OpenAI unveils o3 and o3 mini 🚀, Meta’s AI Watermarking Tool 🧩 , Google Whisk AI 💡

In partnership with

Secure Access to Digital Assets

• Bridging traditional finance & blockchain.

• C$152.4M YTD revenue.

• Global ETP leader.

Explore Now

💬 OpenAI Developments

OpenAI unveils o3 and o3 mini — here’s why these 'reasoning' models are a giant leap

OpenAI's new o3 and o3-mini models represent a significant leap in AI, focusing on enhanced reasoning, coding proficiency, and problem-solving. These models employ extended internal deliberation, enabling them to tackle complex tasks with improved accuracy and adaptability. Highlighted during the "12 Days of OpenAI" event, they are currently in safety testing, emphasizing ethical deployment. Positioned against competitors like Google’s Gemini 2.0, o3 sets itself apart with deliberate reasoning and coding excellence, promising to revolutionize software development, scientific problem-solving, and human-AI collaboration.

🚀 Tech Industry Moves

Google is introducing an AI Mode to its Search platform, enabling conversational answers through a chatbot-like interface, similar to its Gemini model. This new feature will appear as a tab alongside existing ones like Images and Videos, providing conversational responses with links to external websites and a follow-up question option. This move aligns with the trend of integrating AI-powered conversational tools, as seen in platforms like Reddit and OpenAI's ChatGPT. Google aims to enhance search efficiency and discovery with these advanced AI capabilities.

Meta has launched Video Seal, an open-source video watermarking tool designed to address challenges in verifying AI-generated content and managing digital rights. The tool uses a novel approach called temporal watermark propagation, enabling efficient watermarking by processing select video frames and propagating the watermark to neighboring frames. This reduces computational costs while maintaining robustness against manipulations like compression, cropping, and brightness adjustments. By open-sourcing the tool under the MIT license, including its codebase, models, and a demo, Meta aims to foster collaboration and advance the field of video content verification in response to the growing sophistication of AI-generated media.

Google has introduced Whisk, a new AI image generator that creates visuals using other images as prompts instead of relying solely on text. Users can provide multiple images to guide the subject, scene, and style or use a dice icon to generate AI-suggested images. Whisk also allows text prompts for refinement but emphasizes quick visual exploration rather than detailed edits. Powered by the latest Imagen 3 model, Whisk produces editable outputs and supports iterative adjustments. Alongside Whisk, Google unveiled Veo 2, an advanced video generation model with improved cinematic understanding, initially available on Google’s VideoFX and expanding to YouTube Shorts next year.

Meta has integrated AI features into its Ray-Ban smart glasses, enabling voice conversations with the Meta chatbot, real-time live translation for English, Spanish, French, and Italian, and Shazam integration to identify songs. The chatbot supports continuous conversations without requiring activation phrases, while live translation provides spoken or transcribed outputs. These features, part of firmware version 11, are available to early access users in the U.S. and Canada, with Shazam functionality already accessible to all users.

Google is expanding its Gemini AI’s in-depth research mode to 40 additional languages, allowing Google One AI premium users to access an AI-powered research assistant. This tool helps create research plans, finds relevant information, and generates reports through multiple search iterations. While Google focuses on using trustworthy sources and running native-language evaluations to improve accuracy, challenges remain in ensuring factual correctness and grammatical precision in languages like Hindi. The company also employs testing programs and local teams to improve model performance and quality.

⚖️ AI Ethics and Regulation

Mindgard, a British university spinoff, focuses on AI security, addressing risks like jailbreaks and prompt injections that can threaten businesses. Its Dynamic Application Security Testing for AI (DAST-AI) uses automated red teaming to identify vulnerabilities during runtime, particularly in neural networks. Led by Professor Peter Garraghan, Mindgard offers a SaaS platform for enterprises and AI startups, providing tools for AI risk prevention. After raising £3 million in seed funding, the company is now securing $8 million to expand its team and product, with plans to grow internationally, particularly in the U.S. Mindgard aims to ensure AI is used safely and securely.

💡 Emerging AI Technologies and Companies

Boon raises $20.5M to build agentic AI tools for fleets

Boon, a startup focused on AI tools for logistics, has raised $20.5 million to improve fleet operations by integrating data from multiple software systems. Its AI platform helps businesses optimize tasks like routing and staffing, boosting efficiency. The company already serves 35,000 drivers and 10,000 vehicles, with an annual revenue run rate of $1 million. Boon aims to simplify complex logistics workflows and expand its capabilities across the industry.

TuSimple pivot from self-driving to AI animation is complete with CreateAI rebrand

TuSimple has rebranded as CreateAI after pivoting from autonomous trucking to AI animation and gaming. The company is developing AI tools for video game and anime content, including the "Ruyi" image-to-video model. Despite controversy over moving assets to China and potential conflicts of interest, CreateAI plans to leverage its autonomous driving technology for its new venture. The company is also collaborating with anime designer Shōji Kawamori on a project based on "The Three-Body Problem."

🎨 AI in Creative Industries

Instagram is set to introduce generative AI tools for video creators, allowing them to edit appearances and backgrounds using text prompts. Powered by Meta’s Movie Gen AI model, the tools will let users change outfits, backgrounds, and other aspects of their videos. While previewed features, such as altering surroundings or transforming into animated versions, appear seamless, it's unclear if the final user tools will achieve the same results. Meta's announcement follows similar AI models from OpenAI and Adobe, with the tools expected to roll out next year.

🌍 Global AI Race

Reconciling tradition and innovation: Japan's path to global AI leadership

Japan's journey toward global AI leadership balances tradition with innovation, as it strives to integrate AI while respecting cultural values. Despite a cautious approach rooted in risk aversion and consensus-driven decision-making, initiatives like the GENIAC project by METI aim to foster AI adoption without compromising Japan's high ethical standards. Japanese companies, such as Hitachi and NEC, are combining traditional expertise with AI to enhance industries like manufacturing. The media also plays a crucial role in shifting societal attitudes toward risk and failure, supporting a culture that embraces innovation while maintaining trust and ethical rigor.

If you want more updates related to AI, subscribe to our Newsletter