OpenAI Announces o3, GPT-5 Behind Schedule

⚡️ Headlines

🤖 AI

AI Is Marc Benioff's Friend and Foe - Salesforce CEO Marc Benioff navigates the dual challenges and opportunities presented by AI in the tech industry. [The Information].

Backflip Releases AI Model That Turns Text Into Physical Reality, Backed By $30M from NEA and Andreessen Horowitz - Startup Backflip unveils an AI model capable of transforming textual descriptions into tangible objects, securing $30 million in funding from prominent venture capital firms. [Business Wire].

OpenAI Has Edge Over Google in Winning Publishers' Business - OpenAI secures partnerships with major publishers, gaining a competitive advantage over Google in the AI content arena. [The Information].

🦾 Emerging Tech

Trump Appoints Sriram Krishnan as AI Czar; Backlash from MAGA - President-elect Donald Trump appoints Chennai-born venture capitalist Sriram Krishnan as the White House advisor on AI policy, sparking debates among MAGA radicals. [Times of India].

NHTSA Advances Autonomous Vehicle Regulatory Framework - The National Highway Traffic Safety Administration (NHTSA) takes significant steps toward establishing a regulatory framework for autonomous vehicles in the U.S. [The Verge].

🤳 Social Media

MrBeast Plans 100-Hour YouTube Video at Great Pyramids of Giza - YouTuber MrBeast announces an ambitious project to film a 100-hour video at the Great Pyramids of Giza. [Newsweek].

Instagram Previews AI Video Generation Tools - Instagram teases upcoming AI-powered video editing features that promise to revolutionize user-generated content. [Social Media Today].

Bluesky Can Be a Solid Social Media Site for Sports Lovers - Bluesky emerges as a promising platform for sports enthusiasts seeking an alternative social media experience. [Bloomberg].

🔬 Research

APpaREnTLy THiS iS hoW yoU JaIlBreAk AI - Researchers demonstrate methods to bypass AI safety measures, highlighting vulnerabilities in current AI systems. [404 Media].

⚖ Legal

Google's Remedies Proposal in DOJ's Search Distribution Case - Google proposes adjustments to its search distribution practices in response to the Department of Justice's antitrust case. [Google Blog].

Albania Bans TikTok for a Year After Killing of Teenager - Albania announces a one-year ban on TikTok following the tragic death of a teenager, citing concerns over social media's influence on youth violence. [Reuters].

🎱 Random

Texas Instruments Finalizes CHIPS Act Award, Gets $1.6 Billion - Texas Instruments secures $1.6 billion in funding through the CHIPS Act to bolster semiconductor manufacturing. [Bloomberg].

Britannica Embraces Artificial Intelligence - Encyclopaedia Britannica integrates AI to enhance its content creation and curation processes. [The New York Times].

🔌 Plug-Into-This

OpenAI Announces New o3 Model

OpenAI unveiled the o3 and o3-mini models, aimed at improving AI reasoning and problem-solving. The models are in testing, with o3-mini releasing in January 2025, followed by the full o3 version.

• Advanced Reasoning: o3 excels in tasks requiring logical, step-by-step problem-solving.

• Improved Accuracy: Scores three times higher than prior models on reasoning benchmarks.

• Testing Phase: OpenAI is accepting applications for external testers through January 10, 2025.

• Naming Decision: The 'o2' name was skipped to avoid trademark conflicts.

• Safety Innovations: Features like deliberative alignment enhance model reliability.

🧠 o3 positions OpenAI as a leader in AI reasoning, which increasingly seems to be the space they are most easily able to dominate, despite recent challenges from Google’s Flash model.

OpenAI's Next Big AI Effort, GPT-5, Is Behind Schedule and Crazy Expensive

OpenAI's development of GPT-5, code-named Orion, is experiencing significant delays and escalating costs, casting uncertainty over its completion and effectiveness.

• Extended Development Timeline: Initiated over 18 months ago, GPT-5 was anticipated by mid-2024; however, unforeseen challenges have postponed its release.

• Resource Constraints: The project faces potential data shortages, raising concerns about the feasibility of achieving the desired intelligence levels.

• Financial Implications: The escalating expenses associated with GPT-5's development are straining OpenAI's budget, prompting evaluations of the project's sustainability.

• Strategic Partnerships: Microsoft's expectation of the model's mid-2024 arrival underscores the project's significance to OpenAI's key stakeholders.

• Technological Ambitions: GPT-5 aims to represent a substantial leap in AI capabilities, building upon the foundation established by ChatGPT.

⏱️ As competitors are quickly closing the gap on OpenAI’s lead in multiple areas, the pressure on shipping new models quickly has only increased. We’ve seen the discourse on “diminishing returns” coming for AI soon vacillate back and forth this year…perhaps 2025 will serve as a reality check for the high flying industry.

Google Plans to Add 'AI Mode' Option to Search

Google is developing an 'AI Mode' for its search engine, allowing users to access AI-generated responses through a dedicated tab alongside existing search categories.

• Dedicated AI Tab: The 'AI Mode' will appear alongside tabs like 'All,' 'Images,' and 'Videos,' providing a separate interface for AI-generated answers.

• Conversational Interface: This mode will offer a chat-like experience, similar to Google's Gemini AI chatbot, enhancing user interaction with search results.

• Source Integration: AI-generated responses will include links to relevant webpages, ensuring users can access original content for more information.

• Follow-up Queries: A search bar within the AI Mode will prompt users to ask follow-up questions, facilitating deeper exploration of topics.

• Voice Command Support: Future updates may enable users to interact with AI Mode using voice commands, expanding accessibility options.

🔎 Quite a practical move from Google. The backlash to the Gemini summaries being added to all search earlier this year was significant. And truly there’s a role for both forms of search (index of links vs. prosaic summaries) in the modern world.

 🆕 Updates

Meta has updated its Ray-Ban Meta smart glasses to include real-time AI video processing and live language translation.

Why Care?

• Enhanced User Experience: Real-time AI video allows users to receive immediate information about their surroundings, enhancing situational awareness.

• Seamless Communication: Live translation supports English, Spanish, French, and Italian, facilitating conversations across language barriers.

• Integrated Features: The update includes Shazam integration for music identification, expanding the glasses' functionality.

• Hands-Free Convenience: Voice-activated commands and AI assistance enable users to interact with technology without needing to use their hands.

• Competitive Edge: These advancements position Meta's smart glasses as a leading option in the wearable tech market.

Tencent's HunyuanVideo model has introduced an open-source, uncensored AI video generation tool that hobbyists can run locally on hardware like a 24 GB VRAM GPU. It offers flexible fine-tuning capabilities and the ability to handle uncensored outputs, making it accessible for creative, unfiltered video projects.

Why Care?

• Accessibility: Free distribution and local hardware support enable widespread experimentation.

• Customization: Fine-tuning and LoRA compatibility allow users to tailor the model for specific tasks.

• Creative Potential: Uncensored capabilities and realistic rendering broaden artistic and experimental applications.

• Cost-Efficiency: Unlike many models requiring cloud services, it reduces ongoing costs for users with suitable hardware.

• Trend Indicator: Reflects a shift toward open AI tools empowering individual creators.

📽️ Daily Demo

It looks like AI content tools are working their way towards the “all-in-one” solution we’ve been hoping for. These two YouTube demos dive deeper into the capabilities offered under the hood of the two tools referenced in this X clickbait post.

🗣️ Discourse

Veo 2 is still dominating the chatter on X even now over a week since it’s launch, and general AI hyping has shifted squarely into the promise of agents to automate workflows.