Everything about Gemini AI: Google's assistant, Nano, Ultra
Google has recently launched its own Artificial Intelligence (AI) called Gemini. It is a new model that is available on the Bard platform and will gradually be extended to other Google services. Learn more about its features and how you can test this AI that promises to revolutionize the field.
What is Gemini AI?
Google Gemini is the Artificial Intelligence (AI) model created by the Mountain View company. First, we need to clarify that an AI model does not refer to applications or chatbots such as Google Bard, but it is the technology with which these types of applications work. Google wants Gemini to be the successor to PaLM, which is the AI model currently used by Bard. Gemini will gradually replace PaLM. This means that, even if you continue to use Bard, the answers it gives you will be much better.
Gemini is a multimodal model that can understand various types of information, for example, text you type and images, as well as audio and programming code. This makes it a very flexible AI model and one that, at least so far, has not been seen. Unlike GPT, Gemini can natively understand photos, text, audio, and more types of data without having to install any plug-ins.
How does Gemini AI work?
AI models are trained using a large amount of data. Data is collected from the internet and, through training algorithms, the system learns to understand things.
However, the case of Gemini is different, since it has been designed from scratch by Google. From the beginning, it was developed as a multimodal model. This means that from the beginning it has been trained to combine different modalities natively. That is, it can understand both text and images. According to Google it can, for example, relate and order objects in real-time, identify and complement drawings, as well as suggest songs according to the instructions you give it. If you want to see some exercises on how Gemini does all this, you can watch them in this video.
Also, Gemini introduces a new code generation system called AlphaCode2, which improves the understanding of complex mathematics and computer science theory. It is better at reasoning and its ability to understand code, so the answers are more reliable.
What are the types of Gemini AI that exist?
Gemini has three versions:
- Ultra: the most advanced and multimodal.
- Pro: more limited in terms of capabilities and functions.
- Nano: designed for devices with less computing and memory capabilities.
Gemini Nano is a big surprise, as it is an AI that can be implemented directly inside cell phones. This means that you don't have to use an app that connects to an AI server, as is the case with ChatGPT, but Gemini will come directly on the smartphone and you can use it offline.
When will Gemini AI be available?
Gemini and its various versions will arrive in a staggered manner over the next few months. So far, Gemini Pro has started to arrive on Google Bard. At the moment, it is available in English in more than 180 countries, but none of them belong to the European Union. On the other hand, Google is scheduled to release Bard Advance, an enhanced version of Bard with Gemini Ultra integrated. This will happen next year, although exact dates are not yet known. While Gemini Nano will arrive on Pixel 8 Pro handsets.
In addition to Google Bard, Gemini will debut in other services and applications of the company. There will be an AICore, which will be a service for app creators to use and implement in their own apps. Gemini will also come to the Mountain View company's search engine, as well as to Google Ads, Duet AI, and Google Chrome.
Although we don't know the exact dates, we can tell you that from December 13, developers can access Gemini Pro through the API in Google AI Studio or Vertex AI.
How to use Gemini AI
You can use Gemini within Bard. Just log in with your Google account to the Bard website. At the moment, it is a trial and you may encounter bugs or experience other bugs.