- Everyday AI
- Posts
- Humanoids in our world. How it'll work and what's next
Humanoids in our world. How it'll work and what's next
Federal copyright ruling on AI art, GPT-4.5 report on hallucinations, Italian newspaper publishes all AI issue.
Everyday AI
March 19, 2025
š Subscribe Here | š£ Hire Us To Speak | š¤ Partner with Us | š¤ Grow with GenAI
Outsmart The Future
Today in Everyday AI
7 minute read
š Daily Podcast Episode: Humanoids are coming faster than you think. Donāt freak out. Itās actually a good thing. Go listen.
šµļøāāļø Fresh Finds: A viral AI humanoid breakdancing, Microsoft's bold AI leadership shift, and a Deeper Deep Research. Read on for Fresh Finds.
š Byte Sized Daily AI News: Federal copyright ruling on AI art, GPT-4.5 report on hallucinations, Italian newspaper publishes all AI issue. Read on for Byte Sized News.
š§ Leverage AI: Humanoids are equally weird and amazing. Are they coming to take our jobs? Or just make them safer and better? Keep reading for that!
ā©ļø Donāt miss out: Did you miss our last newsletter? We talked about NVIDAās big announcements, Google drops two huge features in Gemini, and Adobe goes agentic. and more.. Check it here!
The Next Frontier of Startups: Domain Experts in the Age of AI
Will we see humanoids in the grocery store soon? š¤
One of the few sectors of AI advancing at the same pace as LLMs is robotics and humanoids.
Spoiler alert ā NVIDA powers the humanoid movement.
So what was announced at NVIDIA ās GTC conference?
Will we be seeing humanoids out and about anytime soon?
Click below to find out from a leading expert at the forefront.
Also on the pod today:
Why humanoids sometimes work 8 hour shifts robot ā°
Whatās new in humanoids from GTC š¤
How everyday businesses might use humanoids š¦¾
Itāll be worth your 22 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Hereās our favorite AI finds from across the web:
New AI Tool Spotlight ā Sider brings visual reports to AI Deep Research, Superlines helps you get discovered in AI search results, Autodraft helps you create 4K animations for YouTube. (With the help of AI, of course.)
AI Productivity ā Maybe āDeep Searchā is so yesterday? Grok reveled āDeeper Searchā
NEWS: Grok has launched "DeeperSearch" - an upgraded version of DeepSearch
ā X Daily News (@xDaily)
11:12 AM ā¢ Mar 19, 2025
Restaurants and AI ā Yum Brands teams up with Nvidia to supercharge AI in restaurants.
AI Models ā OpenAIās Noam Brown reveals how 'reasoning' AI models like o1 couldāve emerged decades earlierāand why academia still has a shot at shaping the future of AI research.
AI and Leadership ā Microsoft shifts HR chief Kathleen Hogan to lead corporate strategy in AI-driven transformation. See why.
Humanoids ā People are losing their minds over this AI-powered, breakdancing robot from Boston Dynamics.
Atlas is demonstrating reinforcement learning policies developed using a motion capture suit. This demonstration was developed in partnership with Boston Dynamics and @rai_inst.
ā Boston Dynamics (@BostonDynamics)
2:01 PM ā¢ Mar 19, 2025
1. No Copyright for AI-Generated Art ā
A federal appeals court in Washington, D.C., has ruled that artworks created solely by artificial intelligence, without human input, cannot be copyrighted under U.S. law. The decision upholds the U.S. Copyright Officeās stance that human authorship is essential for copyright protection, rejecting Stephen Thalerās bid to copyright a piece his AI system, DABUS, generated independently.
This ruling could shape how generative AI tools like Midjourney are used by creators, as the court affirmed that AI-assisted works may differ from those entirely AI-generated. Thaler plans to appeal, arguing the decision could stifle innovation in this rapidly evolving field.
2. Nvidia Joins $100B AI Infrastructure Push š«ø
Nvidia and Elon Musk's xAI have joined a powerhouse consortium, now rebranded as the AI Infrastructure Partnership (AIP), aiming to supercharge U.S. AI infrastructure, Reuters reported Wednesday. Backed by Microsoft, BlackRock, and MGX, the group plans to invest over $30 billion initially, with ambitions to mobilize up to $100 billion for data centers and energy solutions critical for powering AI systems like ChatGPT.
This move comes as competition heats up following President Donald Trump's Stargate initiative, which has a staggering $500 billion target. With Nvidia as a technical advisor and renewable energy firms like NextEra onboard, this signals a major push to scale AI while tackling energy efficiency challenges
3. AI Accuracy Examined: GPT-4.5 Scores Better, but Hallucinations Persist š¤„
OpenAI claims its GPT-4.5 model "hallucinates less" than earlier versions, with a new benchmark called SimpleQA showing it makes mistakes 37% of the time ā a significant improvement over GPT-4o's 62%.
Researchers suggest human oversight and usage metrics could better judge AI reliability, while others warn that adding more data may hit diminishing returns as models reach their limits. Despite improvements, experts agree generative AI will never be fully hallucination-free, pointing to inherent challenges with truth, bias, and user demands for creativity.
4. Italyās First AI-Generated Newspaper Hits Newsstands šļø
In a groundbreaking move, Italy's Il Foglio has become the first newspaper to release an edition entirely created by artificial intelligence, according to its editor, Claudio Cerasa.
The four-page supplement, dubbed Il Foglio AI, handles everything from headlines and writing to humor and reader letters, showcasing how AI could reshape journalism. While some articles tackle global political topics like Trump and Putin, others explore cultural shifts, such as young Europeans avoiding steady relationships. This bold experiment highlights the growing role of AI in media and raises questions about its future impact on traditional journalism.
5. Nvidiaās Huang Unfazed by DeepSeekās AI Progress
Despite Chinaās DeepSeek R1 model shaking up the AI landscape with its low-cost training approach, Nvidia CEO Jensen Huang isnāt concerned. Speaking at GTC 2025, Huang dismissed fears that affordable reasoning models like DeepSeek could undermine Nvidia, arguing that agentic AI applications demand far more computing power than previously thought.
Huang touted Nvidiaās new Blackwell Ultra and Vera Rubin superchips as essential for scaling such advanced AI, signaling confidence in the companyās long-term dominance.
š¦¾How You Can Leverage:
Humanoid robots are working 8-hour warehouse shifts RIGHT NOW (and they're totally judging your lunch break)
āForget ChatGPT. That's baby stuff.
ACTUAL HUMANOID ROBOTS are already replacing humans in warehouses while you're still struggling with prompt engineering.
At NVIDIA GTC, we sat down with humanoid robotics leader Pras Velagapudi, Agility Roboticsās CTO.
He dropped some robotics truths on us : their robot Digit isn't some concept car fantasyāit's clocking actual shifts at GXO logistics facilities under multi-year contracts.
These metal coworkers could work 24/7 but only do 8-hour shifts because the humans around them need inconvenient things like "sleep."
Rude.
So what else do you need to know about humanoids?
Welp, some new announcements at NVIDIAās GTC conference changed whatās possible with humanoids.
Letās dive in. š
1. These robots are literally using Target shopping baskets š
Lab demos are for amateurs.
Real robots are out here DOING WORK in unmodified human spaces.
Pras casually mentioned they bought regular shopping baskets off the internet for their demo at NVIDIAās GTC conference. No specialized robot containers. No fancy custom shelving.
Just off-the-shelf retail equipment that wasn't designed for robots at all. Likeā¦. the stuff us humans would use.
Why?
This "works in normal human spaces" capability is the secret sauce that's changing everything, Pras said.
When humanoids can use standard tools in standard environments, automation suddenly works for jobs that were previously "too awkward" to automate.
Try This:
Pick your most manual warehouse task and time exactly how long workers spend bending, lifting, and walking.
Tomorrow, calculate the hourly cost of that labor including benefits and injury risk. Multiply by 24 hours. That's your humanoid ROI calculator.
Email this number to your ops team with the subject line: "Our 2025 automation target."
Beep boop.
2. AI simulations are teaching robots movement without human programming š¤
Robot programming is soooooo 2024.
Pras revealed their grocery-handling demo uses an "AI trained stack that went directly from NVIDIA's simulation environment to the real robot with no other data."
Zero. Human. Programming.
The simulation teaches the robot. The robot just...does it.
(Yeah, weāre still trying to understand that, too. Donāt worry.)
For their customer Schaeffler, they're building digital twins of entire facilities where virtual robots learn before the physical ones arrive.
It's not coding robots anymore. It's about teaching virtual robots that transfer skills to metal bodies.
Try This:
TODAY: Record a 30-second smartphone video of your three most frequently repeated manual tasks from multiple angles. Save these with detailed notes on weight, dimensions, and variables.
When your vendor asks "what tasks do you need automated?" you'll have exact requirements ready while competitors are still writing vague RFPs. The companies with specific motion libraries will deploy humanoids 6-12 months faster.
3. The home robot timeline is surprisingly specific šļø
Robots doing your laundry? There's actually a roadmap.
Pras told us the exact sequence: warehouses now (already happening), retail backrooms next (restocking shelves), then hospitals (moving supplies), finally homes.
The holdup? Not AI brains. Safety protocols.
Unlike self-driving cars that succeed by "not touching anything," humanoids must constantly handle objects. This creates what Pras called the core challenge: safely applying force around unpredictable humans.
Agility currently uses external safety systems in warehouses, but they're building onboard safety for "cooperative safety" within the next year.
Try This:
Conduct this 10-minute humanoid readiness test: Document three tasks robots could handle overnight when humans aren't present (lowest safety risk).
Print facility maps and highlight "robot-only zones" in green, "supervised robot zones" in yellow, and "human-only zones" in red. Take photos of doorways, measure shelf heights, and note weight requirements.
This exact prep work will put you months ahead when evaluating your first humanoid demo in 2025.
ā
Numbers to watch
$ 20 million
Hallidayās $20M funding aims to make AI agents safely operate on blockchain with 'immutable guardrails
Reply