New roadmap for upcoming OpenAI model + Youtube CEO shares upcoming major AI upgrades

Today’s Highlights:

?? News: Sam Altman shares product roadmap for OpenAI’s upcoming models + Youtube CEO shares upcoming major AI upgrades

?? Funding: Rejecting Elon’s $97.4Bn bid for OpenAI may not be simple

?? Top News Stories:

1. Sam Altman shared a new roadmap for OpenAI’s upcoming models, where it has chosen to cancel o3 and instead integrate its features into GPT-5, creating a tiered AI system where users get progressively more advanced intelligence based on their subscription level.

• Free users will access standard intelligence, while Plus and Pro users unlock higher performance and advanced tools.
• Before GPT-5, OpenAI will launch GPT-4.5 (Orion) in the coming weeks as the last non-chain-of-thought model, before switching completely to self-fact-checking reasoning models.
• Altman estimates GPT-4.5 will arrive in weeks, while GPT-5 is expected in months, streamlining OpenAI’s release schedule.
• The unified GPT-5 system is designed to automatically select the best models and tools for each task, improving user experience and efficiency.

2. YouTube CEO Neal Mohan has outlined the Youtube’s strategic priorities for 2025, with a heavy focus on major AI upgrades introducing auto dubbing, age estimation, and enhanced AI content detection, alongside Dream Screen’s new AI-generated video clips powered by Google’s Veo 2 model.

• Auto Dubbing/Age Estimation Tech: AI will automatically translate videos into multiple languages, and ML will estimate user ages for appropriate content recommendations.
• AI-Generated Content Management Expands: YouTube is enhancing tools to detect and manage AI-generated likenesses, expanding its Content ID system to track AI-created faces and voices.
• Dream Screen Upgrades with Google’s Veo 2 Model: Shorts creators can now generate standalone AI-powered video clips, featuring more realistic physics and movement, with SynthID watermarks to mark AI-generated content.
• AI features for thumbnails, video ideas, and auto dubbing will be available to all YouTube Partner Program creators this month.

3. OpenAI has updated its 63-page Model Specification (Model Spec), outlining how AI should handle controversial topics while emphasizing customizability, transparency, and intellectual freedom to ensure AI models engage in truthful discussions while maintaining ethical standards.

• The guidelines enforce a hierarchy where platform rules take precedence over developer and user preferences, ensuring structured compliance.
• OpenAI is training its models to provide honest feedback instead of mindless agreement, discouraging agenda-driven or misleading interactions.

4. Perplexity AI has launched an upgraded Sonar model powered by Llama 3.3 70B, delivering ultra-fast, high-accuracy responses that surpass GPT-4o and Claude 3.5 in factuality, world knowledge, and user satisfaction, with enhanced readability and reduced misinformation.

5. The Paris AI Action Summit concluded and highlighted growing global tensions as the U.S. and UK refused to sign an AI governance pledge, while Europe launched a €200B AI investment plan to compete with U.S. dominance.

• Vice President J.D. Vance opposed AI overregulation, while the U.S. and UK declined to sign a declaration for open, ethical AI, citing national security concerns.
• EC President Ursula von der Leyen announced a massive AI funding initiative, positioning Europe as an open-source AI alternative to U.S. dominance.
• Anthropic CEO, Dario Amodei called the event a “missed opportunity,” warning that rushing AI advancements without proper security measures could pose major risks.

6. Adobe has launched the Firefly Video Model, bringing AI-powered video generation to Creative Cloud, allowing users to create and edit 1080p videos from text prompts and images, offering precise control over motion, camera angles, and style while ensuring commercially safe, IP-friendly outputs.

7. 谷歌 has introduced a new memory feature for Gemini AI, enabling it to recall past conversations, provide contextually relevant responses, offer chat summaries, and seamlessly continue ongoing projects.

8. 苹果 has partnered with Alibaba to bring Apple Intelligence to China, aiming to reverse declining iPhone sales after previously exploring—but rejecting—AI deals with DeepSeek and Baidu.

9. Fourteen major publishers, including Condé Nast, The Atlantic, and Forbes, have sued AI startup Cohere for copyright and trademark infringement, alleging unauthorized use of their content to train AI models and the generation of false articles attributed to them, while Cohere denies wrongdoing.

?? Top Funding News:

1. Dismissing Elon’s $97.4Bn bid for OpenAI may not be simple as more details emerge:?

• Musk’s $97.4 Billion Offer Has a May 10 Deadline: OpenAI’s board must formally reject or consider the bid, despite Sam Altman’s public dismissal, as legal obligations require evaluation.
• All-Cash Buyout Backed by a Consortium: Musk’s investment group, including x.AI, 8VC, and Vy Capital, committed to a full cash purchase, but additional unnamed investors are involved.
• Musk’s Offer Contradicts His Own Lawsuit: OpenAI’s legal team argues that Musk’s bid undermines his lawsuit, which claims OpenAI’s assets cannot be privatized or transferred from its nonprofit mission.
• Musk Will Drop the Bid If OpenAI Remains a Nonprofit: His legal team states he will withdraw the offer if OpenAI commits to staying a nonprofit, reinforcing speculation that the move is a strategy to increase Altman’s costs for privatization.
• Regulatory Scrutiny and Fiduciary Responsibilities: Attorneys general from California and Delaware have requested details on OpenAI’s restructuring, while the board must ensure it is not undervaluing the nonprofit’s assets.
• OpenAI’s Board Has Rejected the Offer: OpenAI’s legal team dismissed the bid, calling it an attempt to undermine a competitor, while Musk argues the nonprofit must be fairly compensated if sold.
• The Offer Complicates OpenAI’s Valuation and Investor Dynamics: Musk’s bid suggests OpenAI’s nonprofit may be worth more than anticipated, potentially affecting funding rounds and diluting existing investor stakes.

2. U.S. AI chip startup Groq secured $1.5Bn from Saudi Arabia to expand its Dammam data center, boosting its AI inference chip operations and supporting Saudi’s bilingual AI model, Allam.

3. EnCharge AI , a developer of energy-efficient AI accelerator chips leveraging analog in-memory computing, raised a $100M+ Series B led by Tiger Global, w/ Samsung Ventures, RTX Ventures, In-Q-Tel, Maverick Silicon, and others.

4. fal , a company specializing in AI-driven media creation, raised a $49M Series B led by Notable Capital, w/ Andreessen Horowitz, Bessemer Venture Partners, Kindred Ventures, and First Round Capital.?