![](https://static.toiimg.com/thumb/msid-105728786,width-1070,height-580,imgsize-9584,resizemode-75,overlay-toi_sw,pt-32,y_pad-40/photo.jpg)
Google is readying an advanced large language model that could take on OpenAI’s GPT-4, the Gemini. This next-generation AI model of Google is said to have multimodal capabilities, meaning it generates and comprehends texts, images, and more based on a prompt. The word around the town is that the Gemini could outperform OpenAI’s GPT-4 model. While it was expected to debut this year, the launch has been seemingly postponed to next year, as the next-gen model is said to be struggling with non-English languages.
Google first announced the Gemini at its annual developer’s conference earlier this year.At the time, Google said that the Gemini is still created from scratch to be multimodal, and is highly efficient with tool integrations. Still in training at the time, the model was said to show “impressive” model capabilities that have been seen in earlier models. Google said that the model will be made available once tested rigorously and fine-tuned, and appears taking its time to ensure Gemini is thoroughly tested before its release.
According to The Information, two unnamed sources with knowledge of the matter revealed that launch events, which were supposed to take place in New York, Washington, and California next week, have been rescheduled for early 2024. This decision was reportedly made due to concerns about the AI’s reliability in responding to non-English prompts and inquiries.
In an interaction with Business Insider, Sissie Hsiao, Google’s VP and manager of Bard and Google Assistant, shared that Gemini is capable of producing impressive results. For instance, if someone asks Gemini to provide them with three images demonstrating how to ice a three-layer cake while they’re baking, it can create entirely new images to display the steps. Hsiao noted that these are not images sourced from the internet, but rather, unique images that Gemini generates.
Google is anticipated to employ Gemini in Google’s current AI and AI-powered products, such as Bard, Google Assistant, and Search.
Google first announced the Gemini at its annual developer’s conference earlier this year.At the time, Google said that the Gemini is still created from scratch to be multimodal, and is highly efficient with tool integrations. Still in training at the time, the model was said to show “impressive” model capabilities that have been seen in earlier models. Google said that the model will be made available once tested rigorously and fine-tuned, and appears taking its time to ensure Gemini is thoroughly tested before its release.
According to The Information, two unnamed sources with knowledge of the matter revealed that launch events, which were supposed to take place in New York, Washington, and California next week, have been rescheduled for early 2024. This decision was reportedly made due to concerns about the AI’s reliability in responding to non-English prompts and inquiries.
In an interaction with Business Insider, Sissie Hsiao, Google’s VP and manager of Bard and Google Assistant, shared that Gemini is capable of producing impressive results. For instance, if someone asks Gemini to provide them with three images demonstrating how to ice a three-layer cake while they’re baking, it can create entirely new images to display the steps. Hsiao noted that these are not images sourced from the internet, but rather, unique images that Gemini generates.
Google is anticipated to employ Gemini in Google’s current AI and AI-powered products, such as Bard, Google Assistant, and Search.