Google’s Gemini LLM is here to take on Open AI’s GPT-4; here’s what we know

OpenAI’s generative AI chatbot, ChatGPT has been around for precisely one year and one week now, and is by far, one of the biggest names in the world of artificial intelligence. In this context, multiple industry bigwigs have begun to work on and come out with their offerings to take on the same.

One of them happens to be Google, which has just launched its latest Large Language Model (LLM), called Gemini to the public. Here’s all you need to know.

Google Gemini: What it is and what it can do

The Gemini LLM is a successor to Google’s PaLM and PaLM-2, as well as LaMDA LLMs. The new Gemini LLM, Google claims, beats OpenAI’s GPT-4 on 30 out of 32 benchmarks, when tested side by side. Some of these benchmarks are science, technology, engineering, and mathematics (STEM) subjects, commonsense reasoning, code generation, reading comprehension, and more.

Moreover, the Gemini model is also multimodal. That is to say, it can handle multiple types of inputs, in the form of text, code, images, or even audio.

Gemini will be made available to users in three sizes – Gemini Nano, for small, on-device tasks such as that on a smartphone, Gemini Pro – a model that can run both- everyday tasks such as text or image generation, as well complex tasks such as coding and analysis with equal ease, and Gemini Ultra, which according to Google, specialises in highly complex tasks.

When is Gemini being rolled out?

The Gemini Pro and Gemini Nano models are being rolled out to the public starting today on Google Bard, and through its Pixel smartphones respectively. The former can be accessed on both smartphones, as well as laptops. The latter is being integrated into the recorder app for summarising voice notes, as well as smart replies for Gboard on WhatsApp. Moreover, Gemini is also being integrated into Google’s Search Generative Experience (SGE) as well, where, the company claims, it has reduced latency by as much as 40 per cent.

ALSO READ: How to summarise online articles using this AI tool from Google

Gemini Ultra, on the other hand, won’t be rolled out today, as it is still undergoing trust and safety checks. Google says that it will make the same available for testing to developers and select customers starting early next year.

Keeping safety in mind, Google says that it has added new protections to its AI Principles policy. In addition to this, the company also claims it has conducted extensive research into risk areas, applied adversarial testing techniques to identify blind spots, as well as build dedicated filters for filtering out out violence and negative stereotypes.