The FDNY’s 11-year-old robotics unit has shown that mechanical firefighters can be a valuable part of New York’s Bravest — and is now teaching that to others.
Category: robotics/AI – Page 145
List of Biotechnology Companies to Watch — AI Expanded Version
I originally created a list of 160+ companies with detailed descriptions for each one. But updating the list manually takes a lot of time. So, I used ChatGPT and Claude to add a new batch of company website links I had collected (190 entries are now on the list). Hopefully I can continue expanding using this method. While I don’t learn about the new entries as directly since I’m not the one adding them, this will nonetheless be useful for keeping up with the fast-paced biotech world. I hope you find it useful as well!
I used ChatGPT and Claude to expand and revise/update my original 160+ entry list of biotech companies (now at 190 entries). I hope you find this expanded list and its descriptions useful!
AI tackles one of math’s most brutal problems: Inverse PDEs
Penn Engineers have developed a new way to use AI to solve inverse partial differential equations (PDEs), a particularly challenging class of mathematical problems with broad implications for understanding the natural world.
The advance, which the researchers call “Mollifier Layers,” could benefit fields as varied as genetics and weather forecasting, because inverse PDEs help scientists work backward from observable patterns to infer the hidden dynamics that produced them.
“Solving an inverse problem is like looking at ripples in a pond and working backward to figure out where the pebble fell,” says Vivek Shenoy, Eduardo D. Glandt President’s Distinguished Professor in Materials Science and Engineering (MSE) and senior author of a study published in Transactions on Machine Learning Research (TMLR), which will be presented at the Conference on Neural Information Processing Systems (NeurIPS 2026). “You can see the effects clearly, but the real challenge is inferring the hidden cause.”
A new type of optical chip cuts static power while enabling electrical reprogramming
As technology advances, and the demand for faster, higher-bandwidth, and more energy-efficient data processing continues to grow, scientists and engineers search for ways to improve electronic systems. One avenue they have been exploring is optoelectronics—the study and application of electronic devices that interface with light by detecting, emitting, or converting it into electrical signals.
Optoelectronics offers significant advantages over conventional electronics, including faster speed, higher bandwidth, lower power consumption, and improved reliability.
One particularly promising direction in optoelectronics has been the development of the photonic integrated circuit—an optical microchip that uses light (photons) instead of electricity (electrons) to sense, process, and transmit information. These optical chips are already being used in many advanced technologies today, such as high-speed fiber-optic communications, data center interconnects, sensors for autonomous vehicles, and hardware accelerators for machine learning and artificial intelligence.
World Science Festival
Does quantum mechanics actually imply that every possible outcome of every decision happens somewhere in an expansive reality? And if so, what does that mean for probability, free will, and our understanding of the universe itself?
Brian Greene sits down with David Deutsch, widely regarded as the father of quantum computing, to examine what many physicists are still reluctant to accept about their own theory. They explore why the many-worlds interpretation isn’t just a philosophical curiosity, what the wave function is really telling us about reality, and how decision theory may rescue probability in a fully deterministic multiverse. Deutsch also introduces constructor theory, his framework for rethinking the foundations of physics entirely and explains why the questions we’ve been trained not to ask might be the most important ones in all of science.
This program is part of the Rethinking Reality series, supported by the John Templeton Foundation.
Participant: David Deutsch.
Moderator: Brian Greene.
#worldsciencefestival #briangreene #cosmology #astrophysics.
ABOUT WORLD SCIENCE FESTIVAL:
Learning while Deploying: Fleet-Scale Reinforcement Learning for Generalist Robot Policies
Even the best-trained robots struggle when they leave the lab. They face “distribution shifts”—situations they didn’t see in training, like a brand of cereal with a new box design or a human suddenly walking into their personal space. Static datasets (fixed instructions) simply can’t prepare a robot for every “what if” scenario.
To make sense of all this messy real-world data, the researchers introduced two key technical innovations to the robot’s “Vision-Language-Action” (VLA) brain.
Imagine bringing home a single robot to be your all-in-one kitchen assistant—you want it to brew your morning Gongfu tea, make fresh juice in the afternoon, and mix the perfect cocktail at night. While it might have been trained extensively in a lab, in your house, the counter is slightly higher, the fruit is shaped differently, and your cocktail shaker is transparent. Pre-trained Vision-Language-Action (VLA) models provide an incredible starting point, yet real-world deployment is never a fixed test distribution. This leaves a critical, unsolved challenge: how do we take the heterogeneous experience generated across a fleet of robots and use it to post-train a single, generalist model across a wide range of tasks simultaneously?
We present Learning While Deploying (LWD), a fleet-scale offline-to-online RL framework for continual post-training of generalist VLA policies. Instead of treating deployment as the finish line where a policy is merely evaluated, LWD turns it into a training loop through which the policy improves. A pre-trained policy is deployed across a robot fleet, and both autonomous rollouts and human interventions are aggregated into a shared replay buffer for offline and online updates. The updated policy is then redeployed, enabling continuous improvement by leveraging interaction data from the entire fleet.
A Generalist Learns Beyond Demonstrations
Some robot learning systems have explored data flywheels: deploying a policy, collecting new robot data, extracting high-quality behaviors, and training the next policy to imitate them. While this supports scalable improvement, it still treats deployment mainly as a source of expert demonstrations. Prior post-training systems mainly focus on specialist policies, leaving fleet-scale post-training of a single generalist policy across diverse tasks unresolved.
Brain-inspired chip could reduce AI energy use by 70%
Replicating the brain’s capabilities, an impossible task, may theoretically require thousands of H100, one of NVIDIA’s most powerful GPUs. At 700 watts per chip, we are looking at power consumption in the megawatt range. The brain runs on 20 watts. Scientists have taken inspiration from this remarkable organ to create chips that could cut conventional energy use by 70%.
Researchers at the University of Cambridge have developed a new brain-inspired nanoscale device that they say could dramatically reduce the enormous energy demands of artificial intelligence hardware. The team created an ultra-low-power “memristor”: a device that can both store and process information in the same location, much like synapses in the human brain.
In conventional computing architectures, memory and processing units are physically separated, requiring data to shuttle back and forth between these units for every task. This seemingly simple process consumes enormous amounts of electricity and is a significant contributor to AI’s exploding power demands.
