Yesterday, on Wednesday, Google introduced the “Gemini” project, which it considers to be the most capable artificial intelligence model among its products, which demonstrated several capabilities in a video published by Google on its YouTube site.

As reported by CNBC, Gemini includes several types of artificial intelligence, including Gemini Ultra, which is the largest and most capable type; The Gemini Pro fulfills a wide range of tasks and the Gemini Nano is used for specific mobile tasks and devices.

The company now plans to make Gemini available to customers through Google Cloud so they can use it in their own applications.

According to the report, starting December 13, developers and enterprise customers will be able to access Gemini Pro via the Gemini API available in Google AI Studio or Google Cloud Vertex AI through Google Cloud Vertex tools; They are tools that developers use to build apps based on Google’s artificial intelligence technology.

Gemini will also be used to power Google products like Bart Chatbot and the Search Genetic Experience (SGE), which attempts to answer search queries with text in a chat format not yet widely available.

Google released a demonstration video showing off Gemini’s capabilities, in which the video presenter tested artificial intelligence Gemini on several cool challenges, for example:

Learn about ducks

In the first challenge, the program was asked to recognize a diagram drawn like a duck from several lines randomly drawn on a piece of paper, and the program was able to recognize the diagram from the lines.

The video host then showed a toy rubber duck and asked Jiminy: Can this duck float? The program was able to analyze the image and determine that the duck is a rubber toy and that it floats because it is made of a material less dense than water.

Learn about the shot

In one of the video clips shown by the host on the Gemini show, an amateur actor appears to imitate some movements. The host asked Gemini: Can you tell me which movie this is from? The AI ​​replied that it was a famous shot. Dodge the bullet in “The Matrix.”

Connect the lines

In another challenge, a piece of paper appeared with scattered dots and the host asked Jiminy to guess what the map was before connecting the dots, and he was able to find the map before connecting the dots.

sleight of hand

One of the exciting challenges is to put a paper ball under a cup and ask Jiminy under which cup the paper ball is.

Types and capabilities of the Gemini range

Gemini Ultra

Gemini Ultra is the first model to outperform human experts in understanding the MMLU, which tests global knowledge and problem-solving skills using a set of 57 subjects including mathematics, physics, history, law, medicine and ethics.

Gemini Ultra needs to understand nuance and logic in complex topics, the company said in a blog post on Wednesday.

“Gemini is the result of extensive collaborative efforts by teams across Google, including our colleagues at Google Research,” CEO Sundar Pichai wrote in a blog post yesterday.

“It is designed from the ground up to be multimedia, which means it can understand and act on different types of information, including text, code, audio, image and video,” he added.

Gemini Pro

Starting today, Google’s chatbot Bart will use Gemini Pro to help with advanced thinking, planning, comprehension and other complex skills.

Early next year, Google will launch Part Advance, which will use the Gemini Ultra, executives said on a call with reporters on Tuesday. According to the company, this marks the biggest update to the “Bard” program, an automated chat program similar to the “ChatGPT” program.

Gemini Nano

Gemini Nano is used to develop software and applications that can perform more complex tasks on mobile devices, smartphones or tablets.

Comparison with other artificial intelligence programs

The update comes 8 months after search company “Bart” first introduced it, and a year after “OpenAI” launched GPT Chat.

Executives said Tuesday that Gemini Pro outperformed GPT Chat version 3.5, but they avoided comparing it to GPT Chat version 4.

When asked if Google plans to charge for access to Bart Advanced, CC Hsiao, general manager of Bart Google, said it was focused on creating a good experience and that he didn’t have any details on monetizing the program.

Google reportedly delayed Gemini’s launch because it wasn’t ready, reminding followers of the grueling launch process for the company’s artificial intelligence tools faced earlier in the year.

Several reporters asked about the delay, and Collins replied that it would take more time to test more advanced models. Collins said Gemini is the most tested AI model the company has ever built and has “the most comprehensive safety ratings” of any Google model.

“It’s not just more efficient, it’s more efficient,” he added. “We still need significant computing power to train Gemini, but we have become very efficient in our ability to train these models,” he continued.

“This new modeling era is one of the biggest science and engineering efforts we’ve undertaken as a company,” Pichai said in a blog post yesterday, Wednesday. Open to people everywhere.