Unveiled last December, Gemini Pro, Google’s super AI, finally integrates Bard into our regions. It is so powerful that the chatbot will adopt its name and have its own dedicated app as well as its paid subscription.
Little by little, Google is catching up on OpenAI and Microsoft in the race for artificial intelligence! The Internet giant was taken a little by surprise by the release of ChatGPT in November 2022 and by its partnership with the Redmond firm. Although he talked to us about AI throughout 2023, he had difficulty convincing his competitors, particularly during his first presentation of Bard. But things accelerated with the integration of Gemini Pro, one of the company’s most powerful large language models (LLM), into the chatbot a few weeks ago – to put it simply, it’s the competitor of GPT-4. Gemini is a so-called “multimodal” model, that is to say capable of processing several data sources of different nature, such as text, audio or images (see our article). This versatility allows him to tackle complex tasks in various fields, such as mathematics, physics, and even computer programming.
However, the Old Continent was left aside and, unless you spoke English, manually changed the URL and used a VPN, it was impossible to take advantage of it. But Google has announced its official deployment in some 230 countries, including France, and support for 40 additional languages. This is accompanied by new functions, such as response verification and image generation. We can’t wait to try!
Bard Gemini Pro: a free, built-in image generator
Gemini exists in three versions: Gemini Nano – a lightweight version to work locally on a smartphone –, Gemini Pro – a more versatile and advanced version, to carry out many tasks – and Gemini Ultra – the most powerful version, to carry out particularly complex tasks. This is the second option which is integrated here into Bard, replacing the old PaLM 2 language model in order to offer more performance and reliability. We’ve done some testing and the results are impressive. When we asked for information on the How does it work site for example, the chatbot was able to give us the strong points, but also the weak points of the site.
In addition to the quality of the answers, this change is accompanied by the extension of the “Check answer” function to the French language, in order to submit information found on the Internet to the AI for verification – Bard analyzes the request and indicates whether the content found online is true or not. The chatbot also includes a free image generator, called ImageFX. To do this, the AI uses the Imagen 2 model. As with DALL-E 3 – which requires a paid subscription – you just need to enter the description of the desired image for the magic of generative AI to work. . Note, however, that the prompts for this task must always be entered in English. Google explains that it has set up a system of “expressive chips” on which you just have to click to experiment with variations of our creations. We tried to generate images from Bard while located in France, but the chatbot let us know that it did not yet have the function. There is no doubt that the problem should be resolved shortly.
To prevent any drift, the Internet giant uses SynthID to integrate digital watermarks into the pixels of the generated illustrations – they are imperceptible to the human eye – in order to clearly distinguish a human work from an AI-generated visual. To put it simply, if we come across an image generated by Google tools, we will be informed by trying to get more information about it from the search engine or by checking its metadata. Furthermore, Imagen 2’s security measures have been strengthened to block any questionable prompts. It will be difficult, if not impossible, to generate violent, offensive or sexually explicit images. A certain number of filters have also been put in place to prevent images featuring real-world personalities from being generated.
Gemini Pro: how to access it?
For those who would like to test Gemini, it is possible to access the Pro version through Bard, available at this address. To be sure you are using Gemini, and not PaLM 2, go to the “Latest updates” section, which is displayed in a banner at the top of the screen. You should read there the mention of the recent addition of the new model. Pixel 8 Pro owners can also take advantage of Gemini Nano, which allows them to use new capabilities like the “summarize” function of the Recorder application, or automatic replies generated in Gboard – first in WhatsApp, and in other messaging apps starting next year. To ensure you benefit from it, check that the most recent update is on December 6.
Gemini will gradually be rolled out to other tools, such as the Google search engine, Google Ads and the Chrome browser. Besides, a leaked note reveals that Google is preparing to rename Bard to Gemini on February 7 with, as a result, a dedicated application. The chatbot should also welcome the Ultra version of Gemini, within a subscription called Gemini Advanced, which will make it “much more capable of performing highly complex tasks such as coding, logical reasoning, following nuanced instructions, and creative collaboration.”