Google releases AI-driven chatbot and aide, Gemini

First, there were talking digital assistants like Siri, Alexa and Google Assistant. Then there were online chatbots like ChatGPT and Google Bard. Now, the two are merging.

Yesterday, Google introduced Gemini, a smartphone app that behaves like a talking digital assistant as well as a conversational chatbot.

Responding to voice and text requests, it can answer questions, write poetry, generate images, draft emails, analyse personal photos and take other actions, like setting a timer or placing a phone call.

Immediately available to English speakers in more than 150 countries and territories, including the United States, Gemini replaces Bard and Google Assistant. It is underpinned by artificial intelligence technology that the company has been developing since early last year.

The new app is designed to do an array of tasks, including serving as a personal tutor, helping computer programmers with coding tasks and even preparing job hunters for interviews, Google said.

When ChatGPT arrived from OpenAI at the end of 2022, wowing the public with the way it answered questions, wrote term papers and generated computer code, Google found itself playing catch-up. Like other tech giants, the company had spent years developing similar technology but had not released a product as advanced as ChatGPT.

Google released its own chatbot, Bard, in March to middling reviews. In the weeks that followed, it merged its two leading AI labs — Google Brain and DeepMind — and announced that the combined lab was developing new AI technology called Gemini.

Gemini is what researchers call a large language model, or LLM, a mathematical system that can learn skills by analysing vast amounts of data, including books, computer programs and online chatter. By identifying patterns in all that text, an LLM can learn to generate text on its own. That means it can write poetry, generate computer code and chat.

It can also get facts wrong or “hallucinate” — make stuff up.

Gemini is a “multimodal” system, responding to both images and sounds. After analysing a maths problem with graphs, shapes and other images, it could answer the question much as a high school student would.

In December, Google used a limited version of this technology to upgrade Bard. Now, it has retired the Bard name and is releasing a more powerful version of the technology through the Gemini app, available on Android phones and the web. A version for iPhones will arrive “in the coming weeks”, Google said.