We tend to be a little too hasty to equate ChatGPT, in its various versions, with generative artificial intelligence tout court.
However, it is evident, and inevitable, that all the large companies in the tech sector are more or less secretly developing their own AI capable of generating new content, textual and perhaps not only. And to compete with GPT-4, the most advanced OpenAI product so far.
If Apple is working on the elusive Ajax Project, in recent days it has transpired that Meta is also developing a generative artificial intelligence capable of rivaling that of OpenAI, or rather surpassing it in power. AND now there’s news about Google Gemini, developed in the summer, which could even be five times more powerful than GPT-4. Here’s what we know about the project so far.
Google Gemini
The Mountain View company, already in the arena of generative AI with Bard, in recent months is said to have secretly developed much more advanced software, which according to rumors it will be five times more powerful than GPT-4.
It should be called Google Gemini, and its launch should be soon. Here is the (so far little) information we have.
What we know
The news was reported by the Reuters agency, which in an article published on Friday 15 September reports information that The Information would have obtained directly.
The Mountain View company’s project is called Google Gemini, it is ready to compete with GPT-4 and has already been tested by a small number of companies.
Reuters reports that Google Gemini is a collection of large language models, capable of different declinations. The main one remains the textual one, i.e. the role of a conversational chatbot capable of summarizing texts or producing them from scratch based on the user’s requests. And the areas of use can be the most varied, from compilation to creative ones, such as the production of narrative texts or songs.
But the software will also be able to help computer engineers generate images and even write code.
Google Gemini 5 times more powerful than GPT-4?
Already last August 28th an interesting article was published on SemiAnalysis, a very important publication for the semiconductor industry.
Analyzing the structure of Google Gemini, the article shows that the Mountain View software can be trained with 5 times more computing power than GPT-4, and this already by the end of 2023. But by the end of 2024, Gemini’s computing power could be 20 times greater than its competitor developed by OpenAI.
Google Gemini imminent?
Google did not provide responses to Reuters, therefore The release times are not known, but they should be close.
Companies, which have tested a pre-release version of the tool, will be able to access it from Vertex AI.
In August, Alphabet (the company that owns Google) introduced artificial intelligence to its search engine for users in India and Japan, with features capable of showing textual or visual results based on requests received.
Alphabet has also made its AI-powered tools available to enterprise customers, for a monthly fee of $30 per user.
Bard’s latest model
Meanwhile, Google, as we reported in an article, has just presented the most advanced Bard model.
Thanks to the new update, Bard’s responses are faster and more preciseand AI features can be accessed in multiple languages, including Italian.
You can also ask for more personalized answers, or modify them based on specific requests. And Bard can retrieve information simultaneously from several Google tools, such as Gmail, Documents, Drive, Maps, YouTube, Google Flights and Hotels.
AI Smart report arrives in Italy
In collaboration with the School of Management of the Polytechnic of Milan, Google has launched AI Smart Report in our country.
This is a tool dedicated to Italian SMEswhich allows (for free) companies to “discover the most relevant AI applications for their industry and understand how to best leverage them.”
177 AI technologies were analyzed to identify the most effective applications across 8 industries. The tool is designed to reduce certain cost items for SMEs, such as raw materials and energy, by up to 10% and bring “significant overall efficiency improvements with AI-based solutions”.