In an attempt to compete in the quickly expanding field of generative artificial intelligence, Google on Wednesday unveiled its most ambitious project to date: Gemini, an AI model intended to outperform OpenAI’s GPT models and enhance everything from Google’s consumer apps to Android smartphones.
Google’s announcement, which presented Gemini as the company’s largest and most capable AI model and ushered in a “Gemini era” in which the tech giant envisions its model being used in every setting, from large corporations to consumer devices like the Google Pixel 8 Pro, demonstrated the breadth of the company’s ambitions.
In contrast to current AI models, which generally handle a single kind of user prompt, like only text or images, Google claimed that Gemini was designed to be “multimodal.” This indicates that it can receive inputs in the form of various media formats, including text, graphics, audio, video, and programming code.
In a blog post, Google CEO Sundar Pichai stated, “This new era of models represents one of the biggest science and engineering efforts we’ve undertaken as a company.”
Google announced on Wednesday that Bard, its in-house AI chatbot, has already been enhanced with a Gemini model version. The company also plans to add Gemini to other popular products, such as Google’s search engine and Chrome web browser, which are used by billions of people globally.
The announcement is an attempt to recover the initiative after OpenAI’s ChatGPT, which was released abruptly and to great public acclaim a year ago, seemingly caught Google and other tech giants off guard. This led to a rush to develop generative AI tools within the industry and a global conversation about the advantages and disadvantages of AI.
Additionally, it is an attempt to spread generative AI as far as possible within Google’s empire. Gemini 1.0 is available in three different sizes, according to the company: Nano, which is optimised for mobile devices and app developers; Pro, which is the default model made for a variety of tasks and customers; and Ultra, which is Google’s most advanced AI model to date, still undergoing safety testing.
The launch on Wednesday also aimed to highlight Google’s developments in cloud computing, which is an essential tool for AI developers. According to the company, Gemini was trained on a new generation of potent cloud-based processors, which when combined can train large AI models almost three times more quickly than the previous version. This technology, which Google cloud customers will also have access to, has the potential to significantly advance the AI sector by increasing accessibility to AI training and supporting Google’s third-place standing in the public cloud services market. However, it’s unclear how Google’s AI chips compare to those of industry leaders like Nvidia.
According to the company, Google’s Gemini model beat competing AI models in testing on more than two dozen benchmarks that are frequently used by AI researchers to assess an algorithm’s capacity for multistep reasoning, mathematics, and reading comprehension.
On a conference call on Tuesday, Eli Collins, vice president of product at Google DeepMind, informed reporters that they do believe it is opening up new avenues in all areas.
However, he seemed to acknowledge the ongoing risk that users may receive misleading results from AI models, citing concerns expressed by researchers, policymakers, and civil society organisations.
Large language models “are still capable of hallucinating, a term AI researchers use when AI systems make up facts and get stuff wrong – but with extreme confidence,” Collins said. However, Collins acknowledged that Google has made significant progress in improving factuality in Gemini.  They have additional techniques to improve the accuracy of responses when they integrate these models into products like Bard.
Due to these risks, Google announced on Wednesday that Gemini Ultra, the most advanced version of the model, will not be released to enterprise customers or developers until early in the following year. Instead, it will be released gradually to a select group of customers, partners, developers, and safety and responsibility experts for early feedback and experimentation.
As promised to the Biden administration earlier this summer, Gemini Ultra is currently going through third-party safety evaluations, or red-teaming.
