Robert Sutor

Written answer in UK Parliament about AI and intellectual property. Brian Mathew on Department for Science, Innovation and Technology, at TheyWorkForYou https://lnkd.in/exJ4n3k3

1

🔥We filed a U.S. provisional patent for Agentic Enrichment Architecture (AEA). In midst of processing the same in Europe! 🔸AEA is a human-at-the-control enrichment layer that fuses cultural, linguistic, and brand context directly into your content ops. 🔸Why now? LLM-only workflows miss nuance, burn cycles, and inflate costs across markets. AEA cross pollinates w/ your stack—no rip-and-replace: Enrichment API • Locale Profiles • Policy/Style Overlay • Agent Handoffs • Readiness Signals. 🌊 Outcomes: native resonance, fewer review loops, brand integrity, lower risk—with transparency, auditability, and separation of duties built in. 🔸We’ve been solidifying and onboarding strategic partners and we are so proud enriching Legacy models to build future ready right now! 🏗️ Where others choose labels, we architect layers and levels across100+ Geo-markets! Granular Cultural #CX Vibrancy! NO MORE BOX ♒ performative outcomes AEA augments the tech you already have with enterprise security —and so much more. ✔️Check out #CreativeStudioX. 🔸Consult Manta → https://lnkd.in/g8gwvTZi #GlobalizeWe #AEA #AgenticAI #GWAgenticAIInteroperability #Localization #HumanAtTheControl #ChicagoTech #GEO #AEO #SEO #ContentStrategy #BrandAgents #1871Tech #ChicagoTech Betsy Ziegler Lawrence Johnson Krishna Mula GlobalizeWe

15

Open Quantum Design co-founder Roger Melko has been trail-blazing for years. He launched a Perimeter Institute gem - Perimeter’s Quantum Intelligence Lab (PIQuIL) https://sharU9mXdfuHqo7pkjm

50

Springer Nature book on AI ethics riddled with fake citations even as the publisher announces an OA reviews journal — a known magnet for AI slop. https://lnkd.in/e_hEYTuS

1

1 Comment

Raman quantum memory demonstrates near-unity performance #quantum H/T Cecile G Tamura Ever thought of trying to pause a beam of light without losing any of its information. https://lnkd.in/gH9WdTKe That’s essentially the challenge of *quantum memory* a core technology needed for future quantum networks and quantum computers. To be useful, a quantum memory has to store delicate quantum signals with extremely high efficiency and almost no distortion or noise. In practice, though, boosting efficiency often introduces noise, and reducing noise tends to lower efficiency. This trade-off has held back progress for years. A new study reports a breakthrough that finally cracks this problem. Researchers developed a clever way to “reshape” how light gets absorbed and stored in a cloud of warm rubidium atoms. Using a mathematical technique called a *Hankel transform*, they map incoming light into a special kind of spin pattern called a *spinwave* inside the atoms. They then apply an intelligent, adaptive strategy to compact this spinwave, so it stores the information more cleanly and efficiently. https://lnkd.in/gsX_XnDQ The result is a quantum memory that performs far better than previous attempts: * Efficiency of 94.6% meaning almost all the incoming quantum signal is successfully stored. * Extremely low noise, just 0.026 photons per pulse a vital improvement because even tiny noise can ruin quantum information. * Very high fidelity (98.91%), meaning the retrieved quantum state is almost identical to what went in. (That needs to improve - Assume compounding growth with future signal stacking) And it does all this with *broadband, nanosecond-scale signals*, which are essential for high-speed quantum communication. Joe P. In simple terms: this work shows how to store quantum information both *fast* and *accurately*, without the usual compromises. This achievement sets a new practical benchmark for quantum memory and opens the door for faster quantum networks, more precise quantum control, and eventually more scalable quantum computers. #QuantumComputing #quantumphysics #QuantumInformation #physics #RamanScattering #Photonics

16

3 Comments

The relevance of this paper spans a large number of aspects when discussing generative AI, LLMs and the pursuit of consistency and determinism in LLM inference. One area of considerable interest (to me) is reducing nondeterminism in LLM inference when using small language models (which I regret to have started calling 'SLIMs'). What the Thinking Machines Lab paper spotlights is likely the basis for a curious aspect of SLIMS: the past two years of research which reports on how extreme quantization (the basis for SLIMs) intersects with deterministic inference requirements and paradoxically delivers better determinism than conventional LLMs which are orders of magnitude larger. For those of you whom I have not yet cornered and assailed with my line of patter about SLIMS: Small Language Inference Models (SLIMs) deliver AI inference based on various approaches to radical compression. Among the most interesting, the so-called 1-bit model reduces the matrices used in AI to just three values (-1, 0, +1), delivering 6× faster performance while demonstrating the quality and performance of conventional (quantized) LLMs which are orders of magnitude larger and require silicon based on GPUs.  SLIMs compress AI models by as much as 75% yet paradoxically improve reliability—a breakthrough which is beginning to reshape enterprise deployment. These systems promise to deliver mainframe AI power on the class of chips found in today's smartphones. Microsecond decisions for semiconductor fabs. Real-time control for surgical robots. The counterintuitive finding that SLIMs can provide better determinism in those scenarios that need it most - assembly lines, patient monitors, field equipment, power grids and telecomms - challenges conventional assumptions about precision and reproducibility. Get in touch if you want to know more about our on-going watch on small footprint ai. #1bitLLM #SLIMs #EdgeAI #AI #GENAI #AgenticAI

9

Good or bad for us non-billionaires? "White House Launches ‘Genesis Mission’ to Accelerate AI-Enabled Scientific DiscoveryDecember 9, 2025The White House recently issued an executive order launching the “Genesis Mission,” a US Department of Energy–led national initiative designed to accelerate scientific discovery using advanced artificial intelligence. The Mission seeks to unify high-performance computing, secure cloud-based AI environments, and federal scientific datasets into a centralized American Science and Security Platform, which will serve as the basis for developing scientific foundation models, AI agents, and automated research and manufacturing systems across priority technology domains" .Read the full LawFlash on our website > https://lnkd.in/eV5PCZ6r

1

🔬 The quantum computing reality check we may need Reading Josh Luckenbaugh's piece in National Defense Magazine about DARPA's Quantum Benchmarking Initiative reminded me why the quantum revolution seems to be taking longer than the headlines suggest. The fundamental challenge? Error rates. As Juliette Peyronnet from Alice & Bob explains, "we can't really trust the results that we get because there's a really high percentage of errors." Every quantum computation is fighting against decoherence and noise. Most approaches require massive overhead for error correction. Google needs about 1,000 physical qubits to create one reliable logical qubit. That's like needing 1,000 regular computers to make one that works properly. This is why DARPA's approach seems so smart. Rather than betting on one technology, they're stress-testing 18 different quantum computing concepts through their QBI program. Companies like Photonic Inc. are exploring distributed quantum networks that could sidestep scaling constraints entirely. The agency isn't "driving the bus" this time. They're letting innovators propose solutions, then rigorously validating who's "for real and who's not." Given DARPA's track record (internet, GPS, stealth technology), this methodical approach could be what finally delivers practical quantum advantage. Click and follow @thomas-pike-spectrumshield today #QuantumComputing #TechReality #DARPA #Innovation #Engineering

5