Google Delays ‘Gemini’ Launch to 2024

Google has privately delayed plans to launch in-person events to showcase Gemini, widely considered to be its answer to OpenAI’s previously top-of-the-line large language model, GPT-4.

A series of Gemini-related events was scheduled to take place next week in New York, Washington D.C. and California, but those have been scrapped on the orders of Google CEO Sundar Pichai, according to The Information.

Instead, the company will give journalists and software developers a first look through a virtual preview that could happen as soon as this week. Google also has been privately giving demonstrations of Gemini to business partners, but cloud customers would not get access until next year.

The in-person Gemini events are now slated for January. The reported reason for scrapping the prior schedule is due to the model’s poor handling of some non-English queries, although one wonders whether Gemini was retracted for further retooling following OpenAI's upgrade of GPT-4. At OpenAI's first developers conference in November, the startup unveiled GPT-4 Turbo.

Google had been meaning to showcase Gemini to various stakeholders – for example, the Washington event was aimed at policymakers who are in the process of crafting AI regulations.

Google losing ground in AI?

Related:Google DeepMind's Six Levels of AGI

Developed by Google DeepMind, Gemini is among the so-called ‘frontier models,’ foundational large language models that were the focal point of the global AI Safety Summit, where nations met to collaborate on mitigating AI harms ranging from concerns about bias all the way to the threat of human extinction.

Google executives reportedly had been concerned that ChatGPT has become a household name, with Pichai keen for the company to get the initial version of Gemini out as soon as possible to directly challenge OpenAI’s chatbot.

Another concern is that Microsoft's Copilot for its productivity software is becoming a significant business, the publication said. Microsoft has invested $13 billion in OpenAI and is using the startup's large language models to infuse generative AI into its products and services.

Google has largely kept Gemini under wraps since it was first revealed at the company’s I/O event in May, with only a small number of companies given access.

Since then, it has been up there with GPT-5 as being among the most anticipated AI models. Gemini is set to power various generative AI uses on YouTube and Workspace. An October leak saw that Gemini would be used to power tools in Google’s web development platform, Makersuite.

Related:Leaked: New Web Development Tools from Google

Gemini will be multimodal, capable of handling text, videos and images as well as being able to tap into other tools and APIs – similar to how OpenAI has upgraded ChatGPT. Pichai said early tests of the model saw “impressive multimodal capabilities not seen in prior models.”