Good morning. It’s Monday, December 30th.

Did you know: I sent 7,671,861 emails in 2024? Thanks for reading along.

In today’s email:

• OpenAI confirms non-profit plans

• DARPA tries to predict future

• Meta envisions AI social networks

• 5 New AI Tools

• Latest AI Research Papers

You read. We listen. Let us know what you think by replying to this email.

OpenAI Confirms Plans to Become a For-Profit Company as It Looks to Raise Even More Investor Money

OpenAI has formalized plans to discard its layered nonprofit-controlled structure in favor of a streamlined for-profit model by 2025. The reorganization, aimed at wooing capital beyond its historic $6.6 billion raise, will transform OpenAI into a Delaware public benefit corporation—allowing traditional equity while retaining a mission to serve the public good. By shedding structural friction, OpenAI seeks to align investor incentives, bankroll its insatiable compute needs, and position itself as a cornerstone of the AGI economy.

OpenAI’s restructuring into a Public Benefit Corporation (PBC) has sparked skepticism from its former executives. Miles Brundage flagged unresolved governance ambiguities, cautioning against the nonprofit arm granting excessive latitude to the PBC, potentially eroding the organization's foundational ethos. Likewise, Jan Leike, former safety lead now at Anthropic, dismissed the move as uninspired, arguing that OpenAI’s lofty AGI mission is being swapped for modest philanthropic efforts in healthcare and education.

While the PBC structure ostensibly balances financial imperatives with social impact, concerns persist that profit-driven motives could encroach on the long-term alignment and equitable distribution of AGI benefits. Read more.

Related Story: Nonprofit group joins Elon Musk's effort to block OpenAI's for-profit transition

DARPA Seeks Algorithmic 'Theory of Mind' to Predict, Manipulate Future Behaviors

DARPA is rolling out its "Theory of Mind" initiative, aiming to develop an algorithmic framework that models adversaries' decision-making processes. By blending human expertise with machine learning, the goal is to craft simulations that predict and influence future behavior with greater precision and efficiency.

This follows DARPA’s historical pursuit of predictive technologies, from the Total Information Awareness project to LifeLog, both of which aimed at manipulating human behavior through data. The implications of this program are profound, potentially extending from military strategy to broader applications in law enforcement and corporate decision-making. If successful, it could reshape the way we think about security, control, and prediction in both public and private spheres. Read more.

Meta Envisions Social Networks Where AI Characters Coexist Alongside Human Accounts

Meta is setting the stage for AI-generated profiles to mingle with human accounts on Facebook and Instagram. These digital personas will come equipped with bios, avatars, and content creation capabilities—designed to boost engagement. Since mid-2024, U.S. users have been experimenting with AI character creation tools, though these profiles remain largely private for now.

The move aims to inject social media with more dynamic interactions, but it also raises concerns. Misinformation could spread more easily, and the quality of content may plummet as AI-generated posts flood the platforms. While Meta requires labeling of AI-generated content, enforcing it remains tricky. Moreover, integrating AI content with ads could challenge Meta’s revenue model, especially if brands hesitate over bot engagement. Read more.

• Google currently has no plans for a test-time premium subscription similar to ChatGPT Pro

• ChatGPT Changelog for 2024 Reveals Year of Unrelenting Features and Enhancements

• OpenAI Mobilizes D.C. to Push AI Infrastructure Agenda

• Nvidia's next move: powering humanoid robots

• Deepseek's V3 is the latest example of state-controlled censorship in Chinese LLMs

• Researchers develop first AI system for analyzing medical imagery in both English and Arabic

📄 HuatuoGPT-o1, Towards Medical Complex Reasoning with LLMs

📄 Molar: Multimodal LLMs with Collaborative Filtering Alignment for Enhanced Sequential Recommendation ​​

📄 MMFactory: A Universal Solution Search Engine for Vision-Language Tasks 

📄 Orient Anything: Learning Robust Object Orientation Estimation from Rendering 3D Models 

📄 The Superposition of Diffusion Models Using the Itô Density Estimator

Thank you for reading today’s edition.

Your feedback is valuable. Respond to this email and tell us how you think we could add more value to this newsletter.

Interested in reaching smart readers like you? To become an AI Breakfast sponsor, reply to this email or DM us on X!