This week - DeepMind solves protein folding problem; highlights from Cybathlon; 3D engine in DNA; AI Incidents Database; and more!
More Than A Human
Organised by ETH Zurich, CYBATHLON is a unique championship in which people with physical disabilities compete against each other to complete everyday tasks using state-of-the-art technical assistance systems., grouped into six disciplines - Powered Arm Prosthesis Race, Brain-Computer Interface Race, Powered Exoskeleton Race, Functional Electrical Stimulation Bike Race, Powered Leg Prosthesis Race and Powered Wheelchair Race. This video contains the highlights of the recent CYBATHLON and you can find full livestreams on their YouTube channel.
Gabe Newell, co-founder of Valve, sees brain-computer interfaces as the next big thing. "In the brain-computer stuff, we’re way closer to ‘the Matrix’ than people realize", he said in an interview with IGN. Newell notes that BCI is "going to have a huge impact on the kinds of experiences that we can create for people". Maybe sometime in the future we will get Half-Life BCI before Half-Life 3.
Researchers from Tel Aviv University and the Shamir Medical Center used a form of oxygen therapy to reverse two key indicators of biological aging: telomere length and senescent cells accumulation. A clinical trial involving 35 adults over the age of 64 sought to understand whether a method called Hyperbaric Oxygen Therapy could prevent the deterioration of these two hallmarks of the aging process. Scientists claim to have successfully reversed the biological aging process in a group of elderly adults. At the end of the trial, the scientists reported that the participants’ telomeres had increased in length by an average of 20 per cent, while their senescent cells had been reduced by up to 37 per cent. This is the equivalent to how their bodies were at a cellular level 25 years earlier, the researchers reported.
DeepMind announced that their AlphaFold AI system has solved the protein folding problem. The paper describing AlphaFold in details is not available yet but in this post, the team explains their approach to solve the protein folding problem. This is big news if DeepMind's claims are true. It will unlock new possibilities in health and bioengineering, maybe even pave the way towards the kind of nanotechnology we see only in science fiction right now.
Scientists are calling DeepMind to release AlphaFold's source code to verify DeepMind's claims it solved the protein folding problem. They also would like to see how AlphaFold deals with proteins outside Critical Assessment of Structure Prediction (CASP) database. The CASP Chair, however, said: "Fifty years of listening to false claims about this problem has made me the world's biggest skeptic. But I have looked at these results very carefully ... Clearly, this is just the beginning of what DeepMind and others will achieve with these sorts of approaches."
In this video, Lex Fridman explains what is protein folding problem, why it is important and why DeepMind's announcement is so important. Lex even predicts that there will be Nobel Prize based on work done with AlphaFold.
Run by Partnership on AI, AI Incidents Database records accidents caused by intelligent systems. The goal of the database is to be a place where future researchers and developers can look for past accidents and learn from them to mitigate or avoid repeating the same mistakes. The list of all reported incidents is available here.
The State of AI Report analyses the most interesting developments in AI that happened in 2020 so far. It looks at the technology breakthroughs and new capabilities that opened in 2020; the supply, demand and concentration of AI talent; areas of commercial application for AI and its business impact; regulation of AI, its economic implications and the emerging geopolitics of AI.
Do you have problems designing a robot for a particular terrain? It is not an easy task but thanks to a new system from MIT, it gets easier. You start by telling the system, called RoboGrammar, which robot parts are lying around your shop—wheels, joints, etc. You also tell it what terrain your robot will need to navigate. And RoboGrammar does the rest, generating an optimized structure and control program for your robot. The research on RoboGrammar also revealed something interesting. "Most designs did end up being four-legged in the end," says Allan Zhao, the lead author of the paper. Perhaps manual robot designers were right to gravitate toward quadrupeds all along.
With the quarantines and social distancing forced by the Covid-19 pandemic, people are turning to robots to do the work. One might think we are at the beginning of the robotic revolution. This article, however, points out that despite the growth in robotics, we don't have robots everywhere doing everything. Instead, we are in the middle of an incremental robotic evolution - a trend that will likely accelerate over the next five years, particularly when 5G takes center stage and robotics as a field leaves behind imitation and evolves independently.
Subramanian Sundaram, a biological engineer, suggests there are three broad themes that might provide a roadmap of sorts for future robot hand research: emphasizing tactile sensor capabilities over the lifetime of a robot, addressing longevity of the data that is collected, and perhaps the most difficult, addressing the myriad of sensations that are expected of hands—being able to identify a cold coin dropped into the hand in the dark, for example.
A unique installation showcased last March in London showed how robots can enable architects to build structures that would be extremely hard or maybe even impossible to build for humans alone.
Electrical engineers have found a way to use bacteria to manufacture an up-and-coming two-dimensional material called molybdenum disulfide (MoS2), which can form a sheet just a few atoms thick and holds promise for future electronics.
Darwin R. Reyes describes a new method to measure the effectiveness of drugs tested on organs-on-a-chip - devices capable of imitating the interaction of cells in a specific organ such as the lungs or liver. Instead of light, Reyes' method uses physical properties of cells such as their shape, ability to move, and electrical changes between the interior and exterior of the cell to give a real-time and continuous tracking of cells. This can speed up research on new drugs and basic cell biology research.
You won't run DOOM on it yet but it is interesting to see how you can "render" a cube using just DNA arranged in the correct order in test tubes and light.