Tech giant Google is challenging OpenAI with its new natural language model Gemini, which can now be used in its Bard chatbot. Gemini is the company’s most flexible Large Language Model (LLM) to date and is optimized for three different sizes: Ultra, Pro and Nano, which vary by the weight of the model and the types of tasks it can perform.
Google has decided to introduce Gemini to Bard in two phases. First Bard is updated with Gemini Pro version adapted to provide users with the best results and will be published initially in English in 170 countries and then released in other languages and additional countries. It will then integrate the Gemini Ultra version, thus offering the best AI model.
“We are already starting to experiment with Gemini in search, where it makes our Search Generative Experience (SGE) faster for users, with a 40% reduction in English latency in the US and an improvement in quality.”share them in the statement.
Gemini Pro is armed against ChatGPT
Google has been hesitant to introduce Gemini as the company has been working hard to conduct various verification tests and evaluations their performance in a wide variety of tasks. The tests they have taken include understanding natural images and audio and mathematical reasoning. The Google team observed that Gemini’s performance outperformed GPT 3.5 on several tasks.
Aside from that, Has received a score of 90.0% in MMLU (Massive Multitask Language Understanding), outperforming human experts. Our new grassroots approach to MMLU allows Gemini to harness its thinking skills Think more carefully before answering difficult questionsThis results in significant improvements over simply using your first impression.
«This new era of models represents one of the greatest scientific and engineering efforts we have undertaken as a company. I’m really excited about what’s to come and the possibilities Gemini will open up for people around the world.” he shared Sundar Pichai, CEO of Google and Alphabet.
The different versions of Gemini respond to:
- Gemini Ultra – Our largest and most powerful model for highly complex tasks.
- Gemini Pro – Our best model for scaling to a wide range of tasks.
- Gemini Nano – Our most efficient model for on-device tasks.
The statement also talked about Bard Advanced, the version with Gemini Ultra, which can understand and process different types of information such as text, images, audio, video and code as it has multimodal reasoning capabilities. . This update will be the most powerful.
Gemini Abilities
We designed Gemini to be multimodal from the start, pre-trained on different modalities from the start:
- Multimodal thinking that enables you to understand complex visual and written data. This capability makes it easy to extract valuable insights from large amounts of information and enables rapid digital advances in areas such as science and finance by efficiently processing hundreds of thousands of documents.
- Recognize and understand text, images, audio, and more simultaneously, allowing you to better understand nuanced information and answer questions about complicated topics. This makes it particularly suitable for explaining complex issues such as mathematics and physics.
- Multiple programming languages such as Python, Java, C++ and Go, and it stands out as a leading model in coding. Gemini Ultra stands out with key performance ratings such as HumanEval and Natural2Code. In addition, it serves as the basis for advanced coding systems such as AlphaCode and its advanced version AlphaCode 2, which are capable of solving challenges beyond coding, such as mathematical and theoretical computer science problems.