Easetrack introduces Gemini, a new AI model from Google. Gemini outperforms GPT-4 on almost all benchmarks. Gemini is now available on Google Bard.
Artificial intelligence (AI) has been on the rise in recent years due to its ability to think, create, and analyze data in a way that is similar to the human brain. AI comes in many forms and has already become part of our everyday lives, such as self-driving cars, Face ID, smart home devices, and Google Maps.
Recently, Google launched Gemini, its most powerful AI model to date. This is a major step forward in the development of AI, with its wide range of capabilities and its potential to rival OpenAI.
Every technological advancement is an opportunity to accelerate human progress and improve our lives. With the launch of Gemini, Google is opening up a new opportunity for AI to be a force for good in the world.
We all know that the launch of ChatGPT, OpenAI’s large language model, ushered in the era of AI. Since then, the technology has continued to develop and has quickly become a part of our everyday lives.
The success of GPT-4 from OpenAI, which is majority-owned by Microsoft, has led to a flurry of activity in the AI space. From Bard from Google, Erine from Baidu, Meta AI from Meta, Tako from TikTok, to many smaller companies, everyone is vying to raise the bar for AI.
New Gemini AI with cutting-edge potential
Many of you may be familiar with Google’s AI, Bard, but Google’s parent company, Deepmind, has recently launched Gemini, a new language model that has been upgraded and has even more potential than the latest language model, GPT-4.
Gemini was developed from a massive database, leading to the most capable AI ever. The model can understand a variety of input data, from text, code, images, audio, and video, and can run on any device from supercomputers to smartphones.
The key highlight of this AI model is its superior capabilities. From the MMLU test, which combines academic knowledge and general understanding in over 57 fields, Gemini scored a high of 90%, while GPT-4, the latest model from OpenAI, scored 86.4%.
In addition to the MMLU test score, Gemini’s capabilities are superior to GPT-4 in almost every aspect, from mathematics, coding, to understanding various types of data, including images, audio, and video. All have higher understanding capabilities than GPT-4.
With this potential, Gemini will be able to access and understand all forms of data received to process and find answers to reasoning in a logical, comprehensive, and accurate way, so that it can find more complex and reasoned answers in the future.
“As a result of a major collaboration between the Google and Google Research teams, we built Gemini from the ground up to be multi-modal, meaning it can summarize and understand, process, and aggregate different types of data.”
Google Gemini will be divided into 3 sizes, namely:
1. GEMINI Ultra: is the largest model, designed to be used in large data centers or in organizations with complex workflows.
2. GEMINI Pro: is a smaller model, suitable for a variety of uses, and is currently being used as the backend for another Google chatbot called Bard.
3. GEMINI Nano: is the smallest model, supports running on smartphones or other portable devices, and is suitable for general use.
Currently, we can already use Bard with the GEMINI Pro model, but it is still limited to the United States and 170 other English-speaking countries. Other languages or inputs will be updated in the future. Google Pixel 8 Pro users are expected to start seeing new features from GEMINI Nano, and GEMINI Ultra is expected to be available in 2024.
Ready to compete with OpenAI’s GPT-4
Gemini, a new AI model from Google and OpenAI, has been compared to GPT-4. In benchmark tests, Gemini outperformed GPT-4 in 30 of 32 categories. This is because Gemini is designed to handle a variety of data, including text, images, videos, and audio. This gives Gemini a particular advantage in understanding information, including interacting with videos and audio.
Sundar Pichai, CEO of Google, and Demis Hassabis, CEO of Google DeepMind, said that Gemini is the most capable model ever built. The goal of Gemini is to make AI feel less like software and more like a helpful, easy-to-use assistant or expert.
The launch of Gemini is a major step for Google in the development of AI. It may not yet change the world as Google has aimed to, but it may help Google catch up to OpenAI’s technology in the race to create more powerful AI than is currently available. Gemini could become the future of Google, and it’s not too late.
Whether Gemini will be a formidable competitor to ChatGPT is up for you to decide. But Googlers are adamant that this will be the beginning of a new era for Google. They are not stopping there, as they are working hard to expand Gemini’s capabilities into future versions. This includes developing memory planning and adding context to data processing for even better responses. This will undoubtedly change the way billions of people work and live around the world.
Source: nipa, blog.google