Google's artificial intelligence chatbot Gemini, previously under scrutiny for generating historically inaccurate images, is relaunching its people-generation feature with improved safeguards, as reported by TechCrunch. The functionality will initially be available to paid Gemini Advanced, Business, and Enterprise subscribers in an English-language early access test. Google's move aims to gather feedback before a wider release, though a timeline for when the feature will become available to free Gemini users has not been disclosed.
Amidst the February backlash, Sundar Pichai, CEO of Google, extended apologies, while Demis Hassabis of DeepMind promised a swift resolution. Despite the urgency and reports of extensive work hours from Google staff, the revising process was anything but fast. The updated Imagen 3 model, embedded within Gemini, promises to depict people more ethically. Google asserts that it's "significantly reduced the potential for undesirable responses,” according to statements shared with TechCrunch. However, specifics about Imagen 3's training data remained wrapped apart from general reassurances about improved fairness and safety measures.
In addition to image generation, Google is integrating GEMS, an expert system that operates on the Gemini platform, to assist users with various tasks. These AI-powered tools function as subject-matter experts, ranging from coding assistance to event brainstorming. According to TechCrunch, GEMS will be available on desktop and mobile in "most languages" but is not yet supported in Gemini Live.
Despite the recent upgrades, Google has implemented strict content guidelines for Gemini. Users will not be able to create photorealistic images of public figures or minors or engage in generating graphics with gory, violent, or sexual content. Dave Citron, Google's senior director of product management for Gemini, reinforces this approach, stating, "Of course, not every image Gemini creates will be perfect, but we'll continue to listen to feedback from early access Gemini Advanced users as we keep improving," as noted by The Verge. The aim is to refine the tool while closely monitoring its use and output in real-world applications.
.jpg){kind=link}