We used artificial intelligence to test the accuracy of our forecasts
Category: robotics/AI – Page 50
Mission · Oak Lab
Before starting Oak Lab, both Sutton and Javed were working at Keen Technologies—an AGI startup founded by legendary game developer John Carmack. They chose to break away to pursue a fundamentally different path toward understanding intelligence. Sutton bluntly describes current deep learning methods as “weak and inefficient,” arguing that today’s AI models are hitting a wall because of how they learn.
Today’s frontier models (like ChatGPT or Claude) are trained on massive, static, pre-collected internet datasets. Sutton argues this is “learning from someone else’s experience.” Because these datasets are frozen, the models cannot independently discover truly new knowledge, adapt in real time, or evaluate their own outputs.
Sutton’s model explicitly shifts away from the “turn-based” prompt-and-response loop of modern LLMs. By running the FC-STOMP cycle continuously on streaming data, Oak Lab expects to build agents that discover optimal, creative survival and problem-solving strategies that completely bypass the limits of human intuition.
OaK architecture discovers temporal abstractions grounded in experience that are both self-verifiable and useful for planning.
Ray Kurzweil on Why We’re Living in the Singularity | EP #261
In this episode, the mates and Steven Kotler sit down with Ray Kurzweil to discuss AGI, the future, and more.
Get access to metatrends 10+ years before anyone else — https://qr.diamandis.com/metatrends.
Ray Kurzweil is an American inventor and futurist best known for his pioneering work in optical character recognition and his predictions regarding the technological singularity.
Peter H. Diamandis, MD, is the Founder of XPRIZE, Singularity University, ZeroG, and A360.
Salim Ismail is the founder of Open ExO, a GP at Exponential Venture Capital/The Organizational Singularity Fund and a sought after global speaker and thought leader.
Dave Blundin is the founder \& GP of Link Ventures.
Engineers develop AI system to speed satellite tracking of wildfires
A new artificial intelligence system developed by West Virginia University engineers could help firefighters respond to wildfires sooner by enabling satellites to detect blazes and automatically adjust their positions for continued monitoring.
Unlike drones and ground-based sensors, satellites can monitor vast areas of the planet without requiring local infrastructure or routine maintenance. WVU researchers Brycen Pearl, Joshua Warner and Hang Woon Lee developed a framework that allows satellites not only to detect wildfires but also to coordinate with one another and adjust their observation schedules as fires spread.
“Wildfires move quickly—as fast as 15–20 mph (24–32 km/h) under the right conditions—and major wildfires can cover hundreds of thousands of acres,” according to Lee, director of the WVU Space Systems Operations Research Laboratory and assistant professor at the WVU Benjamin M. Statler College of Engineering and Mineral Resources.
Testing the limits of what’s possible (and what isn’t) with AI
When can we trust the results we get from AI, and when is learning impossible? Researchers have shown that there are some problems that even the most powerful AI cannot reliably solve, no matter how much data it is given.
The researchers, from the University of Cambridge and the University of California, Santa Barbara, designed “adversarial” mathematical systems to fool any AI algorithm. Like ethical hackers stress-testing a network’s security, these adversarial systems were designed to map out exactly where and why AI prediction breaks down.
Many real-world systems—like those in the oceans, the human brain or robotics—are too complex to describe neatly with equations, so researchers often learn how they behave by using machine learning. But these AI methods don’t always work well, returning unreliable results or poor predictions.
AI tool improves prediction of who will respond to cancer immunotherapy drugs
Cancer immunotherapy drugs known as immune checkpoint inhibitors (ICIs) can be miracle drugs for cancer patients, curing some and turning deadly disease into a manageable chronic condition in others. But these drugs work for only a subset of patients, with few indications why—a knowledge gap that has detrimental effects on patient prognosis, clinical trial recruitment and research that could lead to new therapies.
A new artificial intelligence model called COMPASS, developed by Harvard Medical School researchers and their colleagues, improves prediction of which patients are most likely to respond to ICIs. Using data from patients treated in the past, the model outperformed the best existing approaches by 8.5%. It makes its predictions based on patients’ tumor gene activity and provides a rationale for its output.
If these results are validated in a future clinical trial, COMPASS could lead to better personalized medicine for cancer patients, more efficient trial enrollment for new therapies and new drug targets for researchers to explore.
Secure glass containers for storing chemical waste through laser welding
As the adoption of electric vehicles continues to grow, so does the need for the safe and permanent storage of battery materials and industrial chemical waste. Certain waste streams require disposal in what are known as Category IV landfills, which impose particularly stringent requirements on storage containers. These must simultaneously ensure environmental protection, safe handling and long-term structural integrity.
Glass is a highly promising material for this application: It is exceptionally chemically inert—meaning it reacts with virtually no other substances—making thick-walled glass containers especially well-suited for the permanent containment of hazardous materials. Glass containers are also of particular interest in the context of potential new recycling methods in the future. The stored residual materials do not react with the containers and can be readily recovered from them.
Until now, these glass containers have been manufactured primarily using thermal gas processes. However, these are limited by uncontrolled heat input, high residual stresses and restricted automation potential. Laser welding, on the other hand, enables high processing speeds and shows excellent potential for automation.
New 200Gbps photodetector doubles optical reception capacity for data centers
Korean researchers have developed, for the first time in Korea, a 200Gbps-class photodetector device for use in hyperscale AI data centers and 5G/6G mobile communications infrastructure. The technology enables ultrahigh-speed data reception fast enough to transmit five 5GB full HD movies per second. The results of this study were presented at OECC 2025, held in Sapporo, Japan, and were recently published in Optics Express.
Electronics and Telecommunications Research Institute (ETRI) announced that it has developed a photodetector device capable of processing 200Gbps-class optical signals per channel. A photodetector is a key semiconductor component that converts optical signals into electrical signals and is essential in determining data reception performance in data centers and communication networks.
The photodetector device developed by the researchers simultaneously achieved a bandwidth of 70GHz or higher, high responsivity of 0.75A/W or greater, and dimensions of 0.5mm × 0.4mm. In particular, applying a “rear-lens integrated structure” that monolithically integrates a convex lens made of indium phosphide (InP) on the back of the chip significantly improved optical reception efficiency and alignment convenience. The entire process, from design to fabrication, was implemented using purely domestic technology.