Trump Admin Eyes “1:1 Chip Rule”

Xicoia Debuts: The First AI Talent Studio to Create Emotionally Intelligent Digital Stars

Particle6 founder and CEO Eline van der Velden used her keynote at the Zurich Film Festival to unveil Xicoia, a new AI talent studio focused on building hyperreal digital stars with personalities, narratives, and emotional depth. Rather than static avatars, Xicoia’s AI personas are meant to perform, engage, and evolve across platforms as creative IP.

Key Points:

1. Avatar engine + storytelling = AI talent with character
   Xicoia is built on Particle6’s proprietary “DeepFame” technology. The AI personas aren’t just superficial visuals — they come with backstories, voices, evolving arcs, and emotional range, allowing them to hold unscripted conversations and respond in real time across platforms.

2. Cross-media presence & monetization focus
   The studio aims to deploy these characters in film, TV, podcasts, social media (TikTok, YouTube, etc.), brand campaigns, and gaming. Xicoia also plans to monetize through IP licensing, fan engagement, and merchandising.

3. Blend of original and legacy personas
   The first AI persona, Tilly Norwood, has been soft-launched on TikTok, Instagram, and YouTube. Beyond wholly synthetic characters, Xicoia will also partner with celebrities and estates to create AI versions of real or legacy figures—recreating not just likenesses but voices, tone, and personality.

Conclusion
Particle6’s launch of Xicoia pushes the boundary of AI in entertainment: from static digital avatars toward “living” synthetic personalities that can act, respond, and evolve. If successfully executed, this could shift how audiences interact with virtual creators and open new models for monetizing AI-driven IP. That said, challenges remain in authenticity, ethics, audience acceptance, and ensuring these AI personas don’t drift into uncanny or misleading territory.

RL Legend Richard Sutton Calls LLMs “Dead Ends,” Proposes OaK as True AGI Blueprint

While much of the AI world races toward GPT-7 or GPT-8, Richard Sutton — one of the founding pillars of reinforcement learning and newly a Turing Award winner — is sounding a warning: “scale” is not the missing piece. In a conversation with Dwarkesh Patel, Sutton asserts that LLMs, as powerful as they are, are structurally incapable of capturing true intelligence. He proposes a new architecture, OaK, aiming for agents that learn through action, surprise, and goals — not static text prediction.

Key Points:

1. LLMs lack goals, surprise, and consequence learning
   They predict language distributions, not outcomes of actions. They can’t adjust based on real-world feedback.

2. Scaling is not the answer
   Bigger models amplify the same limitations. Sutton argues we need architectures that learn at runtime, not only at training time.

3. OaK: Sketch of a new architecture
   An agentic loop that generates subgoals, learns options, builds models, plans, and self-updates — all grounded in real-world interaction.

Conclusion

Sutton’s critique is a powerful counterpoint to the “larger is always better” momentum sweeping AI. Whether or not OaK becomes the next paradigm, the interview reignites crucial questions: Are we chasing scale while ignoring agency? Can learning from scratch ever beat learning with purpose? And perhaps most consequentially: what happens when AI systems begin choosing their own objectives?

New U.S. Policy Would Force Chipmakers to Match Imports with Domestic Production — or Face Tariffs

The Trump administration is reportedly preparing to roll out a sweeping semiconductor policy that ties import volume to domestic manufacturing: for every chip imported into the U.S., a matching amount would need to be made on U.S. soil. Failure to meet this “1:1” ratio could trigger tariffs possibly as steep as 100 %. The move is aimed at reducing dependency on overseas chip supply chains and reviving American chipmaking capacity.

Key Points:

1. Matching rule + tariff penalty
   Reportedly, chipmakers would need to produce an equivalent volume of semiconductors domestically to the ones their customers import. Falling short could incur punitive tariffs — reports suggest up to 100 % import duties.

2. Incentives and grace periods
   To ease into compliance, there’s talk of credit systems: companies pledging U.S. capacity might receive temporary relief or credits, allowing some leeway while new facilities are built.

3. Industry & supply chain shock risk
   The proposal would reshape global semiconductor supply chains. Companies that depend on specialized chips abroad may struggle to match with domestic equivalents. It’s facing skepticism about technical feasibility, enforcement, and cost burdens.

Conclusion

If implemented, this 1:1 import-to-production rule could dramatically recalibrate the semiconductor industry. It would clearly favor firms already investing in U.S. manufacturing, and put intense pressure on chipmakers globally to realign their supply chains. However, the plan also courts serious practical and legal challenges — from defining “equivalent” chips, to enforcing compliance, to navigating WTO and trade law constraints.