Meta has introduced Imagine with Meta AI, an independent AI image generator that transforms text-based queries into high-resolution images swiftly and seamlessly. This free tool leverages Emu, Meta's foundational image model, to compete with counterparts such as OpenAI's Dall-E, Midjourney, and Stable Diffusion.
Previously confined to Meta's messaging platforms, Imagine now emerges as a versatile image generator, producing four images per prompt. The AI-generated images carry both a visible watermark in the bottom left corner and an imperceptible watermark detectable only by a corresponding AI model. This dual watermark system aims to bolster traceability, transparency, and resilience against conventional image manipulations like cropping, color changes, and screenshots.
In response to past controversies, including Meta's racially biased AI sticker generator, concerns persist regarding the safeguards in place for Imagine with Meta, per TechCrunch.
Watermarks Coming Soon
While users were not afforded a pre-launch testing opportunity, Meta has committed to implementing additional watermarks on content generated by Imagine in the coming weeks, reinforcing transparency and traceability. Various companies, including Imatag and Steg.AI, utilize watermarking techniques for generative art, claiming resilience to resizing and edits. Microsoft and Google have adopted AI-based watermarking standards. Pressure on tech firms is increasing to clearly indicate AI-generated works, especially amidst concerns about Deepfakes and AI-generated child abuse images.
However, Gizmodo reported the outcomes from Imagine with Meta AI have been less than stellar. Some of my queries were muddled, providing confusing and unrelated results. In comparison to Dall-E, Midjourney, and even smaller rivals like Playground AI, Meta's offering appears significantly restricted. Nevertheless, it comes with the advantage of being both accessible and unlimited, a feature not shared by all its competitors.
Meta Boosting More AI Features
The evolution of Meta's AI capabilities extends beyond Imagine. The tech giant commits to enhancing its Meta AI assistant, pledging more detailed responses on mobile platforms and more accurate summaries of search results. Users can anticipate the integration of Reels into meta-AI chats. Additionally, Meta's AI characters, inspired by celebrities like Snoop Dogg and Tom Brady, are now fully accessible on WhatsApp, Messenger, and Instagram in the US.
Meta's blog post outlines various updates, including exploring generative AI for offering suggestions, like writing birthday greetings, as per The Verge. The announcement strategically coincides with Google's Gemini release, positioning Meta as a formidable player in the evolving AI landscape. While Google's revelations may garner significant attention, Meta's innovative AI tools showcase practical utility.
Meta is also expanding integration with Microsoft Bing search in its chatbots and experimenting with incorporating "long-term memory" into some bots, according to Barron's. This feature enables bots to learn from user conversations over time, adapting to individual experiences and preferences.
Last week, researchers found AI image generators are environmentally taxing, using as much energy as charging a smartphone for one image. Despite the environmental concerns, the race for more powerful generative AI continues between Meta, Google, and Microsoft, with high energy consumption expected.
Related Article: X Unveils New Feature: Broadcast Community Posts to All Followers
