This will also interest you
[EN VIDÉO] Artificial intelligence that adjusts the angle at which the photo is taken Interactive point-based manipulation of the generative image collector. © DraGAN Project
After a difficult year trying to catch up on OpenAI and its chatbot ChatGPTChatGPT, GoogleGoogle may well have finally succeeded in taking the lead in the race for artificial intelligence. The firm just announced its new AI called Gemini.
Gemini is a natively multimodal AI. In other words, it is not a large language model trained on text and to which other models have been grafted to enable it to understand images or audio. From the beginning, Gemini was designed to understand text, audio, video, and even programming code. This allows it, for example, to understand the actions of a user, to make comments on them and to interact with them, as demonstrated in a demonstration video. Please note that the background of this video is explained on this page. The demonstration was done with static images and text, which is still impressive, but a little less sensationalist.
Demonstration of Gemini’s ability to understand text and image simultaneously. (In English, enable French subtitles.) © Google
Gemini beats humans on one of the main tests
Google says Gemini is able to beat all of its competitors on 30 of 32 major tests intended to evaluate large language models. It’s even the first model to beat humans on the MMLU test (Massive Multitask Language Understanding). Here again, you have to be careful because this is the Ultra version of Gemini, which will not be the one used for Google Bard.
Because Google Gemini will be available in three versions. The basic version is Gemini Pro, and will be integrated into Google products and in particular Bard. Note that it will initially be available in 170 countries in English only, as during the initial launch of Bard. Gemini Nano is a lite version capable of running on devices mobilesmobiles and will be integrated into the Pixel 8 Pro smartphone. Gemini Ultra is a more advanced version, but not yet ready for a mainstream launch. It will be available to developers and businesses early next year.
Google could soon launch Gemini AI, its answer to OpenAI’s GPT-4
Article d’Edward BackEdward Backpublished on 09/16/2023
Google is currently working on a new generation artificial intelligence called Gemini, and expected “for theautumnautumn “. According to the media The InformationGoogle has reportedly already opened access to Gemini for a small number of companies, meaning the general public launch is close.
OpenAI’s ChatGPT took Google by surprise, which has since been trying to catch up. After a rather catastrophic launch of its own Bard AI, operating using the LaMDA language model, the chatbot was switched to the PaLM 2 model. However, it struggles to compete with GPT-4GPT-4.
This is why Google is working on a new artificial intelligence (AI) called Gemini, which is planned for the fall. According to The Information, Google has reportedly allowed a small number of companies access to a preview version of Gemini, meaning the launch is getting closer. Google plans to offer versions of different sizes in order to sell developers access at different prices, and the version currently being tested would be smaller than the final version. According to one of the site’s sources, the final version would be of a size similar to GPT-4. However, this does not mean that its performance will be similar.
An AI that will know users through other Google services
Gemini has the distinction of being a truly multimodal AI, allowing it to analyze or generate images, in addition to generating text. She would be able to explain a graph, for example, or even understand voice commands. In addition, its ability to generate programming code would be greatly improved, allowing it to compete with GitHub Copilot developed by OpenAI. Gemini would have a notable advantage over ChatGPT, since the AI would have access to data from different Google products, allowing it to better understand user intentions to provide them with more precise responses. AI would also be less subject to hallucinationshallucinations.
Gemini AI should be offered to businesses via its platform cloud computing. It will also be used to add new features in Google Workspace and to power its Bard chatbot.
Google is preparing a powerful artificial intelligence for the start of the school year that surpasses ChatGPT
Google intends to go well beyond simple competition with ChatGPT, by developing powerful AI. Called Gemini, it should arrive during the fall.
Article by Sylvain BigetSylvain Bigetpublished on 08/16/2023
Despite the launch of its Bard chatbot, Google has not really recovered from the ground occupied by OpenAI with ChatGPT, then its integration into Microsoft’s Bing. This does not mean that the firm is giving up, since it has been developing artificial intelligence (AI) for a very long time and gradually integrating them into its products. But, according to anonymous internal sources who testified for the site The Information, Google’s labs are reportedly preparing a major novelty based around AI and which could be launched this fall. The project, which would be called “Gemini”, would even be developed by the Google Brain and DeepMind teams. With it, Google would not just compete with ChatGPT, but easily surpass it with a multifunction tool combining the capabilities of its large language model (LLM) and AI image generation.
An AI that concentrates all the talents of AI
In addition to the creation of contextual images, Gemini could make it possible to control software with voice, or to analyze an organization chart, according to the informant. Gemini would be building his learning from texts, images and also transcriptionstranscriptions videos YouTubeYouTube. Google would have made a point of ensuring that the AI does not assimilate data by infringing copyright. Ultimately, AI should naturally fit into Google’s various products, particularly its online office suite. Gemini will also be available as a developer tool through a Google Cloud subscription.
Proof of the importance of this new project for Google, Sergei Brin, the co-founder of the firm, would be personally involved and would notably take care of the recruitment of experts. For the analyst The Informationbeing a candidate for Gemini may not be a good idea, since mixing different types of AI in a single product could also become the main weakness of the tool.
rewrite this content and keep HTML tags