Elon Musk’s venture, Grok, unveiled an advanced AI image-creation tool late Tuesday that mirrors the minimal restrictions of its AI chatbot. This development enables users to craft and directly post manipulated images, such as Donald Trump indulging in cannabis on the Joe Rogan podcast, onto the X platform. Remarkably, the innovation stems from a new startup.
The X social media platform has been overwhelmed by a barrage of provocative images generated by this feature, stirring apprehension as we approach another election cycle. However, the chaos isn’t directly fueled by Musk’s AI firm, but rather by a new partner that aligns with his vision of creating a Grok dubbed as the “anti-woke chatbot,” unfettered by the limitations seen in OpenAI’s Dall-E or Google’s Imagen tools. On Tuesday, a collaboration between xAI and Black Forest Labs, an AI startup specializing in image and video creation and launched on the first of August, was disclosed to empower Grok’s image generator with its FLUX.1 model.
Hailing from Germany, Black Forest Labs emerged from its initial launch phase with a whopping $31 million in seed investment, with Andreessen Horowitz leading the funding round, as detailed in a press statement. Among its eminent investors are Y Combinator’s CEO Garry Tan and Oculus’s former CEO Brendan Iribe. The founding trio, Robin Rombach, Patrick Esser, and Andreas Blattmann, previously contributed as researchers to the creation of Stability AI’s Stable Diffusion models.
According to insights from Artificial Analysis, the FLUX.1 models from Black Forest Lab have overtaken competitors like Midjourney and OpenAI in user-rated quality assessments, as shown in their image arena.
The startup is dedicated to democratizing access to its AI models, offering open source image-generation tools on platforms such as Hugging Face and GitHub. Moreover, there’s an ambition to develop a text-to-video model in the near future.
Black Forest Labs has not yet responded to inquiries from TechCrunch for comments.
In its official release, the company professed their goal to “bolster trust in the safety of these models;” nevertheless, the outpour of AI-generated images on X, has potentially done the opposite. Numerous images, unconventional as Pikachu wielding firearms, generated with the tool from Grok and Black Forest Labs, couldn’t be replicated using the mechanisms provided by Google or OpenAI, highlighting the likely use of copyrighted material in the model’s development process.
That’s Exactly the Intent
This deliberate omission of restraints possibly influenced Musk’s selection for collaboration. Musk has vocally criticized the practice of imposing safeguards on AI, remarking in a 2022 tweet that teaching AI to align with “woke” principles is fundamentally hazardous. “The peril in programming AI with deceit is mortal,” he noted.
Anjney Midha, a board director at Black Forest Labs, shared on X a sequence of compare-and-contrast images from the first day’s output of Google Gemini versus the Grok and Flux collaboration. The conversation highlighted the challenges faced by Google Gemini in accurately representing historical figures, especially in terms of racial diversity. Midha expressed satisfaction with FLUX.1’s ability to circumvent these concerns, crediting xAI’s lead researcher Igor Babuschkin and his team for their decision.
Following the backlash, Google ceased Gemini’s people image generation capabilities in February, a restriction that remains in place.
A Torrent of Misinformation
This liberated approach towards safeguards might present challenges for Musk. The X platform came under scrutiny when it was inundated with AI-created explicit deepfake imagery of Taylor Swift. In addition to this, Grok frequently confronts users on X with hallucinated headlines on a near-weekly basis.
Just recently, five states’ secretaries called upon X to halt the dissemination of false information regarding Kamala Harris. Musk even redistributed a video earlier in the month, employing AI to mimic Harris’ voice admitting to being a “diversity hire.”
It appears Musk is keen on letting such disinformation proliferate on the platform. By allowing the unmarked posting of Grok’s AI-generated images directly on X, he effectively directs an uncontrolled stream of falsehoods into every user’s X feed.
Compiled by Techarena.au.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
