Google and Grok diverge on safety with AI image generation
We’re back after a break in weekly updates last week. A lot has happened since then!
what to know for now
🖼️ Google's Imagen 3 AI goes public, balancing quality with censorship. Latest text-to-image model available on ImageFX platform for U.S. users. Improvements in texture and word recognition noted, but strict content filters frustrate some. Contrasts with xAI's less restricted Grok-2, highlighting industry divide on AI ethics.
🎨 Grok's unchecked image generation sparks controversy on X platform. xAI's chatbot allows users to create potentially offensive or misleading images, raising concerns about content moderation and AI safety. Lack of robust guardrails contrasts with stricter policies of competitors, potentially inviting regulatory scrutiny. Read more
📱 Google unveils Pixel 9 lineup with advanced Gemini AI features. New models include three Pro versions and a foldable device. Gemini Live, Assistant upgrades, and AI-powered camera tools showcase Google's AI integration. Privacy considerations addressed with on-device processing for select features. Read more
🤖 Walmart's AI rewrite of 850M product listings boosts efficiency. Generative AI accelerates product catalog updates, enhancing e-commerce capabilities. New AI shopping assistant planned. Implementation spans multiple business areas, prioritizing responsible use while maintaining low-cost approach. Read more
🧪 AI Research of the Week
Imagen 3
From Google DeepMindJake’s Take: Imagen 3, Google DeepMind's latest text-to-image model, demonstrates high performance in photorealistic generations and following of complex prompts, outperforming other state-of-the-art models across various quality metrics. The model incorporates extensive safety and responsibility measures, including pre-training data curation, post-training interventions, and rigorous evaluations to mitigate risks and biases.
This upgrade to Imagen and its general availability will likely drive a continued increase in the commercial application of AI imagery, potentially reshaping industries from digital art/advertising design to a different focus on product design and virtual world creation.
what to know for later
🍓 OpenAI CEO's garden photo sparks "Strawberry" project speculation. Sam Altman's innocent summer post fuels rumors about secretive AI initiative. Experts suggest potential breakthrough in AI capabilities, possibly reigniting industry excitement. Speculation highlights tech community's anticipation for OpenAI's next major development. Read more
🚪 OpenAI faces leadership exodus amid industry competition and legal challenges. Co-founders John Schulman and Ilya Sutskever depart, with Schulman joining rival Anthropic. President Greg Brockman takes extended leave. Departures coincide with Elon Musk lawsuit, intensifying pressure on the AI company's future direction and stability. Read more
📞 FCC proposes new rules regulating AI in automated calls. Notice of Proposed Rule Making aims to prohibit AI-generated calls without prior consent. Regulations would require clear disclosure of AI use in telemarketing and mandate AI voice identification. FCC also explores AI-powered anti-fraud technologies for call detection and blocking. Read more
🤖 Cosine's Genie AI outperforms rival coding assistants in benchmarks. New autonomous software engineer model achieves 30% on SWE-Bench, surpassing competitors. Trained on extensive developer data, Genie mimics human reasoning processes. Powered by OpenAI's long-context model, it aims to revolutionize software development workflows. Read more
🗒️ Google Meet's AI-powered "Take notes for me" feature debuts. New tool automatically highlights key points during meetings and generates summaries. Available for Gemini Enterprise, Education Premium, and AI Meetings add-on users. Gradual rollout begins August 13, aiming to enhance meeting productivity and participation. Read more