Tech giant Google has officially rolled out Gemini, its latest artificial intelligence model that it claims has outperformed OpenAI’s GPT-4.
On December 6, Sundar Pichai, CEO of Google, and Demis Hassabis, CEO and co-founder of Google DeepMind, announced the launch of Gemini in an enterprise. blog post
The AI model has been optimized for different sizes and use cases (Ultra, Pro, Nano) and designed to be multimodal to understand and combine different types of information.
The model is also advanced in mathematics and specialized coding, compared to OpenAI’s GPT-4, which cannot perform mathematics.
Meanwhile, Google claims its Ultra version achieves “peak performance” on 30 of the 32 academic benchmarks used in LLM (large language model) development.
Additionally, it scores 90% on a massive multitasking language comprehension (MMLU) test, surpassing the performance of human experts, according to Google.
Google’s chief scientist, Jeff Dean, said Gemini Ultra is the first model “to achieve human expert performance on MMLU on 57 subjects with a score above 90%.”
I’m very excited to share our work on Gemini today! Gemini is a family of multimodal models that demonstrate very strong capabilities in the areas of image, audio, video and text. Our highest performing model, Gemini Ultra, advances the state of the art in 30 of 32 benchmarks,… pic.twitter.com/sQfxBy9tpT
– Jeff Dean (@) (@JeffDean) December 6, 2023
The system was also designed from the ground up to reason transparently about text, images, audio and video, giving it a leg up on its competitors.
“We designed Gemini to be multimodal from the start,” Dean said before adding, “rather than starting with a purely textual model and then piggybacking on visual and audio encoders after the fact.”
Gemini also has advanced programming skills, including generating high-quality code using AlphaCode 2, an advanced code generation system. He can also solve complex programming problems and collaborate with developers.
According to AI expert Rowan Cheung, Gemini Pro outperformed GPT-3.5 in six of eight benchmarks, “making it the most powerful free chatbot on the market today.”
Related: IBM, Meta and others form ‘AI Alliance’ to advance AI development
For those interested in trying out the new AI model, a refined version of Gemini Pro has already been rolled out to Google’s version of ChatGPT – known as Bard – according to Google.
“This is the biggest upgrade to Bard since launch. It will be available in English in more than 170 countries and territories, and we plan to expand it to different modalities and support new languages and locations in the near future,” the company said.
Gemini is also rolling out to Google’s flagship phone, the Pixel 8 Pro.
“The Pixel 8 Pro is the first smartphone designed to run Gemini Nano, which brings new features like Summarize in the Recorder app and the rollout of Smart Reply in Gboard, starting with WhatsApp – with more messaging apps to follow come next year,” he said.
It will roll out to more Google products and services such as Search, Ads and Chrome “in the coming months,” he added.
The tech giant has also started experimenting with Gemini to power its web-dominant search engine to make search a generative experience.
Google unveiled Gemini earlier this year, touting its capabilities and claiming it would be more powerful than ChatGPT.
Review: AI Eye: Real-World Uses of AI in Crypto, Google’s GPT-4 Rival, AI Advantage for Bad Employees