This Week in AI

Welcome to another issue of Synthetic. If you only read one or two articles this week, scroll down towards the bottom of this newsletter and read “Powerful AI Is Coming. We’re Not Ready” and the article on AI’s ‘Oppenheimer Moment.’ Then, stay for the cool robots and the big Nvidia update. We hope you’re having a great week.

This week in San Jose, Nvidia CEO Jensen Huang delivered the big keynote at his company’s annual developer conference, GTC. Here are the main takeaways from his lengthy speech:

• The Big Picture: Computing is moving from a retrieval model to one of generation. Rather than retrieving data from an intranet, database, wiki, or other data store, Nvidia expects computers to increasingly use AI to generate information based on ground-truth data used to train their models. “Computers will generate tokens rather than retrieve files,“ said Huang.

• Nvidia announced details on new Blackwell Ultra and Rubin chips. While shipments of Nvidia’s flagship ‘Blackwell’ chips ramp (and Nvidia is SOLD OUT on every chip they can make), the company is already planning to refresh that chip with Blackwell Ultra in the second half of 2025, doubling performance. A new chip architecture named after Vera Rubin, the scientist who discovered dark matter, will follow in the latter part of 2026, doubling performance once again and enabling a single AI server rack (see the headline image if you don’t know what one looks like) to deliver 15 exaFLOPS of performance (That’s a lot!). Rubin Ultra follows in 2027 with a new chip, Feynmann, coming in 2028. Nvidia is on a roll. 📈

• Nvidia sees the AI market progressing in four main steps: Perception AI, Generative AI, Agentic AI, and Physical AI. On physical AI (robots and self-driving cars), Jensen Huang said, “this could be the largest industry of all.”

• Nvidia announced an open-source humanoid robotics platform, GR00T N1, and a new partnership with DeepMind and Disney Research to create a new physics engine called Newton. Newton will power next-generation robots, including Star Wars entertainment robots Disney created for its theme parks. Jensen Huang introduced ‘Blue’ a Disney-built droid in his keynote. 🤖

• Nvidia believes AI data center capital expenditure will top $1 trillion by 2028. For comparison, AI capital expenditure by the top 5 cloud companies (Amazon, Google, Meta, Microsoft, Oracle) in 2024 was $185 billion and is expected to top $320 billion this year. 💰

• Nvidia announced new desktop ‘personal AI supercomputers’ - The machines, named DGX Spark and DGX Station, are designed to develop and run AI applications. Taking a leaf out of IBM/Intel’s book, these new DGX computers will use a standard design and be made by a range of OEMs, including Asus, Dell, HP, Lenovo, and Supermicro. The systems will feature Nvidia Blackwell GPU chips and are expected later this year. The retail price could be around $3,000.

• Nvidia says 100% of its software engineers will be ‘AI-assisted’ by the end of this year and that ultimately, all 30 million software engineers in the world will use AI to generate code. 💻

Synthetic’s Take: Nvidia continues to fire on all cylinders. They are sold out for the foreseeable future and selling every chip their manufacturing partner, TSMC, can make. The DeepSeek moment was a distraction, and Nvidia is way out in front for everything they do. It’s not only about their leading hardware; they have built a wealth of application-specific software to make AI development easy for their partners, from self-driving cars and humanoid robots to scientific simulations, computational biology, and enterprise AI applications. Some dismissed Nvidia’s product roadmap as overkill in a post-DeepSeek world. They don’t get it. The market opportunity for AI is ginormous, and we are only just at the beginning of the AI revolution. In a world of a trillion AI agents and a billion or more humanoid robots, demand for AI (and Nvidia products) will outstrip supply for many years to come.

Quick Hits

• AI Nurses Are Reshaping Hospital Care - And human nurses are pushing back. New AIs like Hippocratic AI’s Ana automate time-consuming tasks normally performed by medical assistants or nurses to reduce workload and prevent burnout. For example, Ana can address patient pre-op questions and concerns. Nursing unions don’t like it, and some practitioners cite dangerous errors. 👨‍⚕️👩‍⚕️

• AI Company Says Massive Shift Coming to Legal Profession - The founder of AI startup Harvey AI says a huge shift is coming and claims “junior folks are incredibly happy about this” as it automates low-value work and accelerates their path to more engaging legal work. 👩‍💼👨‍💼

• Robert De Niro Says ‘Everybody’s Worried’ In Hollywood 🎬 - De Niro’s remarks come after 400 stars and executives sent an open letter to the Trump administration expressing concerns over AI’s impact on creative industries.

• Softbank Purchases Ampere for $6.5 Billion - CEO Renee James (an old colleague at Intel) sells her chip company Ampere to the Japanese tech giant, implying a pivot for Ampere away from CPUs toward AI-focused processors.
  Synthetic’s Take - Softbank overpaid and is only buying Ampere’s engineers and leading-edge silicon know-how.

• OpenAI Says US Will Lose AI Race If It Can’t Steal Copyrighted Work - The company’s defense in ongoing intellectual property lawsuits is that unless it can scrape vast amounts of training data off the internet under ‘fair use,’ China will win the AI race since it faces no such restrictions. 🇨🇳

• OpenAI Calls DeepSeek ‘State-Controlled’ And Calls for Ban on Chinese Models - Some will see this as anti-competitive sour grapes, while others will see it as a vital step to secure US dominance of the AI space (which has profound national security implications). 🇨🇳

Video: Boston Dynamics Atlas Moves Like a Human

In this latest video from robotics pioneer Boston Dynamics, we see their all-electric Atlas humanoid robot jog, crawl, roll, do a handstand, a cartwheel, and even a little light breakdancing (Raygun would be proud). The robot was trained using a combination of reinforcement learning and human motion capture to exhibit smooth, human-like movements. (1m 7s)

Synthetic’s Take: Robots are as bad now as they’ll ever be. AI progress in robotics is exponential. By the end of the decade, humanoid robots will be commonplace in factories, warehouses, and even some homes.

AI Tech and Innovation

Scientists Test Robot With Potential To Revolutionize Food Production 🌱

Roboton is now selling an agricultural robot that uses solar energy to plant, water, fertilize, and tend plants in a field. It even recognizes and pulls out weeds. The robot is designed to address ongoing labor shortages and bring advancements in precision farming. It uses significantly fewer resources by spraying each plant with water, fertilizer, or pesticides. That means less waste, lower cost, and fewer chemicals leaching into groundwater. It conserves up to 83% of water, and by using solar panels to power its operations, it saves around 100 gallons of diesel fuel (400 liters) annually.

Synthetic’s Take: This robot, or something like it, has been around for over five years. What’s new? It’s moving from R&D into production. Deployed at scale, this robot could be a good thing for food prices and the environment.

AI-Generated Research Paper Passes Peer Review 🧬

Japanese startup Sakana built an AI research scientist to write AI research papers. The papers include scientific hypotheses, experiments, experimental code, data analyses, visualizations, and explanations of its work. ICLR, a respected AI conference, accepted a paper written by Sakana, making it the first AI-generated research paper to be peer-reviewed.

Synthetic’s Take: Sakana can’t yet turn out genuinely novel ideas, its capabilities are fragile, and it can make mistakes with citations (a common AI issue with hallucinations), but it clearly illustrates AI’s trajectory. Once AI research can be automated and AI can generate breakthrough ideas using complex reasoning capabilities, we might see a fast take-off of AI capabilities, and AGI may be here before we know it.

How Does ChatGPT Work? OpenAI’s Groundbreaking Chatbot, Explained 🎓

If you’re relatively new to AI and want to learn more about how the technology works, this explainer is for you. It reviews AI’s current capabilities, explains key terms (GPT, tokens, transformers, embedding) in non-technical, accessible terms, and explains how to get better results.

Synthetic’s Take: If you’re looking for another explanation of how modern AIs like ChatGPT work, watch this fun video in which a deep-fake Ryan Gosling explains all you need to know. For a more technical explanation of how large language models work, watch this excellent video about transformers from 3Blue1Brown.

AI Insights

Powerful AI Is Coming. We’re Not Ready.

Human-level AI, so-called AGI, could arrive sooner than we expect. Some believe it could come as soon as next year, and most industry insiders agree that AGI will be with us before the end of the decade. This excellent opinion piece by Kevin Roose, New York Times technology columnist and co-host of the Hard Fork podcast, makes the case that whether it arrives this decade or decades from now, we must start preparing for AGI now. Our leaders have no plan for the significant changes AGI will bring and aren’t even talking about it. Most companies are still grappling with the early stages of building their AI transformation strategies and not yet thinking about how AGI will turn entire industries upside down.

Synthetic’s Take: Everybody should be involved in a serious conversation about AGI's imminent arrival and impact. It should be a primary focus, but it’s not. Don’t underestimate how smart-as- or smarter-than-human AI and robots will impact society, the economy, the social contract, and almost every aspect of human existence. Yes, deploying new technology takes time. Things won’t change overnight. But AGI will diffuse into the world over existing infrastructure (broadband and cloud), so it will happen quite quickly, and we need to start planning NOW.

The ‘Oppenheimer Moment’ That Looms Over Today’s AI Leaders

Does AI present an existential risk to humanity (known colloquially in the AI business as ‘X-Risk’) or an unmissable opportunity to elevate our capabilities, cure disease, and propel ourselves into a new level of being; a world of abundance, leisure, and fulfillment? Our world is about to change, perhaps radically. Frontier AI lab leaders appreciate the significance of their work while simultaneously feeling the need to keep racing forward. Anthropic CEO Dario Amodei recently said, “Almost every decision I make is balanced on the edge of a knife…if we don’t build fast enough then the authoritarian countries could win.” This article by Time explores the significance of the moment and asks what governing bodies society should put in place to keep us all safe.

Tools To Try

Reclaim.AI: The AI assistant for your calendar

Maintain healthy calendar habits, improve your productivity, optimize cross-team meetings, boost collaboration, and improve work-life balance. Reclaim schedules 1:1s and defends your calendar so you can focus. Syncs multiple calendars and integrates with Google Calendar, Zoom, Slack, and HubSpot.

AmpleMarket: Elevate Your Salesforce With AI

A copilot for your salesforce that surfaces fresh leads, maximizes buyer engagement, and boosts email open rates by up to 50%. Empower your sales reps, uncover new opportunities, and grow revenue with an all-in-one, multi-channel AI platform that elevates the capabilities of your sales team. Free 14-day trials.

Looka: AI Logo Design and Branding

Make a logo in minutes—no design skills required! Auto-create a brand kit with fonts and colors to complement your new logo and create a brand identity. Instantly create branded marketing materials, then use over 300 templates to customize branded business cards, social profiles, email signatures, and more.

Firstbase: The easy way to start a new company

Incorporate your company, access one-click growth tools, stay compliant, and manage everything your business needs — all online, from anywhere. Launch your U.S. business in minutes with no paperwork or legal headaches.

Oyster: Find and hire international talent with a simple cross-border HR solution

Need to expand internationally, but not sure where to start? Oyster makes it easy to find, hire, and retain local talent. They handle the details (country-specific labor laws, international tax laws, compliance, and global payroll) so you can focus on finding the right talent and ramping their impact.