and now it’s Google’s turn to play one of its cards. The giant has just released Gemini, a multimodal generative AI model. A direct competitor to OpenAI’s ChatGPT, it’s set to carve out a place for itself in what is now a crowded landscape. Let’s take a closer look at this newcomer
Gemini, Google’s AI with big ambitions
Gemini comes in three versions: Gemini Ultra, Gemini Pro and Gemini Nano. Each is adapted to the task in hand, as well as to the device used – Nano, for example, is designed for mobile devices. Gemini Ultra is the most powerful version, designed for highly complex tasks.
This model is capable of consulting different types of information (texts, images, videos), in order to extract data. According to Google’s chief scientist, Jeff Dean, Gemini is even capable of surpassing human performance when it comes to massive multitask language understanding.
Unlike ChatGPT, Gemini was designed from the outset to be multimodal. In contrast, OpenAI’s AI started from a textual model before extending its capabilities. In addition, Gemini can fetch information from the Web, unlike the free version of ChatGPT, whose data stops at a certain date.
The most powerful AI of the moment?
According to Google, Gemini outperforms ChatGPT 4V on 6 of the 8 benchmarks used to evaluate AI capabilities. This would make it the most powerful model available today. Gemini is capable of generating text, images, lines of code or video.
Comparison between Gemini and GPT-4V on eight benchmarks
Gemini has been integrated into Bard, Google’s text-based generative AI model. So it’s already possible to test its capabilities. Gemini Nano will also be natively integrated into the Pixel 8 Pro, Google’s flagship smartphone at the moment.
Gemini will also be used on other services offered by Google, including search, advertising and the Google Chrome browser. These changes will take place over the coming months, according to the release. Gemini Ultra, the most powerful version of the model, is currently being tested. It will be offered to a limited number of customers and developers early next year.
Indeed, Google is communicating “responsibility” and the need to evolve AI models in a reasoned way. This is a counterpoint to OpenAI, a model that has caused concern in recent weeks. All in all, this shows that the AI race is on, and that Google firmly believes in the future of these technologies.
