This week - the ethics of rebooting the dead; how biased AIs really are; BCI headset for AR/VR; why we need a robot registry; DNA synthesis generates true random numbers; and more!
More Than A Human
This video shows how these simple prosthetic fingers can greatly improve the life of someone who lost their fingers. Plus this prosthetic looks beautiful.
With our personal digital footprint in form of photos, stories, text and video messages growing every day, one can take all your data, feed it into a computer and generate a digital person that resembles you - your digital avatar that speaks, looks and acts like you. Some people have already done that to bring back their dead love ones. We can already do this but should we?
Here's an interview with Hugh Herr, the head of Biomechatronics group at the MIT Media Lab and "Leader of the Bionic Age" as the TIME magazine described him. Herr spearheads the emerging field of biomechatronics and combining human bodies with advanced prosthetic. A double amputee himself, Herr is responsible for breakthrough advances in bionic limbs that provide greater mobility and new hope to those with physical disabilities.
Galea is a new product from OpenBCI designed to attach to both AR and VR headsets. The device is equipped with sensors intended to measure data from the brain, eyes, heart, skin, and muscles. The collected data will allow researchers and developers to measure “human emotions and facial expressions” including happiness, anxiety, depression, attention span, and interest level—and use it to create more immersive content tailored to the individual.
NVIDIA recently released a new platform called Maxine that makes use of spare AI cores on their GPUs to enhance video calls, like removing the background, upscaling video from a lower resolution to safe bandwidth or to remove noise. It can even translate to other languages in real-time or deepfake you to make you look better or turn into Memoji-like avatar.
A report from the United Nations, Europol and cybersecurity company Trend Micro describes how AI and machine learning can be used by cybercriminals, from deepfakes to social engineering at scale. It also looks into the future and describes how drones can be used for precise attack or hacking into autonomous cars to cause chaos on roads. The full report is available here.
New research has shown how biased image recognition algorithms really are. Researchers examined the labels those algorithms from Google, Amazon and Microsoft assigned to different photos and discovered the differences in which AIs see men and women, with women much more likely to be characterized by their appearance. The study adds to evidence that algorithms do not see the world with mathematical detachment but instead tend to replicate or even amplify historical cultural biases.
A part of UK's recent defence spending hike - £16.5 billion ($21.8B) over four years - will go into cutting edge technologies that can “revolutionize” warfare — implying a major role for artificial intelligence and sensor-laden connected hardware in “forging our military assets into a single network designed to overcome the enemy”, said UK Prime Minister Boris Johnson.
To build an AI that uses words for a reason, researchers from the Georgia Institute of Technology in Atlanta and Facebook AI Research combined techniques from natural-language processing and reinforcement learning to make an AI that can play a text-based multiplayer game with humans. The AI was trained to achieve goals in the game by either doing the actions itself or by interacting with humans. For example, if the AI needed a sword, it could choose to steal one or convince another character to hand one over. The results are a bit blunt (when AI needed a bucket, it said: “Give me that bucket or I’ll feed you to my cat!”) but it shows how AI can learn how to argue and persuade.
With robots like Spot Mini starting to be used in public and with more to come, a question arises who should you contact when the robot does something wrong or at least whom to call to report a malfunction. The proposed solution - a robot registry. "Governments could create national databases that require any companies operating robots in public spaces to report the robot make and model, its purpose, and whom to contact if the robot breaks down or causes problems. To allow anyone to use the database, all public robots would have an easily identifiable marker or model number on their bodies. Think of it as a license plate or pet microchip, but for bots", says Stacey Higginbotham, the author of this article.
Construction is one of the industries where robots can improve productivity. Here we have a robot that can drywall or autonomous vehicles to ferry materials and tools around some large site. We also have robots like Boston Dynamic' Spot that scan worksites to check if everything is being built according to the plans or drones for an automated construction site inspection. And there are more to come - forecasts predict that demand for construction robots will grow about 25% annually through 2023.
Roboat II, MIT's autonomous boat, has successfully completed a three-hour course of Amsterdam‘s canals on its own and returned to the starting point with an error of just 0.17 meters. As a next step, researchers are aiming to build a bigger version of the boat that can carry 4-6 passengers or cargo around Amsterdam's canals.
Anthony Townsend shares a vision of a city in 2040, where autonomous vehicles and other robots outnumber people. Fully autonomous fast buses connect neighbourhoods. Heavy freight cars wait till night to deliver their goods. Also at night, a fleet of cleaning robots go out to clean the streets while other robots get cleaned, charged and prepared for the next day. It is a vision of a flexible, automated city capable of quickly adapting to whatever is happening to it.
The PizzaBot 5000 is a pizza making robot that will spread sauce and cheese on a pizza crust, as well as slice and dispense pepperoni. From there the pizza is removed from the machine, either by a human or by another robot, and placed in an oven for cooking. The robot can put together a pizza in under one minute all day long, and uses sensors and some computer vision for precise ingredient dispensing, which reduces food waste and save restaurants money.
Generating true random numbers is a challenge. These numbers need to be truly random, such that they cannot even be predicted by people with detailed knowledge of the method used to generate them. A recent study shows how the synthesis of DNA molecule can be used to generate perfect random numbers and might be useful in improving data encryption algorithms.
A new paper shows how retrons - mysterious complexes of DNA, RNA, and protein found in some bacteria believed to be a part of the bacterias defence system - can be used for genome editing. Some might see parallels with CRISPR but retrons are not ready yet, in part because the technology hasn’t been made to work in mammalian cells. “It could be that retrons will be as revolutionary as CRISPR has been. But until we understand more about the natural biology and synthetic behavior of retrons, it is difficult to say”, said one of the authors of the paper.
Early in the COVID-19 pandemic, scientists in China uploaded the virus's genetic sequence to genetic databases. A Swiss group then synthesized the entire genome and produced the virus from it—essentially teleporting the virus into their laboratory for study without having to wait for physical samples. With the genome synthesis technology becoming better and better, we will be able to synthesise longer genomes. It will allow us to build new organisms or printing organisms on demand.