It will use it in its servers starting this year and it is 30% more efficient than the chips used in these applications today.
Google It has a new chip, designed in-house. Phones don't carry it For the image, Not even speakers Nest and is not available in any consumer products. It is a processor specially developed for performing calculations required in artificial intelligence applications and designed for enterprise servers.
Also known as Chip Print. Google says it's 30% more efficient than the ARM processors currently commonly used in these servers, or 50% more efficient than equivalent chips. Intel And has already started using some of its internal services, such as ads web light, oh Google Earth.
The chip will also be available in the service later this year Google Cloud and other Google tools for third-party companies, and their arrival represents a new twist in the artificial intelligence race facing many major tech companies.
Microsoft oh Amazon, It provides computer infrastructure to other organizations (via Blue Y AWS, respectively) have recently announced their intention to start using their own apps that specialize in artificial intelligence tasks instead. ARM processors or specific Intel models built for servers.
The benefits are obvious. Current artificial intelligence tools require enormous processing power and, therefore, huge energy consumption that hinders the operational efficiency of these network infrastructure services. Jumping to proprietary and efficient chips that specialize in performing the most common types of operations in the field of artificial intelligence allows us to reduce this cost.
And AI
At the annual conference dedicated to the Google Cloud platform that takes place these days las vegas, “Intelligence” and “artificial” are two words that keep ringing in the mouths of corporate executives.
For example, Google has also announced a new version of its artificial intelligence Gemini Gemini Answers is aimed at organizations that reduce the risk of containing incorrect data and allow you to group answers to a specific set of sources and data.
The company also showed off some of the improvements Gemini has made in recent months to the service's tools. workplace, etc. to create more detailed and complex texts Gmail or in instant translations Google Meet.
It also helped launch the conference Gemini 1.5 Pro, A more advanced version of Gemini is intended to compete directly with some of the more advanced language models GPT 4.0 Turbo.
Google faces a growing need to improve the image of its language model, which has been plagued by rollout issues. One of the more well-known (Asian or black players) is the decision to use multiple models of imaging equipment, resulting in historically inaccurate images. Asian or black soldiers in Nazi Germany During World War II.
