Over the past decade, the demand for more and more advanced forms of artificial intelligence has gone up faster than it ever has before. Especially within a particular area of machine learning called deep learning, where an algorithm is programmed to work autonomously without any need for human intervention. As a result, we have been accomplishing a lot in a very short amount of time because of it. However along with that incredible potential comes an equal opposite amount of danger.
If we want to avoid the problem of technological unemployment we need to formulate an open dialogue in order to better channel the impending momentum. Otherwise suspicion of big tech may continue to widen the already gaping divide within our society.
AI can evade censorship
The University of Maryland has succeeded in fabricating an AI that can effectively evading internet censorship. Until now the AI has only been trialed in developing nations such as India, China and Kazahkstan. Nonetheless the creators of the algorithm proclaim that it marks another a significant victory over censorship for those citizens who advocate freedom of speech.
“With Geneva, we are, for the first time, at a major advantage in the censorship arms race,” said Dave Levin, an assistant professor of computer science at UMD and senior author of the paper. “Geneva represents the first step toward a whole new arms race in which artificial intelligence systems of censors and evaders compete with one another. Ultimately, winning this race means bringing free speech and open communication to millions of users around the world who currently don’t have them.”
Currently an authoritarian regime works to prevent their population from accessing any information deemed forbidden by flagging a particular keyword or domain request, redirecting that traffic toward a verified alternative. The University of Marylands subversive AI appears to work by avoiding that entirely instead modifying data packets sent from the original computer, rearranging them in a way that cannot be detected.
Hypothetically speaking their algorithm may eventually be wielded by anyone in a country that does not allow full internet access. The idea being that it would enable them to find the desired information without crossing over whatever invisible trip line has been activated by the regime.
AI can tell if you’re going to die soon
Experts in health are at a loss for explaining how their new algorithm can determine whether or not a patient will pass away soon. After providing the deep learning neural network many sets of data containing cardiovascular information like ECG, it was able to figure out who was going to pass away with much greater accuracy then a medical professional.
The newly found ability brings to the foreground a very real and serious conversation that needs to be had about the role of AI in health care. For example- how do you inform a patient that they are going to die soon without knowing exactly when or why it’ll happen.
It isn’t hard to imagine how that would lead to an unnecessary amount of anxiety on the their part. However on the other hand, if you let them know they are going to pass away it may allow them to spend whatever time they have left with their family, doing what they love.
AI has learned the laws of quantum mechanics
The rules of quantum mechanics seem to defy common sense in many ways that are counter-intuitive. However on the larger evolutionary scale it is quite logical. Human beings have a limited amount of memory that nature has specifically reserved for more immediate matters of procreation and survival. Therefor we are relatively limited in our ability to comprehend what is not a part of our everyday life.
However, artificial intelligence simply does not have that same instinctual boundary. Therefor it can be programmed to work out the more unfathomable aspects of our reality. With that being said – A research team out of the University of Warwick, Berlin, and Luxembourg, have programed an AI that can predict the geometric wave pattern of probability that characterizes a particular compound.
There are a lot of different ways an element on the periodic table may combine together with another one to form a new compound. You could imagine the AI as a tool for eliminating the entire experimental stage of trial and error. Rather then trying out each shape a scientist will now be able to run a simulation to figure out the shape of a compound before even touching the elements that would compose it, therefor reducing the amount of time required to come up with a new material.
In fact several new materials have already been invented by artificial intelligence, including advanced meta material which are a growing area of inquiry among the scientific community due to their unique ability to transform in response to changes in lighting, humidity and temperature.
At a more esoteric level, quantum computers are already capable of using the quantum realm to perform a calculation that would remain otherwise impossible for an analog computer, at least so google claims – and that is not without controversy. However the main point being – what if an AI could not only communicate throughout the quantum realm but also comprehend the very laws that allow that communication to take place?
The energy required to train AI is rising 7 times faster then it used to
Machine learning depends on providing AI with as much information as possible. The more data that it has access to the more likely it can make an accurate prediction. You could imagine that as a form of progressive triangulation. When the AI is going through the information it is also making note of every correlation that it finds, representing them as a grid of interconnecting nodes within an artificial neural network.
Of course all that processing power does not come without a price. The more quickly an algorithm can sort through the data the more energy it will inevitably consume. Unfortunately that fact is still largely overlooked by the community of advocates that support artificial intelligence. Furthermore, the power required for standard operation appears to be rising 7 times faster then it has before. A byproduct of what they calll Moores law which assumes that the potential for computation will double each year.
However, it is very likely that AI will end up solving the problem it created to begin with. Especially if we direct that processing power toward the optimization of renewable energy. For example, an algorithm has already formulated a better way to concentrate solar power, and the result will likely end up replacing industrial furnaces that melt concrete and other construction materials.