The most powerful model created by humans is about to be released!

Google, a technology giant that has been known for its AI capabilities for many years, fell behind in the race against OpenAI's large models in the first half of the year. However, Google may soon have a chance to turn the tables - Gemini, the most powerful model developed jointly by Google Brain and DeepMind, is about to meet the public.

According to insiders cited by The Information, Google has provided an early version of Gemini to a small number of companies, and some external developers have already started using Gemini. Google may integrate Gemini with its family of applications and also sell API interfaces to businesses through Google Cloud.

More Powerful, More Accurate

Gemini directly competes with OpenAI GPT-4 and is the result of the merger of Google's two most powerful AI teams. In addition to the functions commonly found in AI chatbots (summarizing text, composing emails, etc.), Gemini may also support image generation. Furthermore, there may be significant improvements in its coding capabilities.

The development of Gemini is a bold move by Google. The failure of the Bard chatbot in the first half of the year has left Google embarrassed in the face of newcomers like OpenAI. Subsequently, the previously feuding Google Brain and DeepMind teams merged, and the two co-founders, Sergey Brin and Larry Page, who had stepped down, returned urgently to formulate tactics, even personally coding and participating in the training of Gemini.

How does Gemini, which is built by assembling Google's AI talents and consuming massive computing power, perform?

According to those who participated in Gemini testing, Gemini has at least one advantage over GPT-4: Google's proprietary data, which allows Gemini to better understand user needs and seemingly fabricate answers less frequently.

According to The Information, compared to existing large models, Gemini has better code generation capabilities, and Google hopes to gain market share from Microsoft/OpenAI's GitHub Copilot with Gemini.

In addition, Gemini may also have chart analysis capabilities, being able to interpret the meaning of charts and operate browsers and other software through voice and text commands, similar to a more intelligent Siri.

Wallstreetcn previously mentioned that Demis Hassabis, co-founder and CEO of Google DeepMind, confirmed that Gemini combines the technology behind GPT-4 and AlphaGo, and has unprecedented new capabilities, such as the ability to plan or solve problems and text analysis capabilities.

TOB Sales is the Focus: Google Cloud Strives to Catch Up with Microsoft Cloud

Google hopes to attract more users, especially in the cloud computing business, by leveraging Gemini for its products.

Currently, Google Cloud lags behind Amazon AWS and Microsoft Azure in terms of market share. Microsoft executives have stated in their earnings conference call that the integration of OpenAI's GPT large model has greatly enhanced the attractiveness of Azure, and it is expected to generate $10 billion in revenue from AI capabilities in the future, but Microsoft has not provided a specific timeline.

According to The Information, another source revealed that Google plans to allow enterprises to use Gemini through its Google Cloud Vertex AI service and will release versions with different parameters for developers to choose flexibly.

The source added that Google currently provides developers with a version with a larger number of parameters, which is equivalent to GPT-4 in terms of parameter size, but it has not been demonstrated to the public yet.

Since OpenAI started selling GPT-4 earlier this year, Google has actively provided its existing commercial models to more developers in recent months. In May of this year, Google announced that it would provide a set of Palm 2's LLM to Google Cloud customers through Vertex AI. Recently, Google also offered a one-month free trial of its large model to customers through the coding platform startup Replit.