The AI Daily Brief: Artificial Intelligence News and Analysis - Why the Future of AI Has a Body

Episode Date: October 10, 2025

This episode explores why the next era of artificial intelligence will take physical form. It examines the rise of humanoid robots, breakthroughs in embodied AI, and how advances in robotics and action models are closing the gap between digital intelligence and the real world. The discussion also highlights how rapid progress in sensors, movement, and large action models is bringing general-purpose robots closer to everyday use.

Today on the AI Daily Brief, why the future of AI has a body. Before that in the headlines, evidence that using AI at work could be reducing burnout. The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. All right, friends, quick announcements before we dive in. Thank you to our sponsors, Insightwise, KPMG, robots and pencils and notion. To get an ad-free version of the show, go to patreon.com slash AI Daily Brief, or you can subscribe on Apple Podcasts. If you were looking for information about sponsoring the show, speaking, engage, or if you are a kind soul who wants to give me feedback on some ideas for AIDB extensions for

2026, head on over to AIDailyBreef.A.I. And with that, let's dive in. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in around five minutes. We kick off today with a really encouraging report. Something that people have been talking about for a long time is the idea that AI, if done well, should be taking tasks off the plate of humans that are rot, monotonous, difficult, basically the stuff that's really important that has to get done, but that's a total pain to get done. And in so doing, if that actually happens, theoretically work should be a better, more enjoyable experience, right? According to a new survey from HR management platform UKG, that might just be the case. In a sample of 8200 frontline workers,

the percentage of workers who were not using AI who reported burnout was 54%. Among those who were using AI at work, that number dropped to 41%. That means that the burnout rate among those using AI was a full 24% lower than those who weren't using AI. Now, interestingly, in the same study, while about a third of frontline workers reported using AI at work, fears of AI replacement were more widespread than actual use. Two-thirds said they were concerned that AI might replace their job, a quarter said that part of their job had probably already been replaced by AI, and over the next five years,

20% said it was likely that their job would be entirely replaced by AI. At the same time, the survey also found a widespread fear that failing to adopt AI would cause workers to fall behind. 65% of respondents said that they were worried that colleagues more skilled in using AI could take their job. So this is a super interesting report. It is not by an outright AI company. It's got a lot of folks in the sample.

And it's very clearly telling a complex story. There is a lot of fear and concern around AI. There's definitely indication that people are behind in understanding how to use. it, but it's got that really positive glimmer for those who are using it. Said Corey Spencer, Vice President of AI at UKG, the irony, if done the right way, is that AI can empower people to do what they were meant to do. The global study shows that work needs to be done to better educate, train, and explain the why behind AI uses on the front line.

It's about AI and frontline employees working together to move from menial to meaningful work. When AI is deployed with a people-first focus, it doesn't feel like you're using technology. It feels like you're solving problems. That, I think, is interesting context for our next story, which is that Google has introduced a new AI for the workplace product in Gemini Enterprise. On Thursday, Google launched the new product, which represents a fresh start for the way they deliver AI services to corporate clients. Over the past year, Google has, of course, embedded AI tools into Google Workspace,

offering numerous services attached to Google Cloud, but generally dealing with the fact that AI was sort of all over their product sprawled rather than in a clear concentrated bucket. Gemini Enterprise aims to consolidate the disparate products in one. place, with CEO Sunder Pichai branding it as the new front door for Google AI in your workplace. Rather than being an add-on for Google workspace, the product is designed to be an all-in-one AI bundle. On top of the familiar Gemini Assistant, it features a pre-built suite of agents as well as a no-code

agent builder. Google has also improved agent connectivity across ecosystems and tools for this rollout. For the first time, Google's agents will be able to tap into tools like code assist and deep research, and they're also promoting improvements to their agent's ability to access corporate data both within Google workspace and across common platforms including Microsoft 365, Salesforce and SAP. The product also includes a central governance framework to enable agent monitoring, security, and auditing. Google wrote, by bringing all of these components together through a single interface, Gemini Enterprise transforms how teams work. It moves beyond simple tasks

to automate entire workflows and drive smarter business outcomes. This is part of a style of announcement that's happening a lot right now, where the development is not just about a new, more powerful underlying model, but a broader product context that makes it easier to use and actually be functional in real-world deployment. Pichai showcased a powerful but simple example where the user prompted Gemini to, quote, build me an agent that helps me prepare for meetings with customers by analyzing relevant docs, emails, and past meetings. Gemini then created a workflow that taps into those sources across Google Calendar, Gmail, and Google Drive. A very simple use case, but a nudge towards a natural language-prompted agent builder that feels much more accessible than some of the

other UI patterns that we've seen recently. In another demo, the platform was used to create a Halloween marketing campaign. The agent performed research, identified key trends, and checked inventory. It was able to identify a product shortage and rectify it by tapping it as service now. Afterward, the agent drafted emails to store managers about the incoming order, and then finally created materials for social media using Google's image and video generation tools. Generalized interoperability is a huge part of the idea. Box CEO Aaron Levy tweeted, another great win for AI agent interoperability. Box is partnering with Google to make Box AI agents accessible

via Google Gemini Enterprise, so you can easily work with your enterprise content from anywhere. This is what the future of AI looks like. Now, Google shared some adoption statistics alongside the announcement, and based on those, enterprises are ready to take the next step. 65% of Google Cloud customers are now using AI products, and among the AI industry, adoption is even higher. Google boasted that nine of the top 10 AI labs

and almost every AI unicorn is using Google Cloud. Next up, a little bit of fundraising news. Turns out that OpenAI's new agent kit has very much not killed N8N, as that startup has closed a Series C funding round that sees their valuation jump to $2.5 billion. They last raised money in March at a $350 million dollar valuation, achieving a 7X in seven months. The round was led by Excel and also includes Nvidia's venture investment arm.

Among other things, the deal speaks to the red hot demand for top-tier AI startups, previous reporting as the deal was coming together in August, spoke of a bidding war leading to the round. It's also validation that agent building platforms are set to be a major pillar of the AI ecosystem. Sources said that N8N financials showed that they had reached 40 million in ARR and achieved 10x user growth over the past year. Speaking directly to the OpenAI competition, N8N CEO Jan Oberhouser commented, If OpenAI is releasing something, you're going to be locked into the OpenAI model. What makes us special is that you don't have this lock-in. One more big fundraising news, Reflection AI has raised a whopping $2 billion to build a U.S.-based

open source frontier model. The startup was founded in March of last year by a pair of Google DeepMind leaders, Misha Laskin, who led reward modeling for Gemini, and Yuanis Antonaglu, who co-created AlphaGo. They've since recruited around 60 leading researchers and engineers across various specializations. Earlier this year, they released a large-scale reinforcement learning platform capable of training frontier models built on state-of-art mixture of experts' architecture. They said that this was, quote, something once thought possible only inside the world's top labs. Having applied this method of reinforcement learning to build highly-performant coding agents, they want to bring their methods

to general agentic reasoning. This new round, which values the company at $8 billion, allowed reflection to secure a training cluster and begin training a frontier model of their own. Now, a huge part of the pitch is the need for a competitive open source frontier model built in the U.S. Co-founder Laskin said, Deep Seeking Quinn and all these models are our wake-up call because if we don't do anything about it, then effectively the global standard of intelligence will be built by someone else.

It won't be built by America. So you either choose to live at a competitive disadvantage or rise to the occasion. White House AIsar, David Sachs, pointed out that this is a national priority for the government. Retweeting the announcement and adding, It's great to see more American open source AI models. A meaningful segment of the global market

will prefer the cost, customizability, and control that open source offers. We want the U.S. to win this category too. I'm reminded of a Nicki Minaj lyric, 50K for a verse, no album out, or in this case, as Swix put it, $2 billion raised with no product. The team is beyond cracked. That's going to do for the headlines next up, the main episode. As a consultant, responding to proposals can often feel like playing tennis against a wall. You're serving against yourself trying to guess what the client really wants. That all changes with the Insight Wise

proposals platform. Now you've got an AI coach that thinks just like your client. It returns to the brief time and time again, identifying opportunities, showcasing your track record, and making recommendations to improve your pitch. Suddenly you're on center court, but this time you've got a secret weapon. Insight Wise does away with all the time-consuming manual work so you can focus on winning more business more often. Generate reports, pull insights from your own data, build competitive advantage and go to sleep before 2 a.m. When it comes to proposals, you only get one shot. With Insightwise, make yours an ace. Go to Insightwise.A.I. to start your free trial today. What if AI wasn't just a buzzword, but a business imperative? On You Can with AI, we take you

inside the boardrooms and strategy sessions of the world's most forward-thinking enterprises. Hosted by me, Nathaniel Wittamore, and powered by KPMG, the seven-part series delivers real-world insights from leaders who are scaling AI with purpose. From aligning culture and leadership to building trust, data readiness, and deploying AI agents. Whether you're a C-suite executive, strategist, or innovator, this podcast is your front row seat to the future of Enterprise AI. So go check it out at www.kpmG.org.us slash AI podcasts or search you can with AI on Spotify, Apple Podcasts, or wherever you get your podcasts. Today's episode is brought to you by robots and pencils. When competitive advantage lasts

mere moments, speed to value wins the AI race. While big consultancies bury progress under layers of process, robots and pencils builds impact at AI speed. They partner with clients to enhance human potential through AI, modernizing apps, strengthening data pipelines, and accelerating cloud transformation. With AWS certified teams across U.S., Canada, Europe, and Latin America, clients get local expertise and global scale. And with a laser focus on real outcomes, their solutions help organizers work smarter and

serve customers better. They're your nimble, high-service alternative to big integrators. Turn your AI vision into value fast. Stay ahead with a partner built for progress. Partner with Robots and Pencils at Robots and Pencils.com slash AI Daily Brief. Chatbots are great, but they can only take you so far. I've recently been testing Notion's new AI agents, and they are a very different type of experience. These are agents that actually complete entire workflows for you in your style, and best of all,

they work in a channel that you already know and love because they are purpose-built Notion super users. Notion's new AI agents completely expands the range of what Notion can do. It can now build documents from your entire company's knowledge base, organize scattered information into organized reports, basically do tasks that used to take days and get them complete in minutes. These agents don't just help with work, they finish it. Getting started with building on Notion is easier than ever.

Notion agents are now your very own super user to help you onboard in minutes. Your AI teammates are ready to work. Try Notion AI for free at the link in our show notes. Welcome back to the AI Daily Brief. Covering AI is a constant balance of figuring out how much to cover things that matter here and now versus things that will matter in the future. And my bias towards the here and now has meant that I haven't really spent much time on physical AI, embodied AI, aka robotics. And yet, in many contexts, the clear trajectory and future of artificial intelligence is in physical form. It is embodied.

This week we saw the release of the Figure O3 Humanoid Robot, and I thought this was a good of chances any to do a bit of a primer on robotics and embodied AI, everything from the latest technology updates to the state of deployments in the U.S., to the broader global competition, to what technical limitations there still are and how it intersects with artificial intelligence more broadly. Now, coming to the news that inspired this post this week, Figure is one of the most well-established robotic startups in the U.S. You might have seen videos of their robots sorting packages. Earlier versions of their robots are currently deployed in a pilot at the BMW plant in South Carolina. Now, without being an expert in this space, it's very hard to know where the functional

limitations of these seemingly advanced humanoid robots begins an end, but what's clear is that we've been on a rapidly improving trajectory. At the beginning of the year, figure laid out a plan to ship 100,000 humanoid robots over the next four years, splitting their focus between commercial and home use. And that brings us to the new figure 03, the third iteration of figures flagship humanoid robot. So what was improved with this edition? A lot of the improvements are about core functionality and getting the robot ready for widespread use. It now features inductive charging, removing the need to engineer a way to replace battery packs or plug into an outlet. It also now features washable soft fabric covering all pinch points, providing a layer of safety for human interaction

as well as a little more usability. Generally speaking, the design has been re-engineered with mass production in mind. On the technical side, the robot has an improved audio system for voice reasoning, allowing it to hold conversations with people. It has also had a complete overhaul of its sensors and hand design, with the redesign built around Helix, which is Figures' proprietary vision-language action AI model that drives the robot. You might remember last year when Figure actually moved off of OpenAI's models and basically said that embodied AI needed its own custom models and so they were going to keep that in-house. Now, the interaction between mechanical hands, a robot sensors, and the AI system driving a humanoid has been one of the trickiest problems in robotics for decades.

Walking and running are difficult, but one of the Holy Grails in robotics has long been figuring out how to train a robot to grip a fragile object with enough force to avoid dropping it, but not so much force the object is crushed. Humans are able to do this by feeling how fragile an object is and carefully calibrating their grip strength. But pressure sensors generally haven't been sensitive enough for those fine calibrations. In some systems, engineers have tried vision as a substitute. For example, a robot might see that the object they're trying to pick up as an egg

and dial the grip strength way down because of that, but that doesn't necessarily translate well to a generalized model that can be applied to unknown objects when they're encountered for the first time. Figures upgraded hands and AI model seem to have gotten a lot farther on that particular problem. Chris Paxton, a researcher at Agility Robotics posted, Very impressive dexterity from Figure O3. Custom in-house tactile sensor design makes sense. Part of the reason we've seen such a proliferation in open gripper and tactile sensor tech

is because publicly available solutions aren't there yet. Figure has also circulated some impressive videos of the new robot folding clothes, as well as stacking items in a dishwasher without dropping items all over the place. Now, very clearly, part of the vision here is that this will be a very broad general-purpose robot, suitable for a wide range of tasks in the home as well as being versatile and industrial settings. In fact, New Atlas asked, is this the Model T of robots? And that certainly seems to be the design brief here, developing the first U.S. designed humanoid for mass production and broad deployment.

Kern Bashar, the CIO of Brilliant Advice posted, figure O3 is a step towards robots that can live and work alongside humans in everyday places, homes, offices, and factories. Linus Ekinstam wrote, Nobody took this seriously. I've had many conversations this week where most people think this is 10 to 20 years out. They don't get it. But figure O3 with the helix control system is happening right now.

Today. Welcome to the future. Now, while figure O3 was definitely the biggest robotics news, there was another story that came up as I was prepping this. SoftBank announced a multi-billion-dollar deal to acquire the indefinitely. industrial robotics division from Swiss conglomerate ABB. ABB have been planning to spin off the division into a new entity, but instead sold it to SoftB for $5.4 billion.

Now, an important difference here, ABB does not make humanoids. Instead, they exclusively factor robot arms for use and production lines. The arms are currently in service across a range of different applications, including some very fine-grained work in electronics manufacturing. SoftBank CEO Masayoshi San said in a statement, SoftBank's next frontier is physical AI. Together with ABB Robotics, we will unite world-class technology and talent under our shared vision to fuse artificial superintelligence and robotics, driving a groundbreaking evolution

that will propel humanity forward. In other words, even though ABB right now is focused on these industrial robotic arms, it seems pretty clear that SoftBank has a bigger, embodied AI vision behind this acquisition. In fact, robots are central to the vision of the AI future for a lot of tech leaders. Last October, Elon Musk hosted the unveiling event for the long-awaited Tesla cybercabs. However, a swarm of Optimist robots that were catering the event stole the show, serving drinks to guests and dancing on stage. Now, it was later revealed that much of the robot's actions were actually guided by human teleoperators behind the scenes, which made the entire event feel like a little bit of smoke and mirrors, all designed to make the Optimus model seem more advanced than it actually was.

And since then, for Tesla, at least, progress has seemed to slow. In fact, this week we got news that Tesla had scaled back plans to produce thousands of units this year after running into design problems with, you guessed it, the robot's hands. Sources told TechSpot that there's a warehouse full of partially assembled robots missing their forearms and hands. Still, there's clearly still a lot of work being done. Last week, Elon Musk posted a video of Optimist learning kung fu and looking pretty agile on its feet. Several other U.S. companies are getting pretty close with their own humanoid robots.

Alongside their robot dogs and numerous other models, Boston Dynamics have been iterating on a humanoid called Atlas for around 12 years. This week, they unveiled their most dexterous robot hand design yet. The hand only has two fingers and an opposable thumb, but they believe it could outperform humans thanks to a new set of pressure sensors. Another player in the U.S. humanoid market is Apptronic, a small Texas startup. They took investment from Google as a strategic partner in February as part of their Series A, and since then have provided the hardware to test and develop Google's increasingly sophisticated AI robotics model.

In a podcast back in May, CEO Jeff Cardenas remarked on just how fast the space is expanded. It's just amazing to me to hear that there's 100 companies working on humanoid robots. Investors that I talked to two years ago said that humanoids don't make sense. They didn't want to pay attention to hardware. Now they have a humanoid thesis and hardware is the name of the game. Cardenas noted that the first country to develop efficient and effective humanoids will have a huge advantage in manufacturing, going so far as to call robotics the new space race.

Even without fully functional humanoids yet, robots are quickly infiltrating the manufacturing space. A report from the International Federation of Robotics found that over 500,000 robots had been deployed globally for the fourth year in a row in 2024, with the U.S. representing over 35,000 of those units. A report from Grandview Research found that automotive electronics and metals and heavy machinery are the major sectors using robotics and manufacturing. Interestingly, as robotics advanced, the share of robots being deployed in electronics is going up as they become more capable of the fine motor skills required for the job. Outside of manufacturing, Amazon has been a massive driver of

in warehousing and logistics. They currently have a range of different arms in production for picking and sorting, as well as various robotic forklifts and other stock-moving robots. Amazon alone currently has 750,000 robots deployed across their logistics network. Now, interestingly, growth in robotics deployments in the U.S. has generally been slow and steady, averaging around 5% per year since 2018. That growth, however, has been very dependent on the economy. There was actually a 5% contraction in robot installs in the U.S. in 2023 due to poor economic conditions in manufacturing and autos. Total installs fell to 38,000 that year. Generally, there's a sense that while robots have found a solid niche in industrial settings, they certainly haven't hit

an inflection point in capabilities that forces firms to install them as fast as possible. Which, of course, brings us to embodied AI. The idea is simple, pairing a version of the type of AI models that power LLMs with robot bodies. Now, this field has been coming along in leaps and bounds over recent years. Nvidia has recently been putting a ton of resources into working on AI models that can drive robots, which they call large action models. The idea is similar to multimodal LLMs. However, the inputs are based on the robot sensors and the outputs are actions the robots can take, rather than words on the screen or other generated media. Google are also working on these action models, but have taken it a step further. They're applying reinforcement learning to what they're calling embodied reasoning models.

Separate to work on models that drive robots, there are also world models that are being designed to act as virtual training environments for robots. The idea is that training with a sufficiently accurate model of real-world physics should be mostly indistinguishable from training in the real world. The benefit is, of course, that you can run the simulations and reinforcement learning entirely in software. So a large action model can carry out millions of trials per minute, rather than a single one if it were bound to a robot body. And yet, as much progress has been made, there is still a fair way to go. At the moment, these action models aren't anywhere as close to being as capable as their LLM counterparts. Data about text is wildly abundant compared

to data about physical movement. Now, some are trying to put together troves of video data in an attempt to train foundation action models, but others suspect the training set will need to include telemetry data from robot sensors instead. That data is even more relatively scarce than videos of movement in the physical world, so it likely involves projects of dedicated data creation. This is one of the reasons that people are so excited about the possibilities of world models, which could theoretically be a way to generate enough synthetic data for very large training sets very quickly. For some time now, robots have been very good when trained on specific tasks, but very primitive when they come across edge cases or tasks they haven't seen before.

More than ever, however, the intersection with advanced AI is seen as a path to much more generalized physical intelligence. Now, one last note before we get out of here, any primer on the state of robotics cannot ignore China, who seem to be leaps and bounds ahead of the U.S. From a pure scale perspective, China is the premier robotics-enabled manufacturing hub. They have millions of robots in operation already and are scaling up faster than the rest of the world combined. In 2023, they installed 276,000 industrial robots compared to 38,000 in the U.S. You might have also heard of dark factories where robots work in darkness without a human in the building. There are also an endless stream of videos out of China showing how advanced

their humanoids are. Although at this point, it's very difficult to get a good read on how real these videos are. The polish of a demo video is extremely easy to fake with teleoperators or simply a defined routine. When we get footage of Chinese humanoids completing an unscripted task, they don't seem to be light years ahead. Still, humanoids are one of the big focuses for investment dollars in China, and you can see that investment showing up. A few months ago, Chubby posted a robot traffic cop in Shanghai writing, The Difference to the West, China tries out much more in practice, be it a robot marathon or, as here, as a policeman. Effectively, what's undeniable is that China has an order of magnitude more experience

in building and deploying industrial robots and a significantly larger number of startups working on the technology. It also appears that so far, Chinese investors have been more interested in funding the next big robotics breakthrough because they've seen how transformative industrial robots have been to the country. All of that together has allowed Chinese companies like Unitri to get humanoids into production and shipping at scale. The G1 model humanoid is available in China.

and was even recently listed on Walmart's U.S. website for the honestly not crazy price of $21,600. Again, it remains hard to tell just how good the Unitary G1 is, but the point is that to the extent that there is a race, China is racing. Obviously, there are many, many applications of AI that will not be embodied or physical. Many of you listeners, very understandably, are more focused on how AI is going to help me do my job better in the next few weeks than how physical AI might shift how tasks get done a decade in the future. but there is no doubt that that intersection is coming and it's coming faster than it seems. So hopefully this was a useful primer.

That's going to do it for today's AI Daily Brief. Appreciate you listening or watching as always, and until next time, peace.