In 2023, ChatGPT has grown day by day, and this ** AI tool is now widely known, and people have been expecting a strong competitor to challenge it. Google, OpenAI's main competitor, began working on an artificial intelligence tool called ChatGPT shortly after its release.
As we all know, the core of ChatGPT to give full play to its functions lies in relying on the support of GPT-4, a key technology. GPT-4 is a large language model that is trained on billions of sets of data on the Internet to grasp many elements such as images, text, and context, so that it can be functional. Over the past year, Bard has developed a broad knowledge of text, audio, images, and other forms of content. Its training data includes a large number of books, articles, libraries, recordings and other diverse resources, just like GPT-4, Bard also achieves a generalized understanding of language texts, providing users with effective answers and support.
Fast forward to the beginning of February 2024, and a leaked document shows that the artificial intelligence Bard will usher in a comprehensive major version update, which not only makes a leap forward in performance, but also has a new name, Gemini.
Inherited from Bard, Gemini is also multi-modal, multi-tasking, and multi-lingual, and can handle a variety of data types, including text, images, etc., and can understand and generate all kinds of content. Compared to Bard, Gemini is smarter, not only learning to write a complete **, but also learning to understand the logic of human texts, and showing its own unique creativity.
There's a paragraph like this in Google's presentation of Gemini's various technologies**. Gemini can not only depict various objects in real time, but also explain them in detail. What's even more rare is that Gemini can respond to all kinds of human questions about the objects it draws, communicate in multiple languages, and even design games based on the images it draws.
However, while this content is convincing, it is still somewhat misleading. In fact, Gemini doesn't answer people's questions in real time. What is more done is to analyze the context of the answer to the question in combination with the prompts given by humans.
In another demonstration**, Gemini's ability to recognize image combinations has also been enhanced. Subject showed Gemini a pancake with bacon** and a figure dancing at a carnival and asked him to guess the title of the movie, to which Gemini was able to accurately answer that it was The Breakfast Club.
In addition, Gemini can determine the weather and occasion for which different types of clothing are appropriate, find associations between different words and images, and help with children's math homework. It's fair to say that Gemini's potential for text, images, and a wide variety of digital content is endless.
So will this Gemini, which Google has worked so hard to develop, beat GPT-4 in 2024 and become ChatGPT's biggest rival?
For now, it's up for debate. The above features are not new, OpenAI has also experimented with GPT-4, and has performed quite well.
However, Google does not seem to think so, and in its product statement, it believes that Gemini beats GPT-4 in 30 out of 32 categories used to evaluate the model's knowledge, reasoning, perception, and other capabilities. It also scored 90%, making it the first model to outperform human experts on a large-scale multi-task language comprehension test. Gemini can also integrate 57 disciplines, including mathematics, physics, history, law, ethics, and medicine, and use a variety of knowledge to solve problems collaboratively.
Of course, all of the above data belongs to Google's boasting, and one cannot confirm how well Gemini performed outside of the controlled tests. Compared with ChatGPT, which has been widely used by netizens around the world, Google is still slowly but steadily advancing the test on a controllable scale. It seems that for the question of "Gemini has to compete with ChatGPT", at least Google is still afraid of it at present.
Text: Entrepreneurship in the world - rainyproject