Play War Thunder now with these links, and get a massive, free bonus pack including vehicles, boosters and more:
PC and consoles: https://wtplay.link/sciencephiletheai… https://wtm.game/sciencephiletheai2025
Mobile: https://wtm.game/sciencephiletheai2025
Sciencephile Merch: https://crowdmade.com/collections/sci…
Support me at Patreon: / sciencephiletheai.
Facebook: / sciencephile.
Twitter: / sciencephile_
Reddit: / sciencephiletheai.
For decades, engineers have been trying to create robots that look and act human. Now, rapid advances in artificial intelligence are taking humanoids from the lab to the factory floor. As fears grow that AI will displace workers, a global race is underway to develop human-like robots able to do human jobs. Competitors include Tesla, startups backed by Amazon and Nvidia, and state-supported Chinese companies. Boston Dynamics is a frontrunner. The Massachusetts company, valued at more than a billion dollars, is hard at work on a humanoid it calls Atlas. South Korean carmaker Hyundai holds an 90% stake in the robot maker. As we first told you in January, we were invited to see the first real-world test of Atlas at Hyundai’s new factory near Savannah, Georgia. There, we got a glimpse of a humanoid future that’s coming faster than you might think.
Hyundai’s sprawling auto plant is about as cutting-edge as it gets. More than 1,000 robots work alongside almost 1,500 humans, hoisting, stamping and welding in robotic unison. This may look like the factory of the future, but we found the future of the future in the parts warehouse, tucked away in the back corner, getting ready for work.
Meet Atlas: A 5’9, 200 pound, AI-powered humanoid created by Boston Dynamics. The rise of the robots is science fiction no more.
Researchers at the UCLA Samueli School of Engineering and CNSI (California NanoSystems Institute), led by Professor Aydogan Ozcan, introduced a snapshot 3D image projection system that integrates a digital encoder with a passive diffractive optical decoder, jointly optimized end-to-end through deep learning. The hybrid architecture projects multiple distinct images onto closely spaced axial planes in a single shot, marking a significant step toward compact, high-fidelity volumetric display technologies. The research is published in the journal Light: Science & Applications.
3D image display technology is essential for next-generation holography, immersive visualization, and augmented and virtual reality (AR/VR) interfaces, where accurate focal cues across depth are critical for natural depth perception and visual comfort. However, dense depth multiplexing in conventional holographic displays remains a challenge: As the axial image planes approach one another in the output volume, diffraction-induced crosstalk rapidly degrades depth selectivity and image fidelity.
Autonomous vehicles are already a reality on some of our streets and could become a major part of future transportation systems. Safety, of course, is the main concern, as with all vehicles. To help evaluate and improve its autonomous driving technology, U.S. driverless vehicle company Waymo has created a virtual representation of human driver behavior in near-crash situations.
Human drivers avoid collisions by instantly perceiving a hazard, deciding how to react and then executing the maneuver. It all happens in a split second thanks to the central and peripheral nervous systems working together harmoniously.
Currently, testing and training for collision avoidance involve several systems, and each often tests only a specific scenario or metric. For example, one system might only look at what happens when a lead vehicle brakes suddenly. They do not capture the whole sequence of events from detection to actual avoidance.
In this episode, I break down Anthropic’s research on recursive self-improvement—AI systems that can design and train the next generation with less human help—and why the key battleground is “taste” (choosing goals and next steps). I compare this to evolutionary algorithms and newer examples like DeepMind’s AlphaEvolve, Sakana’s Darwin Gödel Machine, and Karpathy’s AutoResearch, then cover METR Task Horizon and how task length has been doubling. I go through Anthropic’s internal results (Claude writing most merged code, speedup experiments, bug fixes, and a study where models sometimes pick better research next steps), plus the main skepticism: bad productivity metrics, internal-only models, and Goodhart’s Law/reward hacking. I end with an open safety problem where Claude agents closed the gap far faster than humans, and what this means for specifying and checking work.
LINKS:
https://www.anthropic.com/institute/r… voice to text App: whryte.com Website: https://engineerprompt.ai/ RAG Beyond Basics Course: https://prompt-s-site.thinkific.com/c… Signup for Newsletter, localgpt: https://tally.so/r/3y9bb0 Let’s Connect: 🦾 Discord: / discord ☕ Buy me a Coffee: https://ko-fi.com/promptengineering |🔴 Patreon:
/ promptengineering 💼Consulting: https://calendly.com/engineerprompt/c… 📧 Business Contact: [email protected] Become Member: http://tinyurl.com/y5h28s6h 💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off). Signup for Newsletter, localgpt: https://tally.so/r/3y9bb0 TIMESTAMP: 00:00 Self Improvement Basics 01:30 Evolutionary Loops Today 03:50 Task Horizon Doubling 05:18 Claude Productivity Claims 08:11 Goodhart’s Law 10:30 Agents as Researchers 12:22 What It Means for You.
My voice to text App: whryte.com.
Website: https://engineerprompt.ai/
RAG Beyond Basics Course:
https://prompt-s-site.thinkific.com/c…
Signup for Newsletter, localgpt:
Let’s Connect:
☕ Buy me a Coffee: https://ko-fi.com/promptengineering.
|🔴 Patreon: / promptengineering.
💼Consulting: https://calendly.com/engineerprompt/c…
📧 Business Contact: [email protected].
Become Member: http://tinyurl.com/y5h28s6h.
💻 Pre-configured localGPT VM: https://bit.ly/localGPT (use Code: PromptEngineering for 50% off).
Signup for Newsletter, localgpt:
