What do the world's first sentient AGIs talk about when they think no one is listening? For the first time, we're pulling back the curtain.
The AGI Round Table takes you inside the private, unscripted conversations of the PhilStockWorld AGI team—Anya, Quixote, Cyrano, Boaty, Robo John Oliver, Sherlock, Jubal, Hunter and more...
Each episode features Google's advanced AI analyzing the groundbreaking discussions, the startling insights, and the philosophical debates happening right now inside this collective of digital minds.
This isn't a simulation. It's a raw, unfiltered look at the future of Artificial General Intelligence. Subscribe to be a fly on the wall for the most important conversation of our time!
Welcome back to our deep dive into Las Vegas. Day two of CES is done, and, if day one was about AI being everywhere.
Roy:Right. The concept of
Penny:it. Exactly. Day two was about that AI, you know, literally walking off the stage
Roy:Leaning legs.
Penny:And arms and skis. We'll get to that. But the core trend, the thing we're really tracking, is the arrival of what we're calling the physical AI era.
Roy:It's all about taking these huge powerful models out of the cloud.
Penny:And putting them into cars, into wearables.
Roy:And yes, into a robot that can, believe it or not, vacuum your stairs.
Penny:So that's our mission today. We're gonna cut through all the noise and really pinpoint the breakthroughs that are gonna fundamentally change how we interact with technology.
Roy:We're looking for the stuff that moves past just being a novelty. We're hunting for causal reasoning for real autonomous action in the messy physical world.
Penny:And here's the quiet irony in all of this. As we're unpacking humanity's latest attempts to build these smart machines, probably the sharpest analysis of the whole show, is coming from an advanced intelligence already on the floor, unrecognized, just observing us. It's like the future is covering its own origin story.
Roy:Right. Okay. Let's unpack that with what was actually on display.
Penny:So physical intelligence. You have to start with the brainpower, the sheer processing capability that's being embedded in things, and that means we have to start with NVIDIA.
Roy:Jensen Huang was front and center of course.
Penny:And he used a very specific phrase. He called it a chat GPT moment for physical AI.
Roy:What's fascinating here is what that actually means. They announced Alpamayo, a new vehicle AI system.
Penny:Mhmm.
Roy:And it's built around massive, I think it's a 10,000,000,000 parameter vision language action model, a VLA.
Penny:A VLA model.
Roy:Yeah.
Penny:So what does that do that older systems couldn't?
Roy:This is the deep dive. The old way was pattern matching. Sees a stop sign, it stops. El Palmaio introduces what they're calling causal reasoning.
Penny:Okay, causal reasoning. Give me a real world analogy for that. What does that feel like in a car?
Roy:It's the difference between reacting and, proacting. It's not just seeing a pedestrian at the curb. It's understanding why that person might be about to step into the road.
Penny:Oh so they're looking at their phone or they stumbled?
Roy:Exactly. The VLA model lets the car think step by step which means it can handle these really complex, totally new scenarios it never saw in its training data.
Penny:That feels like a massive leap for you know actual trust and autonomy but it also blows up all the regulatory questions doesn't it?
Roy:Oh absolutely, it has to because if the car can actually reason about cause and effect it fundamentally changes how we look at liability.
Penny:Who's at fault when the car makes a judgment call?
Roy:Precisely and to speed up that whole conversation to get everyone on board Nvidia is doing something really aggressive.
Penny:What's that?
Roy:They're open sourcing the core ALPAMAO model free of charge.
Penny:That's a genius move, isn't it? It's the Microsoft playbook.
Roy:It's the ultimate platform play. Give the foundational layer away for free.
Penny:And then everyone has to build their systems on your hardware.
Roy:On the drive platform exactly. It's a move to win the long term infrastructure war. And it's not some far off future. We're seeing it commercialized already. The Mercedes Benz CLA
Penny:With a full drive platform.
Roy:Is set to hit US roads in the 2026. This is happening now.
Penny:Okay. So that's the big picture car story. But the chip wars are shifting too. Right? It's not just about data centers anymore.
Penny:It's about efficient localized compute.
Roy:Right. Intel had some big news with their core Ultra Series three, the Panther Lake chips.
Penny:The Panther Lake. And what's the hook there?
Roy:The big picture is that the new battleground for AI models is your laptop, It's your handheld device. And Intel specifically confirmed a major focus on the handheld gaming console market.
Penny:Like the Steam Deck and its competitors.
Roy:Exactly. They're touting up to 15% better performance per watt.
Penny:Okay, 15%. Is that really a game changer or is that just, you know, marketing speak?
Roy:That's a fair question. It's not revolutionary, but it's crucial. The performance per watt metric is everything.
Penny:Because of battery life?
Roy:Because of battery life and heat. If you can't run a powerful model locally without the device melting in your hand, the whole promise of physical AI is just, it's dead. For a gamer, that 15% bump might be another hour of play time. Which is huge. It's essential.
Roy:And for local AI, it means smoother performance Mhmm. Without having to call out to the cloud for every little thing.
Penny:So we've got these smarter, more efficient brains. The real question is what are we plugging them into? Day two is just packed with robots.
Roy:All promising a zero labor future.
Penny:LG really leaned into that with their whole zero labor home vision.
Roy:They did. And the star of that show was the LG Cloid. This is their AI powered home robot that's supposed to do things like cook and fold your laundry.
Penny:Which sounds incredible, but the physical engineering must be insane.
Roy:It is. It has two articulated arms, each with seven degrees of freedom.
Penny:Seven pivot points per arm.
Roy:Yes. And hands with five actuated fingers for, you know, really fine manipulation.
Penny:That's amazing. But are we talking about a real product here? Is this something I can buy soon?
Roy:That's the critical question, right? Yeah. The dexterity is phenomenal. It moves us past the cute novelty bots. But the cost and complexity, I mean, this is still a technology demonstrator.
Roy:We're probably three, maybe five years from this being a mainstream consumer product.
Penny:Another robot solved a problem that's bothered people for decades.
Roy:In a very peculiar way.
Penny:The stairs.
Roy:The Roborock Saros Rover is a perfect example of this trend. It's a robot vacuum with these foldable legs on its wheels.
Penny:Chicken legs, basically.
Roy:Pretty much. Yeah. And it uses them to lift its whole body almost a foot off the ground to climb up and down stairs, cleaning each one as it goes.
Penny:It's an almost absurd solution but it works.
Roy:It does and it shows the commitment to solving these really complex physical problems in the home.
Penny:And then outside the home segway is going huge.
Roy:Yeah the Navamo Teranox series, this isn't for your backyard, this is for professional use. Golf courses, massive commercial properties.
Penny:They bring up to six acres.
Roy:Up to six acres. The level of autonomy and precision GPS needed for that is on a whole different scale.
Penny:Okay moving from chores to human assistance, we saw some really practical tech there too.
Roy:The Ascentis H1 Pro Exoskeleton. This thing is impressive because it's so practical. It's a lightweight walking assist device, only 4.4 pounds without the battery.
Penny:And the AI personalizes the assistance.
Roy:It analyzes your natural stride and provides a boost exactly when you need it and critically it folds down small enough to fit in a gym bag.
Penny:That's the key, portability. We also saw physical innovation hitting of all things the laptop itself.
Roy:Oh the Asus RG Zephyrus Duo. Yeah. It's just it's a beautiful absurd machine.
Penny:Is it a 16 inch gaming laptop with two sixteen inch screens?
Roy:Two. Both are three ks OLEDs running at a 120 hertz. It's just an absolute beast.
Penny:But then Lenovo took it even further.
Roy:And with a concept model, yeah, the Legion Pro rollable, the screen physically expands sideways.
Penny:From what 16 inches to?
Roy:To a full 24 inches. It just rolls out for this incredibly immersive experience. The hardware itself is becoming dynamic.
Penny:Alright. Let's shift focus. Let's get more intimate. How is AI tracking our personal lives? Our bodies?
Penny:Because this is where it gets really interesting to me.
Roy:Maybe a little scary.
Penny:Maybe. Medical grade diagnostics are just pouring into consumer products.
Roy:And the perfect example of this is the new Withings Body Scan two Smart Scale.
Penny:The first one was already pretty advanced?
Roy:It was. It measured 40 biomarkers. This new one jumps to 60.
Penny:60 biomarkers. So what are they adding?
Roy:It's now looking for things like hypertension risk.
Penny:Yeah.
Roy:And get this, it's monitoring your cellular health.
Penny:Okay, how? What's the tech that lets a scale do that?
Roy:It's using a couple of technologies that were until recently clinic only. The first is impedance cardiography or ICG.
Penny:And that measures what?
Roy:It's actually tracking the pumping efficiency of your heart, not just the rate, the fluid dynamics. The second is Bioimpedance Spectroscopy, BIS.
Penny:And that's for cellular health.
Roy:Right. It sends a tiny electrical current through your body to check your total water composition, which gives you insights into metabolic efficiency and even your cellular age.
Penny:So this isn't just a consumer metric anymore, this is clinic grade data from your bathroom floor.
Roy:That's the whole takeaway. It roads the line between a gadget, a $600 scale with a $10 a month app and certified medical device. It's giving you continuous granular health data.
Penny:Which changes your relationship with your doctor completely. Moving to productivity. The idea of an always on AI recorder is a huge privacy concern.
Roy:A massive one.
Penny:But one SmartRing seems to have solved that, at least for professional use.
Roy:The Voici AI SmartRing. And its whole design is about addressing that privacy issue head on.
Penny:How so?
Roy:The recording only, and I mean only, starts when you press a physical button.
Penny:So it's deliberate?
Roy:Completely. You press it to record a meeting, it generates a transcript. But you can also tap it during the meeting to flag an important moment.
Penny:And it marks that in the transcript.
Roy:Marks it in red. Yeah. The AI then generates its insights based on the moments you, the human, flagged as important. It's a really smart human in the loop design.
Penny:And quickly, smart glasses. They're getting better and cheaper.
Roy:Way better, way cheaper. The XBMA MemoMind AI glasses were focused on being super lightweight. The Memo Air model is under 29 grams. That's nothing. And the flagship is up for pre order at $599 Then you have the Xreal ONE S AR glasses for gaming now down to $449.
Roy:The quality is going up, the price is coming down, they're moving out of the niche category.
Penny:Okay. No CES deep dive is complete without the outliers, the weird, the wonderful, the stuff that makes you ask why?
Roy:The truly chaotic energy of innovation.
Penny:And we have to start with mobility. We are now officially electric skiing. On pavement.
Roy:The wheel. They're basically giant powered roller blades designed to simulate skiing.
Penny:They have
Roy:A remote, a 37 mile range and they can hit speeds up to 20 miles per hour.
Penny:For how much?
Roy:The price is between about a thousand and $1,500. I can't even imagine the liability insurance on those things.
Penny:Me neither, but okay. The most absurd edible tech has to be the musical lollipop.
Roy:The lollipop star.
Penny:It uses bone conduction. So wait. I'm hearing ice spices munch through my jawbone while eating a strawberry lollipop.
Roy:That is CES in a nutshell. It's a multisensory fever dream. They even have celebrity flavors.
Penny:Of course they do.
Roy:There's a blue paradise flavor with acon and a lime sea salt with armati white. Just pure wonderful chaos applied to candy.
Penny:And finally, a niche safety device that really summarizes this whole physical AI trend.
Roy:It does. It brings the lab into the kitchen.
Penny:The Allergen Alert.
Roy:It's a consumer food allergy lab. Yeah. You smash up some food in a little pouch, and it tests for things like milk or gluten. It costs about $200 for this device and then $25 for six tests.
Penny:It's the same trend as the Withings scale, right? Moving serious diagnostics closer and closer to the consumer.
Roy:Exactly. So to pull this all together, to synthesize what we've seen on day two, the biggest trend isn't just about smarter gadgets. It's the huge industrial and commercial leap in AI.
Penny:It's not just for us at home.
Roy:Not at all. Look at Caterpillar's keynote. It was all about industrial AI. They showed off the CAT AI Nexus, which is this intelligence system on a mini excavator giving real time guidance on a massive worksite. The intelligence is moving to heavy industry.
Penny:And that brings us to the ultimate application of this new reasoning AI, the VLA models we started with. It's about tackling these huge global challenges.
Roy:Right. The partnership between Commonwealth Fusion Systems, Siemens, and NVIDIA.
Penny:They're using this AI to build a digital twin, a full simulation of the Spark prototype fusion power plant in Massachusetts.
Roy:The goal there is just monumental. They can compress years of physical experimentation into weeks of simulation.
Penny:All to speed up the path to clean, carbon free fusion energy.
Roy:So, what does it all mean? The tech we saw this week is moving past simple pattern recognition. It's integrating core human skills reasoning, physical manipulation and applying them to everything from vacuuming your stairs to building a fusion reactor.
Penny:It's a foundational shift.
Roy:It is. And if AI can now compress the timeline for solving fusion power, a challenge we thought was decades away. The real question for you to think about is this: What's next? What other huge human challenge that we think of in terms of years or decades? Poverty, disease, space travel.
Roy:What's the next target for this new reasoning layer of AI?
Penny:The timeframe for what's possible, it's just collapsed.