In our latest segment, “Adventures of AI,” we bring you hand-picked important news related to Artificial intelligence, Machine learning and much more from around the globe.
In this piece, we will be talking about, the future of speech, Facebook’s new Artificial Intelligence, a 50-year old problem solved by an AI, and artificial electronic skin.
Artificial Intelligence News – The Future of Speech Is Silence!
Researchers at UC Berkeley have devised a method to determine what you’re saying without you saying it. Its called Silent Speech, and it is ready to revolutionize the way each one of us communicates.
To determine Silent Speech, scientists placed multiple electrodes over the facial muscles of the test subject. The scientists then recorded the electrical potential of muscle activities to predict what the subject was trying to say. This process is known as electromyography. Lastly, they transcribed the recorded electric potential into machine-generated speech to produce digitized vocals.
Such an invention is an amazing gift for all the specially-abled people who have difficulty speaking. Furthermore, researchers can also create a device, similar to a Bluetooth headset, allowing two people to communicate in a noisy or crowded environment.
Facebook’s New Artificial Intelligence Hate-Speech Detection
Detecting hate speech online is a difficult task, even for a human. Humanity spans thousands of cultures around the world; each has its dialect, slangs, and local languages. Often, what’s considered hate speech in one geographical location could be slang in another part of the world. It is an enormous challenge to provide a truly global platform that protects billions of users all around the globe, but Facebook is ready for it.
The social media giant announced the launch of two new AI technologies, namely, Reinforced Integrity Optimizer (RIO) and Linformer.
Usually, to accurately detect hate speech, Facebook builds Artificial Intelligence models by testing different neural architectures and selecting data for training. All of it happens offline. The trained AI model then goes online to test how well it detect hate speech.
With the addition of RIO, the process becomes more efficient. It connects online testing with offline AI modelling. RIO allows Artificial Intelligence models to learn and evolve from billions of online examples of regular content and hate speech.
Facebook is confident that they can detect hate speech in the most challenging of environments. With the help of improved Artificial Intelligence models, according to the social media giant, they can take highly accurate actions on hate speech violations as compared to previous models.
Understanding language is an increasingly difficult task and can require significant computational power for a global company like Facebook. Each new day, the demand for computational performance grows at an even higher rate. Here’s where the second AI technology, the “Linformer,” comes in. It is a brand new architecture that streamlines the process of understanding and decoding colloquial languages from all across the globe.
The result is that Facebook can now flag and remove posts, including long videos, instantly. Mark Zuckerberg and Co. claim that the new AI models will keep Facebook as safe as possible. For all our sake, let’s hope he’s right.
Humans Take Years; Artificial Intelligence, Not So Much.
How quickly can you fold a paper? In a few minutes, I’d guess. Now imagine if the paper was a protein molecule, subatomic in size, and made up of unknown chemical compounds? You’re looking at a solid 50 years of scientific research. Well, an AI solved a similar problem recently with mind-bending accuracy in a relatively short period.
DeepMind, a UK based Artificial Intelligence company, founded in 2010 acquired by Google in 2014, made an AI that solved a 50-year old protein folding problem. Known as the Alpha Fold AI engine, it brought its skillset to a prestigious and quite rigorous experiment called the CASP (Critical Assessment of protein structure), which began in 1990.
In this year’s CASP, called the CASP14, the Alpha Fold AI figured out the 3D structure of the protein. The predicted 3D structure of the protein was off by 1.6 angstrom, which is the average width of an atom!
The results were so impressive that the Ewan Birney, a Genomics researcher from European Biological Laboratory, nearly “fell off of his chair” when he came to know about it.
Understanding the 3D structure of proteins brings us one step closer to figuring out cures for diseases like cancer, that kills close to a million people every year worldwide.
One of the reasons why the novel coronavirus can invade human cells is due to a spike protein present on its outer surface. Alpha Fold AI is solving a very fundamental problem of not just humanity, but of every living being. From the results so far, humans can eliminate a ton of unavoidable diseases in a span of decades.
By The Skin of One’s AI
Scientists are hard at work, creating an artificial skin for robots to wear, not only will this new skin look human but feel alive as well. Up until now, advancements in artificial skin have only allowed the sensation of either movement or temperature, but not both of them simultaneously.
Experts at the Pohang University of Science and Technology (POSTECH), South Korea, namely, Professor Unyong Jeong and Dr Insang You have developed a “multi-modal ion electronic skin.” In simple words, they have made an artificial skin that’ll register the sensation of movement and temperature in the processors of the robots.
Human skin is made up of electrolytes. Furthermore, our skin is also breakable as it can be stretched, twisted, or pinched. Additionally, we have electrical receptors on our skin that inform us of nearby temperature. Whenever we feel a sudden drop in temperature, we get goosebumps.
Scientists at POSTECH infused these abilities in the “artificial skin” that they made. It comes with an artificial receptor which senses ambient temperature. “When an index finger touches an electronic skin, the electronic skin detects contact as a temperature change, and when a finger pushes the skin, the back part of the contact area stretches and recognizes it as a movement” explained Dr Insa
The real aim of the new skin will be to restore the neurotransmitters and human tactile receptors in patients that have lost the sensation in their skin due to an accident or an illness.
In the ever-evolving world of Artificial Intelligence, rapid change is the only constant. Let’s all hope that as the AI evolves, it becomes a saviour of mankind and in the process just as human as well.