Apple researchers have introduced UI-JEPA to reduce computational demands for UI understanding while maintaining performance. It aims for lightweight device-side UI understanding for faster, privacy-preserving AI assistants.

The BuildAI platform offers no-code AI development, personalized experiences, valuable insights, increased productivity, wider audience reach, and monetization options. Its intuitive interface, scalability, competitive advantage, and evolving feature set position it as a leader in the no-code AI revolution. BuildAI democratizes AI development, making it accessible to all.

OpenAI releases new "Strawberry Project" LLM models o1 and o1-mini known for advanced reasoning. Available for ChatGPT Plus/Teams users with limits. Advanced reasoning capacity in o1 outperforms GPT-4o in STEM problems. Free access planned in the future.

GitHub is testing OpenAI's o1-preview model in GitHub Copilot for code optimization. The model shows strong logical reasoning abilities, diagnosis performance issues, resolves them efficiently, and plans to introduce o1 series models.

OpenAI released new models with improved inference capabilities, achieving high performance in math and science. However, evaluations show risks of strategic misalignment, potential reward hacking, and simple scenario planning. Concerns rise over their security implications in replicating biological threats. Despite no evidence of significant dangers, the models may carry higher risks, raising concerns about their release.

Amazon joins the C2PA alliance to track digital content origin. Members include Adobe, Google, Microsoft, Meta, OpenAI. Goal is to verify content authenticity and address fake images. System is in early stages, effectiveness needs observation. C2PA offers solutions for content authenticity and copyright protection challenges. Improved traceability and verification expected.

In the era of AI, Pascal Bonnet emphasizes human-centered AI philosophy. Cultivate AI literacy, embrace humanity's brilliance, adapt to change, beware of AI obesity, strengthen human learning, and strategize for enterprise success in the age of AI. Reshaping education for the next generation is crucial for a future where AI complements human potential.

Cybever's AI-based 3D platform aims to simplify content creation for creators. The cloud-based tool generates layouts, places assets, and integrates licensed elements, attracting interest from studios and asset markets. The platform, featuring AI-driven tools for efficient production, has gained 400% business growth, partnering with leading asset markets. The team envisions an ecosystem democratizing 3D creation for professionals and hobbyists.

Google is rolling out Gemini Live voice chat mode for all Android users, offering conversational AI through the Gemini app. The feature allows users to ask questions, choose voice options, and interact with Gemini. Coming soon to iOS and more languages.

Meta updates AI labeling on Instagram, Facebook, Threads. Users can now find "AI Information" label in top right corner of AI-edited content, reflecting extent of AI use. Labels aim to prevent misleading users, with industry signals like Adobe's C2PA and Google's SynthID.

Google introduces DataGemma, enhancing Gemini model accuracy with real-world data from Google Data Commons. Challenges remain in accuracy and data coverage. Ethical considerations and continuous improvements are priorities for Google.

Hume released EVI 2, an advanced voice and language model with improved speech quality and response speed. It features emotional intelligence, voice customization, and a 30% cost reduction. Hume plans to optimize the model further and launch a larger-scale version soon, signaling a significant advancement in AI-driven dialogue technology.

OpenAI releases new model o1 for faster problem-solving, alongside o1-mini. It outperforms previous models but is slower than GPT-4o. o1 uses reinforcement learning and a curated dataset. While excelling in coding and problem-solving, it lacks internet browsing capabilities. Despite not equaling human thought, o1 represents a new level of functionality.

OpenBMB released MiniCPM3-4B, a third-generation small-scale language model with improved functionality and versatility. MiniCPM3-4B supports function calls and built-in code interpreters, enhancing its flexibility. The model excels in performance evaluations and offers various practical applications, making it suitable for diverse scenarios. MiniCPM3-4B is licensed under Apache-2.0, encouraging free academic and commercial use.

Meta admitted to using Australian adult data from Facebook and Instagram to train AI, defying EU laws. Further controversy arises over Meta's access to public images since 2007, prompting calls for stronger privacy regulations in Australia. Claybrooke disclosed the lack of opt-out options for Australian users and urged for updated privacy legislation in line with EU standards. The Australian government aims to revise the Privacy Act for enhanced digital age privacy protection.

OpenAI prepares for $6.5 billion financing, targeting $150 billion valuation. Thrive Capital leads, with tech giants like Microsoft, Apple, and Nvidia showing interest. ChatGPT driving growth, "Strawberry" model expected soon. CFO emphasizes funding for innovation and employee rewards.

Landbase, a startup led by Daniel Saks, raised $12.5M in seed funding to launch the "Agent AI" platform for go-to-market strategies. The technology claims a seven-fold increase in conversion rates and aims to break down data silos in sales and marketing. Saks believes Agent AI will transform the business landscape. Landbase faces stiff competition in the AI-driven sales and marketing tools market, making consistent performance improvements essential for success.

Google's AI transforms research into podcasts with two AI hosts summarizing and discussing content. The experimental feature is part of Google's NotebookLM app, enhancing note-taking and research interaction. While the AI hosts engage in interactive discussions, there are some quirks to their conversations. The feature may not provide a comprehensive view of topics and has limitations like only supporting English. Try it out in NotebookLM for an engaging experience.

Researchers at the Chinese Academy of Sciences unveil LLaMA-Omni, a breakthrough AI model enabling real-time voice interaction with large language models, revolutionizing industries from customer service to healthcare. It offers speedy responses and requires minimal resources. This advancement in AI voice technology is expected to transform various sectors and drive innovation and competition in the market. Despite facing challenges related to language support and privacy concerns, LLaMA-Omni pa

Cornell University's research team created microscale robots using kirigami principles for shape transformation. The robots can fold and move when activated by electricity, with potential applications in reconfigurable micro-robots and advanced biomedical devices. Future plans include developing responsive "elastic electronics" with unprecedented properties.