Tech Brew Ride Home - Tue. 05/20 – Microsoft Build 2025

Episode Date: May 20, 2025

He noted the firm has the OVX computer that is meant to do simulation and graphic simulation physics engine, and it is used to synthesize and generate data. And this data is consumed by the DGX computer, which are used to train foundation models. And then it is deployed to the HX computer, which is the runtime on the edge for platforms like humanoid robots. Sponsors: 1password.com/ride Links: GitHub’s new AI coding agent can fix bugs for you (The Verge) The new Microsoft Discovery agentic platform targets scientists and researchers (Neowin) Trump signs the Take It Down Act into law (The Verge) Stablecoin Bill Advances in US Senate in Big Win for Crypto (Bloomberg) EU to impose €2 tax on low-cost items in blow to Temu and Shein (FT) Nvidia charges ahead with humanoid robotics aided by the cloud (GamesBeat) Autonomous cars with ‘social sensitivity’ cut threat to road users, study finds (FT) Mountainhead Trailer Learn more about your ad choices. Visit megaphone.fm/adchoices

On April 4th, 2023, around 2 in the morning, a man was found stabbed multiple times on a sidewalk in downtown San Francisco. Hey, who did this to you? What happened next turned the story into a political firestorm. Reports have identified the victim as Bob Lee, the founder of Cash App. From Bloomberg Podcasts, this is Foundering, the Killing of Bob Lee, beginning April 16. Welcome to the Tech meme right home for Tuesday, May 20th, 2025. I'm Brian McCullough today. All the headlines from the Microsoft Build 2025 keynote. A whole slew of legislation that relates to tech is making its way through Congress. Temu and Sheehan can't catch a break. Invita is serious about using AI for robotics and how something called social sensitivity can make autonomous cars even more safe. Here's what you missed today in the world of tech. They did this to me last year, I think, as well. The Microsoft Build Conference kicked off yesterday. and Google I.O. is today, so it's hard to cover both at the same time without shortchanging one or the other. So, we'll cover build today and I.O. tomorrow. Build first, beginning with Microsoft-owned GitHub

debuting an AI coding agent for GitHub copilot that can fix bugs, add features, improved documentation, and more. And they are open-sourcing GitHub copilot in VS code, quoting the verge. To complete its work, GitHub says the AI coding agent will automatically boot a virtual machine, clone the repository, and analyze the codebase. It also saves its changes as it works, while providing a rundown of its reasoning in session logs. When it's finished, GitHub says the agent will tag you for review. Developers can then leave comments that the agent will automatically address. The agent also incorporates context from related issue or PR pool request discussions and follows any custom repository instructions, allowing it to understand both the intent behind the task and the coding standards of the project, GitHub says. The new code

agent is available to copilot Enterprise and Copilot Pro Plus users through GitHub site, its mobile app, and the GitHub command line interface tool. Microsoft also announced that it's open sourcing GitHub copilot in Visual Studio code, which means developers will be able to build upon the tool's AI capabilities, end quote. But that wasn't the end of the open sourcing because Microsoft also open source the Windows subsystem for Linux and released its code on GitHub, except for a few Windows-specific components. Microsoft also launched NL Web, an open project that lets developers add a conversational interface to their website with a few lines of code, an AI model, and data. Quoting TechCrunch. You can use the AI model of your choice and your own data. A retailer could

use NL Web to create a chat bot that helps users choose clothing for specific trips, for example, while a cooking site could use it to build a bot that suggests dishes to pair with a recipe. Web pages built using NLWeb can optionally make their content discoverable and accessible to AI platforms that support MCP, Anthropics Standard for Connecting AI models to the systems where data resides. We believe NLWeb can play a similar role to HTML for the Agentic Web, writes Microsoft in press materials provided to TechCrunch. It allows users to interact directly with the web content in a rich semantic manner. Microsoft didn't say either way, but NLWeb may have its origins in tech from chat GPT Maker OpenAI, Microsoft's close collaborator, end quote. Microsoft also expanded Entra, Defender, and Perview,

embedding them directly into Azure AI Foundry and Copilot Studio to help organizations secure AI apps and agents. For its co-pilot Studio agents, Microsoft announced a computer use feature available in its frontier program for select US users, WhatsApp integration, and more. They added GROC 3 and GROC 3 Mini to their Azure AI Foundry service, and sources say Satcha Nadella had been pushing for Microsoft to host GROC, quoting The Verge. It's a surprise edition that could prove controversial internally and further inflamed tensions with Microsoft's partner OpenAI. Microsoft has been steadily growing its Azure AI foundry business over the past year and has been quick to embrace models from a variety of AI labs that compete with its

OpenAI partner. In January, I reported in Notepad that Microsoft CEO Satcha Nadella had moved with haste to get engineers to test and deploy Deepseek R1 as it made headlines around the world. engineers didn't sleep much over those days while they worked over time to get R1 ready for Azure AI Foundry. Sources tell me Nadella has been pushing for Microsoft to host Grok, and he's eager for Microsoft to be seen as the hosting provider for any popular or emerging AI models. Grok is the latest model to join the Azure AI Foundry, which is quickly becoming an important AI service for Microsoft as it seeks to be seen as the platform to host AI

models for businesses and app developers, end quote. Microsoft also debuted Foundry Local for Windows and MacOS on Onix runtime to let developers build cross-platform AI apps that can run models, tools, and agents on device. And finally, Microsoft announced Microsoft Discovery, an agenetic platform to accelerate scientific discovery and enterprise R&D efforts. Quoting NeoWin, Microsoft Discovery doesn't lock researchers in with Microsoft's own tools. Instead, it has been built to be highly extensible, allowing researchers to integrate with models from other partners and even their own models, tools, and data sets when required.

Also, it is built on top of a graph-based knowledge engine to provide a deep understanding of conflicting theories, diverse experimental results, and more. Researchers can also validate and understand every step or make changes if required. Microsoft says its own researchers use Microsoft Discovery's AI models and HPC simulation tools to discover a new coolant prototype to be used in data centers in around 200 hours, a process that Microsoft says would have taken months or years otherwise. Microsoft is already working with a select set of Microsoft customers from various industries, including chemistry and materials, silicon design, energy manufacturing, and pharma

to develop Microsoft Discovery's capabilities. With its extensible architecture and AI capabilities, Microsoft Discovery promises to significantly shorten research timelines across various scientific fields. Back in February, Google announced a somewhat similar product called AI Co-Scientist. It's a multi-agent AI system built with Gemini, designed to serve as a virtual scientific collaborator that helps scientists generate novel hypotheses and research proposals. It will be interesting to see Microsoft and Google compete to win over researchers and scientists in the coming years, end quote. President Trump has signed the Take It

Down Act, criminalizing the distribution of non-consensual intimate images, basically deepfakes, and requiring platforms to promptly remove them when notified. Quoting the verge, The bill sailed through both chambers of Congress with several tech companies, parent and youth advocates, and First Lady Melania Trump championing the issue. But critics, including a group that's made its mission to combat the distribution of such images, warn that its approach could backfire and harm the very survivors it seeks to protect. The law makes publishing NCI, whether real or AI generated, criminally punishable by up to three years in prison plus fines. It also requires social media platforms to have processes to remove end. NCI within 48 hours of being notified and, quote, make reasonable efforts to remove any copies.

The Federal Trade Commission is tasked with enforcing the law and companies have a year to comply. Under any other administration, the Take It Down Act would likely see much of the pushback it does today by groups like the Electronic Frontier Foundation and Center for Democracy and Technology, which warn the takedown provisions could be used to remove or chill a wider array of content than intended, as well as threatened privacy protecting technologies like encryption, since services that use it would have no way of seeing or removing the messages between users. But actions by the Trump administration in his first 100 days in office, including breaching Supreme Court precedent by firing the two Democratic Minority Commissioners at the FTC may have added another layer of fear for some of the law's

critics who worry it could be used to threaten or stifle political opponents. Trump, after all, said during an address to Congress this year that once he signed the bill, quote, I'm going to use that bill for myself, too, if you don't mind, because nobody gets treated worse than I do online. Nobody, end quote. The Cyber Civil Rights Initiative, which advocates for legislation combating image-based abuse, has long pushed for the criminalization of non-consensual distribution of intimate images or NDI, but the CCRI said it could not support the Take It Down Act because it may ultimately provide survivors with, quote, false hope. On blue sky CCRI, President Mary Ann Franks called the take down provision a poison pill that will likely end up hurting victims more than it helps.

Platforms that feel confident that they are unlikely to be targeted by the FTC, for example, platforms that are closely aligned with the current administration may feel emboldened to simply ignore reports of NDII, they wrote. Platforms attempting to identify authentic complaints may encounter a sea of false reports that could overwhelm their efforts and jeopardize their ability to operate at all, end quote. But that's not all on the legislative front because the U.S. Senate has advance the Genius Act, a stable coin bill after a group of Democratic senators dropped their opposition to the bill, marking a major win for the crypto industry. Quoting Bloomberg. The industry-backed regulatory bill is now set for debate on the Senate floor with a bipartisan

group hoping to pass it as soon as this week, although senators said a final vote could slip until after the Memorial Day recess. Democrats had united to filibuster the legislation earlier this month amid a furor over President Donald Trump's crypto dealings, along with other concerns related to the regulation of stable coins. But the Senate voted 66 to 32 on Monday night to end the filibuster. Crypto-friendly Democrats led by Senators Kristen Gillibrand of New York and Angela also Brooks of Maryland negotiated modifications to the legislation and urged their colleagues to support it, even without a ban on Trump profiting from his family's many crypto ventures while in office. But Democratic Senator Mark Warner of Virginia, an influential moderate on the

Banking Committee announced Monday he would support the measure, adding that concerns over the Trump family's business dealings shouldn't sideline broader stable coin legislation. The legislation is, quote, not perfect, but it's far better than the status quo, Warner said, end quote. And one more bit of congressional news. The U.S. House Budget Committee has advanced a budget bill that would ban U.S. states from enforcing any law regulating AI for 10 years. The bill now goes to the House. The hits keep coming for Temu and Sheehan. Sources and a document have revealed that the EU is planning to levy a flat two-euro fee on billions of small packages that enter the block mainly from China on a daily basis. So pretty negative for the business model of you know who.

Quoting the F.T. The European Commission circulated a draft proposal on a handling fee on Monday after pressure from other member states whose customs authorities are overwhelmed by the 4.6 billion items annually imported directly to people's homes. The proposal, seen by the FTE, does not set the fee level, but people familiar with the commission's thinking suggested it would be about two euros. Some of the money would cover customs costs, but also go into the EU budget, adding billions annually. Trade Commissioner Maros Sefcovic has promised to tackle the surge in packages, which he said had led to an increase in dangerous and non-compliant goods and complaints by EU retailers of unfair competition, end quote.

NVIDIA has unveiled Isaac Grute N1.5, an open, customizable AI model for humanoid reasoning and skills, and also released Groot Dreams, a tool for generating synthetic motion data, quoting Games Beat. Nvidia said it is racing ahead with humanoid robotics technology providing a custom foundation model for humanoid reasoning, a blueprint for generating synthetic motion data, and more Blackwell systems to accelerate humanoid robots. development. At the Computex 2025 trade show in Taiwan, NVIDIA unveiled Isaac Grute N1.5, the first update to Nvidia's open, generalized, fully customizable foundation model for humanoid reasoning and skills.

NVIDIA Isaac Grute Dreams is a blueprint for generating synthetic motion data and Nvidia Blackwell systems to accelerate humanoid robot development. Humanoid and robotics, agility robotics, Boston Dynamics, Fourier, FoxLink, Galbot, Mantee Robotics, Neuro Robotics, General Robotics, Skilled AI, and XPeng Robotics are adopting NVIDIA-Isaac platform technologies to advance humanoid robot development and deployment. Showcase in CEO Jensen Huang's CompuTech's keynote address, Invita Isaac Dreams, is a blueprint that helps generate vast amounts of synthetic motion data, aka neural trajectories, that physical AI developers can use to teach robots new behaviors, including how to adapt to changing environments.

Developers can first post-trained Cosmos predict world foundation models for their robot, then using a single image as the input. Groot Dreams generates videos of the robot performing new tasks in new environments. The blueprint then extracts action tokens, compressed digital pieces of data, that are used to teach robots how to perform these new tasks. The Grute Dreams blueprint complements the Isaac Grute Mimic blueprint, which was released at the Nvidia GTC conference in March, while Groot Mimic uses the Nvidia Omniverse and Nvidia Cosmos platforms to augment existing data,

Groot Dreams uses Cosmos to generate entirely new data. Jim Fan, director of AI and distinguished scientists at Nvidia said in a press briefing, NVIDIA has a very strong robotic strategy, and it is centered around what Jensen calls the three-computer problem. He noted the firm has the OVX computer that is meant to do simulation and graphics, simulation physics engines, and is used to synthesize and generate data. And this data is consumed by the DGX computer, which are used to train foundation models. And then it is deployed to the HX computer, which is the runtime on the edge for platforms like humanoid robots.

Groot is the lifecycle of physical AI and robot-based workflows, fans said. It is an instantiation of the three computer problem, he said, end quote. Finally today, this is interesting. Autonomous vehicles train to use what they're calling social sensitivity in assessing the collective impact of multiple hazards leads to fewer injuries during road accidents involving these autonomous vehicles. Quoting the F.T. Autonomous cars that are trained to respond more like humans to danger will cause fewer injuries during road accidents according to a study that shows how driverless vehicles might be made safer. Vulnerable groups such as cyclists, pedestrians, and motorcyclists

saw the biggest gain in protection when driverless cars used social sensitivity in assessing the collective impact of multiple hazards. The study published in the U.S. proceedings of the National Academy of Sciences highlights growing efforts to balance AV's efficient operation with the need for them to minimize damage and collisions. The issue of AV ethics is attracting increasing attention as growing use of the cars offers the prospect of eliminating driver problems such as spatial misjudgments and fatigue. The study suggests human behavioral methods could, quote, provide an effective scaffold for AVs to address future ethical challenges, said it's China and U.S.-based authors led by Hong Lang Lu of the Hong Kong University of Science and Technology.

Based on social concern and human plausible cognitive encoding, we enable AVs to exhibit social sensitivity in ethical decision-making, they said. Such social sensitivity can help AVs better integrate into today's driving communities. Social sensitivity included being attuned, like human drivers, to the vulnerabilities of specific road users and being able to judge who was likely or to be more seriously hurt during a crash. The researchers drew on evidence from neuroscience and behavioral science that humans navigate using a cognitive map to interpret the world and adapt accordingly. The scientists based their instructions for the AV on a concept known as successor representation,

which encodes predictions of how different elements in an environment will interrelate across time and space. They examine the results of harnessing their model to ethical planner, a system AVs used to make decisions accounting for various risk considerations, The researchers modeled 2,000 benchmark scenarios measuring total risk of each one by assessing the probability of collision and the likely severity of harm for the people involved. The scientists found that using their human-inspired model with ethical planner cut overall risks to all parties by 26.3% and by 22.9% for vulnerable road users compared with using ethical planner alone. In crash scenarios, all road users suffered 17.6% less harm, rising to 51.7% for vulnerable users. The occupants of the AV were also better off, experiencing 8.3% less harm, end quote. Today I learned that Jesse Armstrong, that dude who did the TV show Succession, has a movie coming out about tech billionaires.

It's actually coming out at the end of the month on HBO, I guess. It's called Mountain Head. Must see viewing for us, I guess, right? Link to the trailer in the show notes today. Talk to you tomorrow.