Google to Relaunch AI Image Generation Tool After Pausing It

Google Debuts Improved Version of Gemini AI Tool

Google’s artificial intelligence image generation tool offered through Gemini is set to be relaunched in the next few weeks after being paused due to inaccuracies in historical depictions.

The CEO of Google DeepMind, Demis Hassabis, announced the news at the Mobile World Congress in Barcelona, Reuters reported Monday (Feb. 26).

Google introduced earlier this month the Gemini AI models for image generation, which allow users to create images of people, the report said. The tool was taken offline after some users pointed out inaccuracies in the generated images.

Alphabet’s shares saw a decline of 3.5% following Monday’s news, impacting the S&P 500 index, per the report.

Previously known as Bard, the AI chatbot was rebranded as Gemini and rolled out with paid subscription plans for better reasoning capabilities, according to the report. Google has been in a race to develop AI software to rival that of OpenAI’s ChatGPT, with the aim to enhance generative AI technology.

Bob O’Donnell, chief analyst at TECHnalysis Research, told Reuters that persisting glitches and inaccuracies in AI development could raise concerns among users.

The move by Google comes after a similar incident last year, where inaccuracies in promotional videos caused a drop in shares by as much as 9%, per the report. The company is now focused on refining the AI tool to ensure accuracy and reliability.

In December, Google postponed the launch of Gemini as the AI offering faced challenges with non-English queries.

In that case, sources within Google revealed that the AI technology in Gemini did not reliably process and respond to certain non-English queries, leading to the postponement.

On Feb. 8, when rebranding Bard as Gemini, Google also launched Gemini Advanced, a new chatbot giving users access to the company’s largest AI model, Ultra 1.0.

A week later, on Feb. 15, the company said its Gemini AI tool received a new update called Gemini 1.5. Hassabis said at the time that “Gemini 1.5 delivers dramatically enhanced performance” and can process more information than competing AI models.

For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.