Researchers have made an AI that can scan brain waves and determine what humans are thinking. LILA Gazette explains.
By Eden Perkins – 8th grade
For Artificial Intelligence, the future is already happening. At Purdue University, researchers are using artificially created neural networks to decode real neural networks. They did an experiment, scanning three womens’ brains as they watched short videos. They then recorded brain activity and one algorithm learned how their brains would respond to new clips. Another algorithm read their brains and predicted what they would see next. The reconstructions were pixelated, white images, but they could predict what category the image was, such as “bird” or “face.” It had 50% accuracy when predicting the category. If the computer was using data from a different woman’s brain, it could still predict it with 25% accuracy.
Now, what are neural networks? Artificial Neural Networks, often called Neural Networks or “Neural Nets,” are an artificial recreation of human neural networks. Basically, they use artificial neurons that work together to solve a problem. It is entirely based on humans. Artificially made neurons are software modules called “nodes.” At their core, nodes are algorithms that use computing systems to solve mathematical equations. So far, A.I. is not developed enough to have logic like humans do, so they substitute it for hard facts like math. Computers often use functions as they are very straightforward.
For example, let’s say there are three values: a, b, and c. Now let’s say that an input “a” will lead to output “b,” and input “b” will lead to output “c.” Following the order, input “c” will lead to input “a.” These are hard facts. “a” will never lead to “c” as “b” will never lead to “a.” For a computer, this is as simple as 1+1=2.
Answers to math problems
The nodes that were mentioned earlier work together to solve incredibly hard math problems with countless different values and unknown variables. Although A.I. is created by humans, they can come up with answers to math problems that might take months for humans to solve. In some ways, they are more advanced, notably in hard facts such as math. And then there are the subjects in which current A.I. can’t even get near, in the emotional aspect. A.I. is nowhere near as complicated as humans, as they are unable to have feelings and emotions. They can “comfort” people but will never be able to feel their pain. A.I. mostly relies on math rather than common sense and logic.
Artificial neural networks are functions. They start with an input layer, have a hidden layer where the nodes solve the problem, and an output layer.
Scan brain waves
Currently, A.I. can’t “read” minds as we define it, but they can scan brain waves and make educated guesses on what we’re thinking. 31 participants participated in a study where their brain waves were recorded as they looked at hundreds of A.I. generated faces. They were asked to focus on certain features, such as the faces being old or smiling faces. In real time, the participants’ brain waves were sent to a computer, which inferred whether a participant’s brain saw a face that matched the feature the participant was looking for. The computer used this information to estimate what features the participants were looking for. Next, the computer generated a face. The participants then verified if this was the face they were looking for. The accuracy of each face averaged 83%.
A.I. might seem a little like magic. It can solve incredibly hard riddles in less than half the time it would take an average person, scan brain waves to figure out what they’re thinking, and many other things. Humans are entering a new age. Artificial Intelligence may turn into Artificial Life. When humans start to play God, who knows what might come next?