Tech News Review Week 15-21 September 2020, by Stelios Iordanis

Space Engineering

New measurements show moon has hazardous radiation levels

Astronauts would get 200 to 1,000 times more radiation on the moon than what we experience on Earth—or five to 10 times more than passengers on a trans-Atlantic airline flight.

While envisaged space colonization is the definition of Keynesian spending, and it is presented as the only hope for human species survival in the long run, all available options in the solar system, i.e. the Moon and Mars, are literally hellish places to inhabit, and no truly practical solutions are proposed for the myriad of emerging insurmountable problems.



Can Neural Networks Show Imagination? DeepMind Thinks they Can | by Jesus Rodriguez | DataSeries | Sep, 2020 | Medium

Imagination-Augmented Agents for Deep Reinforcement Learning: Paper and Code – CatalyzeX

Imagination-Augmented Agents (I2As), is a novel architecture for deep reinforcement learning Techniques like I2A are still in the very nascent state but can become a key building block of reinforcement learning architectures in which agents are not only able to learn the present but to “imagine” the future.


Google AI Blog: Improving Sparse Training with RigL

Modern deep neural network architectures are often highly redundan], making it possible to remove a significant fraction of connections without harming performance. The sparse neural networks that result are more efficient compared to dense networks, and, in many cases, can significantly decrease all inference times.

RigL, a new algorithm for training sparse neural networks identifies which neurons should be active during training, which helps the optimization process to utilize the most relevant connections and results in better sparse solutions


victordibia/neuralqa: NeuralQA: A Usable Library for Question Answering on Large Datasets with BERT

NeuralQA integrates well with existing infrastructure like ElasticSearch instances and reader models) using a masked language model (MLM) as well as relevant snippets (RelSnip) – a method for condensing large documents into smaller passages that can be speedily processed by a document reader model.

Search engine technology and Natural Language Processing are revolutionized every week by neural networks; for instance, applications in the medical and legal corpora can multiply the productivity of scientists and legal experts hundreds of times.


Provably exact artificial intelligence for nuclear and particle physics

It is fascinating how machine learning inspires physics and how the opposite applies, as well.



Mosa Meat preparing for large-scale cultured meat production

Sufficient meat supply for the entire planet is an unsolved problem, and cultured meat production could be the only viable alternative.


A single blood plasma protein can transfer the anti-aging benefits of exercise between mice

The potential of Gpld1 to treat brain aging, and help to prevent neurodegenerative diseases like Alzheimer’s and Parkinson’s, is an intriguing discovery, indeed.


Tiny, magnetically controlled robots coax neurons to connect | Science News

Tiny robots can operate as nerve cell connectors, bridging gaps between two distinct groups of cells. These microscopic patches may lead to more sophisticated ways to grow networks of nerve cells in the laboratory, and perhaps even illuminate ways to repair severed nerve cells in people.


Can AI and Automation Deliver a COVID-19 Antiviral While It Still Matters? – IEEE Spectrum

The claims of researchers are betting that new tools can cut drug discovery from five years to six months are very conservative and realistic. The other great advantage of AI based drug discovery is that we do not have to rely on endangered herbal species so much in the future.


Researchers Invent $1 Hearing Aid

Medical equipment has an even higher margin than illegal psychotropic substances, but due to open source technology, this is bound to change soon.


Using deep learning to control the unconsciousness level of patients in an anesthetic state

Optimizing the levels of unconsciousness of patients who require anesthesia for a medical procedure using deep reinforcement learning to control may further improve surgery safety and even have other therapeutic applications.



Airbus Just Unveiled Three New Zero-Emission Concept Aircraft

Air transportation has been demonized to an extreme degree lately; hydrogen could change that, in theory, however, broad adoption will be very slow, as liquid hydrogen is the most dangerous fuel (excluding nuclear sources).



Why Boston Dynamics Is Putting Legged Robots in Hospitals – IEEE Spectrum

The human factor must be minimized in hospital environments, and productivity significantly boosted, as shown by the first wave of COVID-19; robotics is essential towards this goal.


Reprogrammable shape morphing of magnetic soft machines

Shape-morphing magnetic soft machines have myriads of applications in minimally invasive medicine, wearable devices, and soft robotics; breakthroughs like this will transform everyday life and economy output in 10 years from now.


Amazon announced Ring’s new indoor security drone: How will Always Home Cam work?

Privacy issues have already been reported for the Alexa system by Amazon, however, concerns that have already emerged about this security drone are exaggerated.