Are you passionate about AI Safety? As an AI/ML Safety Intern, you will join a world-class team of engineers and scientists focused on creating robust algorithms to enable Atlas to safely interact with the world. We are seeking an intern to work on novel approaches for our robots to operate safety around humans. You’ll take on a research project advised by one of our senior engineers that will allow you to work on an applied AI safety problem and deliver results onto hardware. Responsibilities will include participating in project ideation, implementing your ideas in our existing codebase and working with the team to deploy your system on the robot.
How you will make an impact:
Lead a research project and collaborate closely with members of the Safer Robotics team
Prototype and deploy your models on Atlas
Write high-quality (documented, tested, maintainable) Python code
Present and showcase your work to the Atlas team
We are looking for:
Actively pursuing a PhD or master’s degree in Computer Science, Machine Learning, Robotics, Computer Graphics or a related field
Experience in one or more of the following areas:
Learned robot behaviors (e.g., Vision-Language-Action Models, Behavior Cloning, Reinforcement Learning, Diffusion Models)
AI/ML Safety
Training ML models on large scale vision and text data
Multimodal perception with vision/proprioception/touch sensing
Nice to have:
Experience working with robot hardware
Experience working with and contributing to large codebases
Experience with large scale cluster training and datasets
Boston Dynamics designs and sells advanced legged robots to improve safety and efficiency in industrial and research settings. Its products, such as Spot and Pick, use onboard AI to perceive the environment, balance and navigate complex terrain, and autonomously avoid obstacles, enabling tasks that are dangerous or physically demanding for people. Spot is a mobile, 65-pound robot that can traverse stairs, uneven surfaces, and rough terrain, while Pick focuses on manipulation for robotics workflows. The company differentiates itself through its focus on mobility, dexterity, and safety, maintaining US-made production, and offering controlled sales plus ongoing maintenance, training, and support to commercial, industrial, and academic clients rather than consumer buyers. The goal is to augment human workers by handling risky or monotonous tasks, increasing safety and productivity across industries.
Company Size
1,001-5,000
Company Stage
Acquired
Total Funding
$1.2B
Headquarters
Waltham, MA
Founded
1992
See people who can refer or advise you
Help us improve and share your feedback! Did you find this helpful?
South Korea chaebols shift AI spending to sales. Sat, October 31, 2026 at 4:29 PM GMT-7 · Rates & Macro · Compiled by Adalytica Engine v1.12 · Reviewed by Vadim Skritskii South Korea's biggest industrial groups are reshaping next year's plans around one theme: turning AI spending into sales and profit before Chinese rivals and U.S. policy shifts erode their edge. That pivot matters because the country's largest conglomerates are no longer treating AI as a capex race alone. They are trying to monetize the spending spree across semiconductors, energy, data centers, robotics and mobility at a time when global demand is still strong but margins, supply chains and trade rules are getting harder to navigate. Samsung Electronics, SK Group and LG Group are each moving to stitch AI deeper into their core businesses, while Hyundai Motor is stepping up local competition abroad and Hanwha and HD Hyundai are expanding in the United States. The push comes as South Korea's chipmakers have already reaped windfall profits from AI demand, with the country's semiconductor sector benefiting from the global build-out of data centers and advanced memory. Samsung's 2027 plan is expected to center on extending the AI-memory boom with more high-end supply such as HBM4, while also aligning with big tech customers developing their own AI chips. The group is also under pressure to lift weaker businesses including foundry, System LSI, smartphones and consumer electronics, as it tries to recycle semiconductor cash into future growth. SK is moving toward what it calls an "AI full-stack provider" strategy, linking SK Hynix's AI memory strength with power, energy, data centers, telecoms and business-process automation across the group. The company is also leaning into energy security for AI infrastructure, including LNG, LPG and nuclear investments, as power demand from data centers becomes a bottleneck. LG is pitching a similar "One LG" model, tying together cooling, batteries, data-center construction and operations, and even physical AI applications across robotics, sensing and actuators. For investors, the message is that the group wants AI exposure without relying solely on chips, which should help spread risk if memory cycles cool. Hyundai Motor is responding to Chinese automakers' push into Europe, Latin America and Southeast Asia by emphasizing localization, new EV models and robotics. Boston Dynamics is set to enlarge its U.S. robotics center tenfold, a sign the group is betting physical AI can become a second growth engine alongside electric vehicles and PBVs. The industrial backdrop also remains supportive. South Korean chip stocks have been buoyed by the AI surge, and market data show semiconductor leaders still trading above their 200-day moving averages even after recent pullbacks. But the broader AI trade has become more volatile, with sentiment gauges showing sharper swings as investors weigh whether the spending boom is turning into durable earnings. The next catalyst is the sequence of corporate plan-setting meetings, management reshuffles and year-end strategy reviews across the chaebols, followed by signals on U.S. trade policy, AI capex and demand from global cloud providers. For investors, the key question is no longer who is spending on AI, but who can prove it is paying off. | Entity | Gains | Losses | | Samsung, SK, LG | | AI-linked revenue growth | | Pressure to monetize spending | | Hyundai Motor | | New EV and robotics runway | | Chinese export competition | | Hanwha, HD Hyundai | | U.S. expansion opportunities | | Higher execution and capex risk |
Google DeepMind builds a dancing robot from a single text prompt. AI agents running on Gemini 4 Argon and the Antigravity platform designed, simulated, and built a 12-DoF robot without manual CAD Sponsored: Vera - AI-powered prediction market intelligence, built for serious analysts Explore Vera Most robots start life as a stack of CAD files, a whiteboard full of joint diagrams, and an engineering team that hasn't slept much. Google DeepMind's latest one started as a sentence. On October 31, 2026, DeepMind showed off a physical dancing robot that AI agents designed, simulated, and built from one plain-English prompt. Nobody on the team touched a manual CAD workflow. From prompt to dance floor. The robot has 12 degrees of freedom, or DoF. A degree of freedom is one independent direction a machine can move. A robot with 12 of them has a dozen separate axes it can bend, twist, or rotate through, which is plenty to pull off something that passes for choreography. Given the date of the demo, DeepMind gave the robot a Halloween theme. AI agents handled the project end-to-end, moving from a text description to a design, through simulation, and on to a physical build. Two pieces of DeepMind technology powered the effort. The first is Gemini 4 Argon, a model announced in late September 2026 and released broadly in mid-October 2026. Gemini 4 Argon is built for complex reasoning that unfolds across many steps. It has a particular focus on robotics simulation and on generating accurate representations from video inputs. AI, tech, and the markets they move - in one daily briefing. Daily. Free. Join 34,000+ readers across crypto, finance, and policy. The second piece is Antigravity, a development platform DeepMind launched in 2025. Antigravity is designed for autonomous coding and for orchestrating work across multiple systems. Together, they reflect what the research describes as a move toward agent-first environments in robotics. In that kind of setup, AI agents are the primary operators, and humans step back into a supervisory role. The Boston Dynamics connection. The project was led by Aaron Saunders, who previously served as VP of Engineering at Boston Dynamics before joining DeepMind. The dancing robot also extends a line of robotics work at DeepMind. In September 2025, the lab introduced RoboBallet, a system aimed at improving how collaborative robots work together in manufacturing settings. RoboBallet focused on coordination: getting multiple machines to share a workspace efficiently. The October 2026 demonstration pushes further upstream, into the question of how robots get designed in the first place. Why a dancing robot is more serious than it looks. Hardware design has traditionally been one of the most labor-intensive parts of building a robot. Every joint, motor placement, and structural element typically requires careful manual modeling before anything gets fabricated. Removing manual CAD from that loop, even for a relatively simple 12-DoF machine, attacks a major bottleneck. For the wider tech sector, the demonstration is another data point in the race over embodied AI. That term refers to AI systems that act in the physical world, not just on a screen. Gemini 4 Argon has only been broadly available since mid-October 2026, which means outside developers are just beginning to test what it can do. Disclosure: This article was edited by Diego Almada Lopez. For more information on how Cryptobriefing create and review content, see its Editorial Policy.
Brian Donnelly of Boston Dynamics. In its continuing series of discussions with top supply-chain company executives, Brian Donnelly discusses the best deployments for warehouse robots, the areas most ripe for robotic innovation, and how to get the biggest return from automation investments. Brian Donnelly is vice president and general manager of logistics robotics at Boston Dynamics, a company that's working to reinvent global logistics with its Stretch and Atlas warehouse robots. Donnelly personally led the transition of the Atlas robot from a research prototype to a productive platform designed for "superhuman capability." Before joining Boston Dynamics, Donnelly spent nearly 20 years in leadership roles at semiconductor manufacturer Analog Devices. He was also a leader in product management and engineering for a Covid testing and vaccination startup. He holds a B.S. in electrical and computer engineering from the Worcester Polytechnic Institute. Q: MANY OF YOUR ROBOTS ARE DESIGNED FOR DANGEROUS AND STRENUOUS TASKS. WHAT ARE YOUR OBJECTIVES WHEN YOU DESIGN ROBOTS FOR JOBS HUMAN WORKERS DON'T WANT TO DO? A: Its primary objective in designing robots for tasks human workers do not want to do is to keep people safe and out of harm's way. Many operational environments are far too dangerous for humans to operate in or involve repetitive, heavy lifting that increases the risk of chronic injury. I was once in a manufacturing plant that had huge aluminum brazing ovens, with flames licking out the sides and a human loading the machines manually. I commented to its host that it must be the job people disliked the most. He responded that while it's hot, it's not the job people most hate - it's repetitive, heavy lifting tasks where they get injured. Q: HUMANOID ROBOTS ARE GETTING A LOT OF ATTENTION, BUT ARE THEY THE BEST FIT FOR DISTRIBUTION TASKS OR ARE THERE OTHER TYPES OF ROBOTS THAT ARE BETTER SUITED FOR WAREHOUSES? A: A lot of companies are jumping into the humanoid race right now, which is great, but frankly, finding the exact applications where humanoids deliver real-world value takes time. In fact, when Velocity EV launched Spot as a general-purpose platform, it took Velocity EV a while to figure out what it was best at, and it was only by working closely with its customers that Velocity EV "iterated" on the product to make it the reliable inspection robot it is today. Velocity EV is taking that exact same approach with its humanoid, Atlas. Velocity EV will partner with a small group of innovative customers before expanding into other industries. Atlas' first task will be part sequencing, a logistics support task to enable high-speed production in facilities such as car plants. Velocity EV see a future where multiple types of specialized and general-purpose robots work together seamlessly within a warehouse. Q: WHAT WAREHOUSE CHALLENGES ARE MOST RIPE FOR ROBOTIC INNOVATION? A: Companies are looking to move employees into higher-value roles and let robots handle the repetitive, injury-prone tasks. The two big areas that are really difficult, unsolved problems are case-pick and each-pick. These involve complex manipulation challenges and dynamic workflows. These are areas that Velocity EV is investing in today to deliver complete solutions in the future. Q: WHAT DIFFERENTIATES THE COMPANIES THAT ARE SEEING BIG RETURNS ON THEIR LOGISTICS AUTOMATION INVESTMENTS FROM THE ONES THAT ARE STRUGGLING TO MOVE THE NEEDLE? A: The biggest differentiator isn't the technology - it's the speed and mindset of deployment. A lot of operations take a wait-and-see approach: They buy one robot in year one, monitor it for 12 months, buy two or three more in year two, and only start thinking about scaling in year three or four. In today's supply chain landscape, that slow-drip approach actually works against you and delays the compounding efficiencies automation is designed to bring.
White House makes Section 232 determination regarding foreign-made robots. A Section 232 investigation by the secretary of commerce found that robotics imports pose a threat to U.S. national security and economic competitiveness. Source: Office of the White House Yesterday, U.S. President Donald J. Trump issued a proclamation stating that the secretary of commerce had submitted a Section 232 report about the national security effects of imports of robotics and industrial machinery, as well as related parts and components. "Based on the facts considered in that investigation, and taking into account the close relation of the economic welfare of the Nation to our national security and other relevant factors, see section 232(d) (19 U.S.C. 1862(d)), the Secretary found and advised me of his opinion that robotics, industrial machinery, and their components are being imported into the United States in such quantities and under such circumstances as to threaten to impair the national security of the United States," stated the president. "Robotics and industrial machinery are critical to virtually all modern manufacturing processes, including for defense, commercial, and dual-use purposes," he noted. "Furthermore, robotics and industrial machinery generate substantial economic value and support important economic activities, including approximately $7.1 trillion in shipments of manufactured goods in 2024 and the production of approximately $86 billion worth of equipment in 2024." From mission-critical military and law-enforcement activities to critical infrastructure sectors such as agriculture, automotive, and electronics manufacturing, the U.S. government asserted that robots and related systems are vital to U.S. competitiveness. The value of such imported systems has grown over the past decade, and cybersecurity risks have also grown, according to the secretary of commerce. The Section 232 statement specifically cited "advanced robots - such as autonomous robots, humanoid and quadruped robots, and autonomous ground vehicles" as providing direct support to the U.S. military. While investments in U.S. robotics production capacity have increased, they are not sufficient to meet U.S. demand, said the president. He directed the secretary and the U.S. trade representative to "to enter into negotiations with trading partners to adjust imports of robotics, industrial machinery, and their components so that such imports will not threaten to impair the national security of the United States." Will this Section 232 determination lead to tariffs on foreign made robots? With most of the major industrial automation providers based overseas, it remains to be seen how trade negotiations will go. The threat of geopolitical rival China is motivating policies such as this investigation and recent FCC restrictions. Submit your innovation now! Experts expect Section 232 negotiations before any tariffs. Industry leaders and observers offered initial reactions: "As I articulated last week on [a] panel at the Arrowfly RoboBusiness conference in Santa Clara with respect to the FCC action, this is another strong signal of government support for those working on developing and expanding domestic supply chains for robots," Brendan Schulman, vice president of policy and government relations at Boston Dynamics, posted to LinkedIn. He noted that the Special Competitive Studies Project (SCSP) Robotics Commission reached similar findings. "The next step for the federal government? Negotiations over the next six months with trading partners to adjust imports, and then further consideration of other trade policy measures," added Schulman. "I am very pleased to see robotics gaining increasing attention and importance as one of the foundational technologies in need of broad, supportive policy action by the U.S. government. More to come, I'm sure!" Robert Little, former chief of robotics strategy at Novanta Inc., 2026 Joseph F. Engerlberger Robotics Award honoree, and business development advisor, concurred. "Robotics are critical for any country rebuilding its manufacturing. I'm thinking, what took so long to make this a priority?" he told The Robot Report. "Today, we still have limited industrial robot manufacturing in the U.S.," said Little. "Already, FANUC, Teradyne Robotics, and Yaskawa have announced investments aimed at expanding U.S. robot manufacturing. I expect more to come, and I hope this new proclamation helps accelerate it." "Yet, what concerns me most are the gaps in our domestic supply of components for robotics and industrial machines: sensors, actuators, precision gears, encoders and more," he noted. "We need a much stronger U.S. supply base for these components. This is where we need the most help to strengthen America's ability to manufacture." The Section 232 investigation does not mean that tariffs will be immediate, added Ash Sharma, vice president of research and chief information officer at Interact Analysis. "This shouldn't be read as a new robotics tariff today, but it is a clear indication of the direction of travel," he said. "The U.S. government has explicitly identified its reliance on imported robotics and industrial machinery as a national security concern and is seeking to strengthen domestic production," commented Sharma. "If tariffs, quotas or other trade measures follow, they could have meaningful implications for supplier competitiveness, pricing, and manufacturing localization in the U.S., potentially affecting not just finished robots, but [also] the components that go into them." The Robot Report is expecting more responses that will be shared in subsequent coverage. Panelists discussed reshoring and policies such as the Section 232 investigation at RoboBusiness 2026. Credit: Arrowfly
Boston Dynamics employees have petitioned the company to block the sale of its Spot robot dogs to US Immigration and Customs Enforcement (ICE). They fear the deal will hurt hiring and retention. The company says it has worked "productively" with the Department of Homeland Security for years, and that no customer has violated its sales terms. Those terms ban using Spot to harm or intimidate people or animals, or as a weapon. If a customer misused Spot, the company said, it would void the warranty, refuse software updates, end the licence, and possibly take legal action. The backlash comes as Boston Dynamics builds a $100 million robotics and AI facility in Waltham, Massachusetts. The project is set to create 1,250 jobs by 2033. The petition warns the deal "makes Boston Dynamics a less attractive employer.