Google to fix glitches in Gemini AI image tool and relaunch it soon

Google has paused its AI tool that creates images of people, after users reported inaccuracies in some historical depictions. Google Deep Mind CEO Demis Hassabis said that the company aims to relaunch the tool in a few weeks, after fixing the issues.

Google launched the image-generation capability earlier this month, as part of its Gemini AI models, which also power its chatbot and assistant. The tool allows users to enter prompts, such as “a medieval knight” or “a famous singer”, and get an image of a person that matches the description.

However, some users noticed that the tool generated historical images that were sometimes inaccurate or misleading. For example, when asked for a German soldier in 1943, the tool showed a racially diverse set of soldiers wearing Nazi uniforms. When asked for a historically accurate depiction of a medieval British king, the tool showed a woman ruler and a man of Asian descent, among others.

The tool also failed to recognize some prominent figures, such as Google’s own founders, Larry Page and Sergey Brin, whom it depicted as Asian men. Users shared screenshots of these errors on social media, and criticized Google for its lack of quality control and cultural sensitivity.

Google takes the tool offline and promises to fix it

Google took the tool offline on Thursday, and apologized for the mistakes. Hassabis said that the tool was not working the way the company intended, and that it was working on fixing the glitches. He said that the company plans to relaunch the tool in the next few weeks.

Hassabis said that the tool was based on a large corpus of data, and that it was not possible to manually check every possible prompt and image. He said that the company was using automated methods to detect and correct the errors, and that it was also seeking feedback from users and experts.

Hassabis said that the tool was meant to be a fun and creative way for users to explore the possibilities of generative AI, and that it was not intended to be a source of historical or factual information. He said that the company was committed to ensuring that the tool was accurate, respectful, and ethical.

Google faces competition and challenges in generative AI

The image-generation tool is part of Google’s efforts to compete with OpenAI, a Microsoft-backed company that created ChatGPT, a powerful generative AI chatbot that can produce coherent and realistic text on various topics. Google launched its own chatbot, Bard, in 2020, and renamed it Gemini earlier this month. Gemini also offers paid subscription plans, which users can choose for better reasoning capabilities from the AI model.

Generative AI is a branch of artificial intelligence that can create new content, such as text, images, music, or code, based on a given input or data. Generative AI has many potential applications, such as enhancing creativity, personalization, and communication, but it also poses many challenges, such as privacy, bias, and misinformation.

Google is not the only company that is developing generative AI tools. Anthropic, a San Francisco-based startup, is working on Claude 2, a rival chatbot that claims to be more ethical and transparent than ChatGPT. Anthropic has attracted investments from Google and Amazon, who are also interested in the potential of generative AI for their cloud services.

Leave a Reply

Your email address will not be published. Required fields are marked *