- Google has launched Gemini AI, the company's most potent and capable model yet.
- Gemini AI is fully multimodal and can understand all input to give crazily accurate responses.
- The model will come with three variations and will fully launch next year. However, a version of it is already live on Google Bard.
Even though it is 2023, the AI race between companies continues to move at a huge pace. And while ChatGPT has had the upper hand for a long time, it appears that might not be the case anymore. Google has launched Gemini, its largest, most powerful, and capable AI model yet.
As detailed in a blog post released recently, Google detailed that its Gemini AI has been built from the ground up to be completely multimodal. The model has been pre-trained from the start for different modalities.
It was then further optimized and trained to accept all kinds of outputs to give out all kinds of output. This effectively translates to the AI model fully understanding text, images, videos, audio, code, and responding back.
Google showcased the same when it fed the model a simple video of two yarns and showed suggestions of what to make. The same case appeared when Gemini built a code for a dot matrix pattern it was shown. The fact the model can read and translate videos into outputs is crazy.
Gemini Beats OpenAI’s GPT-4
Besides showcasing its multimodality, Google also released a few benchmark figures comparing Gemini to OpenAI’s GPT-4 model. These include the Massive Multitask Language Understanding (MMLU) test, one of the most popular benchmarks. Surprisingly, Gemini AI beat GPT-4 in various areas, including reasoning, math, and even code.
Considering how ChatGPT is heralded as the prime GPT model when it comes to AI, Bard beating it is quite a huge deal that will surely cause upwaves.
Gemini AI: Three Different Variations
Google is going to release Gemini AI in three different models titled “Ultra,” “Pro,” and “Nano.” The most capable Ultra model is the most capable and made for highly complex tasks. The Pro is the best model for scaling, while Nano is compact.
Google has already planned a next year release for Ultra while the Gemini Pro model will be powering Bard amid other services. Gemini Nano will be inside the Pixel 8 phone, helping out with reply suggestions and summarising audio recordings.
Google’s Gemini AI has been trained and will be available in the English language and will be available in over 170 countries. The ability to analyze images and sounds will roll out later for Google Bard.
So, what do you think about Google’s Gemini? Do you think it can beat GPT-4? Drop your thoughts in the comments below.