Google Gemini: A Game-changing AI That Surpasses GPT-4 and Human Experts

Google has announced the launch of a new AI model called Gemini, which the company claims can outperform both OpenAI’s GPT-4 model and “expert level” humans in various intelligence tests. Gemini is capable of handling text, audio, and video. The model is available in three versions: Nano, Pro, and Ultra, with increasing size and capability. The Nano version has 1.8 billion parameters for slower phones and 3.25 billion parameters for more powerful devices. It is designed to run locally on smartphones. The Pro and Ultra versions have not disclosed their size, parameters, or training data. However, Google claims that the Ultra version exceeds the capabilities of all existing AI models and scored 90% on the industry-standard MMLU benchmark, surpassing the performance of “expert level” humans. The Gemini model has beaten other models, such as OpenAI’s GPT3.5 and Anthropic’s Claude 2, in multiple benchmark tests. Google plans to integrate the Pro model into its online chatbot, Bard, and release a more advanced version called Bard Advanced early next year, featuring the larger Gemini Ultra model. The new version of Bard will be initially available in English in over 170 countries, with support for other languages and additional capabilities like audio and image interaction coming in the future. Gemini has been trained on text, images, and sound and demonstrates adaptability to different tasks and mediums. It is considered Google’s largest, most capable, and most general model to date, with claimed superiority in software development as well.