Google Releases Gemini Generative AI Model Designed to Match and Exceed OpenAI’s GPT-4 and ChatGPT
Google has officially unveiled its new Gemini generative AI model after nearly a year of anticipation. Gemini comes in three sizes and is officially multimodal out of the gate, though with limited availability for many of the features showcased by Google in an avalanche of demos and announcements that seem to subtly suggest Google wants Gemini to immediately replicate and improve upon everything OpenAI has done with GPT-4, ChatGPT, and other generative AI products in the last year or so.
Gemini is available in three sizes, in ascending order: Nano, Pro, and Ultra. Gemini Nano is optimized for native and on-edge devices for offline use on Android devices. Gemini Pro is set to power many Google AI services, including Google’s generative AI chatbot Bard. Gemini Ultra caters to data centers and enterprise applications as well as many multimodal services. Google is also pitching Gemini as a great coding assistant, with the new AlphaCode 2 system released in tandem with Gemini to help programmers do anything from brainstorming ideas to writing code. All the Gemini versions employ Google’s Tensor Processing Unit (TPU) accelerators to operate.
“AI has the potential to create opportunities — from the everyday to the extraordinary — for people everywhere. It will bring new waves of innovation and economic progress and drive knowledge, learning, creativity and productivity on a scale we haven’t seen before,” Google CEO Sundar Pichai wrote in a blog post about Gemini. “These are the first models of the Gemini era and the first realization of the vision we had when we formed Google DeepMind earlier this year. This new era of models represents one of the biggest science and engineering efforts we’ve undertaken as a company. I’m genuinely excited for what’s ahead, and for the opportunities Gemini will unlock for people everywhere.”
Google tested Gemini in nearly three dozen benchmark tests, notably competing mainly against OpenAI’s GPT-4 and its variants. The researchers claim Gemini outperformed GPT-4 in 30 out of 32 tests. Gemini reportedly excelled particularly in understanding and interacting with video and audio, stemming from Google’s multimodal foundational approach. You can see one example of its reasoning capacity and visual acumen below.
Double Star AI
Gemini’s official launch is tied to only very limited accessibility for now. Google Pixel 8 Pro phones are rolling out updated generative AI and synthetic media tools powered by Gemini Nano, but only in English. Most other Gemini experiences are still in early access or still being tested. For instance, Google Search is experimenting with Gemini to speed up results, while Bard is still trialing a fine-tuned version of Gemini Pro. Gemini Ultra isn’t slated for release until early next year.
While framing Gemini’s formidable abilities, Google repeatedly underscored its meticulous vetting to uphold reliability and safety standards at scale. These span bias, toxicity, misinformation, and security evaluations via leading research processes and partners – applying lessons from Google’s extensive technology accountability experience. That’s partly why Google postponed releasing Gemini Ultra, as it has more tests to pass.
“Gemini is the result of large-scale collaborative efforts by teams across Google, including our colleagues at Google Research. It was built from the ground up to be multimodal, which means it can generalize and seamlessly understand, operate across and combine different types of information including text, code, audio, image and video,” Google DeepMind CEO Demis Hassabis wrote in the announcement. “Gemini is also our most flexible model yet — able to efficiently run on everything from data centers to mobile devices. Its state-of-the-art capabilities will significantly enhance the way developers and enterprise customers build and scale with AI.”
Still, developers and enterprise customers will only have access to Gemini Pro through Google Generative AI Studio or Vertex AI in Google Cloud starting Dec. 13. Vertex has been a hub for Google’s generative AI offerings this year. That includes the Vertex AI Search and Vertex AI Conversation app pipelines for search and conversational AI fueled by LLMs. Meanwhile, the new Duet generative AI assistant for Google Workspace apps provides natural language assistance to automate tasks, generate content, answer questions, and recommend actions based on Workspace usage and context. Clearly, Google has decided to theme its generative AI products around duos, like duetting singers and the Gemini twins of myth and constellation fame.