Timothy Chou

The science is clear. When an ad fits the moment, the brain stays in the story and remembers the brand. Inside the article: a simple model for attention quality, the design rules that reduce load, and the metrics that price attention. Highlights: ->Context and continuity produce longer attentive seconds ->Embedded units can more than double attention at a lower price per second ->AQ thinking strengthens both outcomes and renewals Read the article below. If you want our attention, quality checklist, and examples, connect with me.

3

Richard Sutton Says LLMs Are a Dead End. He's Right! I was listening to Dwarkesh Podcast interview with Richard Sutton. His point in the podcast? LLMs just imitate what people say. They don't pursue goals, interact with the world, and learn from experience. For true general intelligence, you need exactly that - experience, reward signals, and world models. Not just predicting the next word. Sutton is absolutely right about the limitations. But from all my research what I see is that the systems being built around LLMs are already solving real problems. No, this isn't AGI. Yes, it's useful. Suttons assessment: Today's models don't learn on the job. In production - you wouldn’t let a model rewrite itself during day-to-day use. Teams retrain models offline and roll out updates under change control. LLMs aren't goal-seeking agents. Out of the box, an LLM predicts text. It doesn't pick actions to maximize reward in its environment. But just because something isn't AGI doesn't mean it can't transform how we work. Especially in healthcare – I don’t need AGI – I need technology that reduces pajama time, improves patient care and management and if at all possible gets rid of the ridiculous back and forth of insurance issues. The vendors that really understand these technologies aren't hand-waving. They're shipping real capabilities: Memory you can govern. OpenAI, Anthropic, and Google have added persistent memory outside the model - context across sessions without risky on-the-fly changes. Retrieval instead of guessing. RAG lets systems look things up in live knowledge bases and cite sources. In healthcare, this means pulling from current clinical guidelines instead of hallucinating. Planning and tool use. There are techniques that push models to plan multi-step work, call tools, and check results. That's not "learning," but it does delivers better outcomes. The product is now LLM + memory + retrieval + workflow engine. What May Get Us Beyond "LLM Only" The path forward from the existing players seems to be to pair LLMs with components that act and learn: - World model agents "imagine" futures before acting - Robotics systems transfer web knowledge into physical control   - Enterprise agents in Office tools where most white-collar work lives Together, they add experience, control, and feedback loops around the LLM core. My thoughts: Sutton's warning is healthy. But the market isn't waiting for AGI. The pattern is always the same in business and healthcare: we don't wait for perfect. We deploy what works, measure rigorously, and iterate. By wrapping LLMs with memory, retrieval, and multi-step control - and gradually adding experience-based components - I am seeing that groups are building systems that help real people do real work today. If you are waiting for the magic of AGI (however you define it) you may be too late. https://lnkd.in/eaUr48ZC

6

In #musictech, we decompose audio into spectrograms and lyrics at scale. At Moodagent, we trained models to learn and tag tracks with mood, genre, instrument and even emotional scores - like distinguishing cool jazz from modal jazz. Subtle to the ear, but clear in the data. Now apply that to #healthcare. Patient communication - voice and text - is just another high-dimensional signal. Using #AI (NLP techniques and audio models), we can surface topics patients actually care about. That’s exactly what Stanford did by analyzing 600k+ cancer patient messages. Same core idea: structured insight from noisy, unstructured input. Mature in music. Underused in health. Let’s close that gap.

7

Sierra C. of Barclo is building at the frontier of quantum, AI, and deep tech. In this past episode of The Jason Cavness Experience. Sierra shares her journey from TEDx production and anti-trafficking advocacy to founding a venture studio focused on future technologies and systems innovation. We also talk about the realities of deep tech startups, venture culture in Seattle, and the big questions around AI and quantum computing. Full conversation is on YouTube at The Jason Cavness Experience if you want the long version. Link is in the comments. Topics include: • The future of quantum and AI • Building a venture studio • Challenges facing deep tech founders • Big Tech’s influence on innovation • AI ethics and energy demands #podcast #deeptech #artificialintelligence #quantumcomputing #venturecapital

7

1 Comment

What happens when a father and son set out to teach generative AI across the globe? We found out on this week’s Sidecar Sync podcast, where Mallory Mejias, AAiP and I had the pleasure of sitting down with Finn and Conor Grennan, the first father-son duo we’ve had on the show. Conor is NYU Stern School of Business's Chief AI Architect, advising top-tier organizations like NASA, Google, and PwC on AI adoption. Finn? He’s just finished his sophomore year of high school…and is already shaping his school’s AI strategy and leading workshops in Nepal. Yes, Nepal. What stood out? - AI feels like magic. Whether you’re a nonprofit leader in the U.S. or an educator in Nepal, the awe, curiosity, and fear are shared across borders. - Peer-to-peer learning matters. When Finn, a teenager, led sessions on AI for educators and NGOs, the response was powerful. Why? Because when the messenger feels relatable, the message lands deeper. - Education is ripe for disruption. For centuries, the cycle has been teach → memorize → test. GenAI challenges this, making it harder to tell if it’s the student or the AI being tested, and that forces us to rethink what learning truly means. - The AI adoption blueprint is surprisingly universal. Whether you're in pharma, finance, or a remote village, it's less about prompts and tools, and more about starting with what you actually do. From there, AI becomes your second brain. I’m inspired by Finn’s clarity, Conor’s humility, and their shared mission to democratize AI education. For #associations, there’s a key takeaway: build guardrails, not roadblocks. Empower your members to experiment safely. Encourage bottom-up innovation. Catch the full episode on Sidecar Sync. It’s heartwarming, eye-opening, and a blueprint for how associations can lead in the AI era. Link in comments. Timestamps: 0:00 Introduction 2:09 Backgrounds in AI: From NYU Stern to Sophomore Year 8:52 How Schools Handle AI Guidelines 15:55 The Nepal Trip: Why They Went and What They Did 19:32 Structuring the AI Workshops for Teachers and NGOs 23:06 Cross-Cultural AI Challenges and Common Concerns 28:30 Addressing Critical Thinking and Student Cheating Fears 35:02 Peer-to-Peer Influence and Student-Led AI Learning 42:45 Democratizing AI Education 48:45 Final Thoughts #AIforGood #GenerativeAI #AIInEducation #Nonprofits #DigitalTransformation #Leadership #SidecarSync #AssociationInnovation #AIAdoption #FutureOfWork

21

6 Comments

Deeply thoughtful conversation with Zico Kolter, board member at OpenAI and head of the machine learning department at Carnegie Mellon University, about AI safety, AI security, agents and frontier AI Episode available on Spotify, Apple Podcasts and below on YouTube: 00:00 Intro 01:32 OpenAI board role and Safety & Security Committee 03:53 How OpenAI reviews major model releases 05:33 OpenAI’s preparedness framework explained 09:46 Are frontier AI models getting safer? 12:33 Why AI safety does not come from scale 15:23 The four categories of AI risk 19:38 Doomerism vs accelerationism in AI 24:11 The six-month AI pause debate 26:20 AI safety as a global effort 28:04 How Zico Kolter got into machine learning 31:05 OpenAI in the early days 34:14 Why Carnegie Mellon became an AI powerhouse 38:43 What Gray Swan does in AI security 40:44 AI safety vs AI security 43:15 The GCG jailbreak paper 49:19 How AI labs responded to jailbreak research 50:19 State-of-the-art AI defenses 52:32 State-of-the-art AI attacks 54:22 Why AI agents expand the attack surface 58:39 Are AI agents ready for production? 59:40 Mechanistic interpretability explained 1:02:31 Will AI be safer in two years? 1:03:46 Reinforcement learning and self-improving models 1:08:09 Do post-transformer architectures matter 1:09:29 Best research directions in AI now 1:11:00 Zico Kolter’s Intro to Modern AI course 1:14:53 Why modern AI is simpler than people think https://lnkd.in/ecxD8puf

25

3 Comments

Incredible Israeli Innovation Newly developed AI by teams at NVIDIA Israel and the Weizmann Institute of Science can identify a person’s risk of developing type 2 diabetes up to 12 years before diagnosis, using as little as one to two weeks of glucose data. Published in Nature, this breakthrough could transform prevention, enabling earlier, more precise interventions and reshaping the future of healthcare. Read more here: https://lnkd.in/eFGrYEjH #standwithIsrael

38

3 Comments

Excellent Tech Crunch interview with former iRobot CEO, Colin Angle, about how the FTC caused the iRobot bankruptcy. It's ver sobering, and explains how US regulators have effectively shut down innovation in the US. https://lnkd.in/eKcNGQ2M

27

Foundation, an SF based humanoid robot company, is offering to weaponize their robots for U.S. defense purposes and are currently lobbying in Washington DC. https://lnkd.in/gYtJvp6w Humanoids — robots in human form — are starting to be used at factories nationwide; Morgan Stanley estimates there will be more than 1 billion in operation globally by 2030. Investors have poured a record $2.7 billion into humanoid robotics so far this year, according to PitchBook. More than half of the top robotics firms are based in China. 

25

1 Comment

Incredible day for Spore.Bio who are launching Spore Labs today ❤️ Following Demis Hassabis and John Jumper's Nobel Prize in Chemistry for AlphaFold, Google.org launched the AI for Science Fund, a multi-million dollar initiative to support organizations using AI to solve the most complex scientific challenges of our time. Spore.Bio was selected as the sole start-up recipient. Not just validation - a mandate. Google.org is betting on Spore.Bio’s vision: that AI can fundamentally change how we detect, understand, and fight microbial threats. That the convergence of photonics, hardware engineering, and artificial intelligence can unlock answers we couldn't even ask before. With this funding, Spore.Bio is building Spore.Labs an AI-native research powerhouse that will develop open datasets, will publish breakthrough findings, and collaborate with academic and clinical partners worldwide. The microbial challenges of the 21st century demand 21st-century solutions. 💡 The team is killing it and they’re hiring 🚀 cc: LocalGlobe Amine Raji Maxime Mistretta, PhD. Mohamed Tazi Huge thank you to Dorothy Chou 🤩 for your partnership on AI in Science

30

1 Comment

🚨 New #RFP Alert: Fueling the Next Quantum Breakthroughs Big news out of Chicago - P33 Chicago and the National Quantum Algorithm Center (NQAC) just launched the Grand Challenges Program, awarding two $125,000 grants to advance real-world quantum algorithm research through collaboration between universities, quantum companies, and industry partners. Here are 5 things you should know: 1️⃣ $125K proof-of-concept awards to fund breakthrough quantum research. 2️⃣ Each proposal must include an Illinois university, a quantum company, and a non-quantum industry partner. 3️⃣ Applications are open now through Dec. 10, 2025. 4️⃣ This builds on major momentum at "Quantum Shore” following PsiQuantum’s groundbreaking. 5️⃣ Projects will translate quantum discoveries into real-world impact across finance, materials, and energy. 📅 Q&A: Nov. 10, 10:30 AM CT 📧 Submit by Dec. 10 → NQAC@IQMP.org Learn more 👉 https://lnkd.in/gUda39Pr

12

Jensen Huang told Carnegie Mellon's graduating class something worth repeating for every organization deploying AI right now. AI automates tasks but elevates workers. A radiologist does not just read scans. They care for patients. AI handles the task. The purpose remains human. Most organizations are asking which jobs AI will replace. The better question is which human capabilities get amplified when the routine work falls away. Read more here: https://lnkd.in/eEiGxMsc #AI #EnterpriseAI #HumanAI #ITLeadership #MidMarket

6

1 Comment

AI success that scales comes from systems design. Specifically it’s how systems learn and how they develop expertise over time. As we’ve built our AI Operating System (AIOS), we’ve made deliberate architectural choices that don’t show up in demos, but compound into learning advantages. It helps us scale over firms that have built individual agents. Three aspects of what Abdul and I found out over the past two years: 1) Memory as the foundation of learning. Most AI systems treat context as disposable. We treat it as cumulative. Our architecture centers on persistent, evolving memory that enables continuous learning across interactions, personalization at the OS level and compounding context that adapts performance over time. Without memory, there is no real learning and the system needs to repeat many steps. 2) Model layers as pathways to expertise. We’ve moved away from single agents on monolithic models toward modular, layered systems: - Routing to specialized models, fine tuned to develop domain-specific expertise - Orchestration layers that route tasks to the right “expert”, human or AI - Continuous evaluation and learning capabilities that upgrade without breaking the system Expertise emerges not from one agent doing everything, but from systems that know which intelligence to apply, when. 3) Multiplayer AI and reinforcement learning from experts. The next frontier is smarter learning loops. We’re building toward a multiplayer paradigm where: - Networks of experts (human + AI) provide reinforcement signals - Systems learn from real-world judgment, not just static data - Expertise is encoded, refined, and scaled across the network Most teams are still optimizing for benchmarks and surface-level capabilities. But durable advantage in AI will come from systems that learn continuously, architectures that compound expertise, and feedback loops that improve with scale The companies that win will build systems that get smarter, more specialized, and more valuable with every interaction.

9

3 Comments

I keep hearing about the "hype bubble" around AI. If there is a bubble its a man-made bubble. The bubble is caused by technologist types who are enamored by technical achievements of new models that wow and amaze us rather than actually looking for how to apply these advances in ways that ACTUALLY improve business or life. The irony is that in many cases the simplest uses of AI with most basic models add the most value. These patterns can be implemented and integrated in minutes but almost seem too trivial to many so instead we have complex patterns that never deliver on value and AI gets looked at as hype. If you think there is a hype bubble or you think AI is game change for your business and your productivity, you are probably right.

43

This critical technology for the UK and US is only one of the material #quantum transactions you will see DCVC lead as we continue our *two decade* track record of the broadest and most impactful support for quantum computing, sensing, and security companies 💪

14

VERSES is playing a key role in the collaboration and growth of Active Inference researchers and practitioners around the world.

19

In the latest episode of the OpenAI podcast I got talk to OpenAI CFO Sarah Friar and legend himself Vinod Khosla about the AI industry. Sarah explained the relation between compute and revenue (OpenAI's revenue keeps 3x'ing each year) and Vinod provided an insightful way to tell if we're in a bubble: API calls. Are they going up or down? (Spoiler: Up, up, up!) https://lnkd.in/g2tTQBdp

24

1 Comment