Major Update for Google Bard: More Language Support and AI Image Generation

Google Bard will use SynthID to make all AI images easily identifiable from real photographs.

On February 1, Google Bard received a significant update that gives the chatbot driven by artificial intelligence (AI) several new features. The most significant improvement is creating images from text inputs; this is now possible. It is still unable to generate image-to-image outputs, though. In addition, the tech behemoth announced that Google Bard will now support over 40 languages and has been expanded to over 230 countries and territories. Just one day prior, during Google's quarterly earnings call, the company disclosed that Google Bard Advanced, which is powered by Gemini Ultra, will come with a paid subscription.

Google announced the news in a blog post that included a summary of the AI chatbot's updates. Given that many of its competitors, including Baidu's Ernie Bot, Microsoft Copilot, and OpenAI's ChatGPT Plus, have had this feature for some time, the company's late but significant addition of the AI image generator is noteworthy. The Imagen 2 model, the same one that powers under-tested ImageFX and Vertex AI, is the source of Google Bard's image generation capabilities.

Features of Google Bard's AI image generator

Multiple paragraph-long prompts will be supported for text input by the new AI image generator. Google promises that the resulting images will be wide-ranging, realistic, and of excellent quality. When we at Gadgets 360 tested the feature, we discovered that the AI model was able to produce reasonably well-reproduced, high-quality images that complied with the prompt. Nevertheless, the resolution of 1536 x 1536 is fixed for all generated images. Additionally, the images lack photorealism and are, for the most part, easily identifiable as digitally produced. Additionally, the chatbot declines to respond to any requests requiring it to produce images of actual people, which is probably going to reduce the chances of deepfakes (creating AI-generated images of people and objects, which appear real).

Furthermore, Google has employed SynthID to clearly identify the images produced by Google Bard as AI-generated. In August 2023, Google's DeepMind division unveiled SynthID, a tool for watermarking and identifying images created by artificial intelligence. The company stated at the time of launch that "this technology embeds a digital watermark directly into the pixels of an image, making it detectable for identification but imperceptible to the human eye."

In addition to enhancing image creation capabilities, Google has extended Bard's reach to over 230 nations and regions. More than 40 languages are now supported by it, including Arabic, Bengali, Tamil, and Urdu. It was previously limited to supporting English and available in 170 countries.

Additionally, the multilingual 'double-check' feature has been added by the tech giant. Users can identify the sections of the response that are sourced from the internet along with citations by using the double check feature. It also draws attention to the sections that lack any kind of reference. This feature was introduced to reduce the frequency of AI hallucinations, or the times when AI confidently responds incorrectly. Clicking the G icon located beneath the generated response will allow you to access it.