Solving Hypercomputation using Neural Networks
Alan Turing, arguably the father of modern computer science, laid out the revolutionary idea of a “know-it-all” machine, which he called the “oracle.” He never specified how such a machine could be achieved or if it was even possible. Most computer scientists and logicians deny the idea that an “oracle” could ever exist because it would go against how information and energy flow in the universe. A machine can’t answer every question (think Deep Thought from The Hitchhiker’s Guide to the Galaxy). Computers are essentially restricted by our instructions and codes, meaning they share our blind spots in logic. Can a modern computer define the truth of the paradox “this statement is false?” Gödel already demonstrated that any logical axioms have unprovable statements. Turing himself showed computers built on logic alone won’t be an “oracle.” So are we out of luck?
Well according to two researchers, Emmett Redd and Steven Younger, a super-Turing computer is close to completion. They take note of Hava Siegelmann, a professor of computer science, who stumbled onto a potential solution by accident. She researched neural networks to mimic the human brain, where the output of one can processor act as the input of another. Each input are weighed differently based on influence. It essentially becomes a self-learning machine. Siegelmann’s research revolved around proving the theoretical limits of the neural networks, that is to show that they could never have the full logical capabilities of a conventional Turing machine. She ended up proving the opposite. By weighing the inputs with irrational numbers, the network could perform tasks that most modern logical computational base couldn’t solve, such as driving cars or medical scans.
With Siegelmann on board, Redd and Younger plan to create a chaotic system where each response is sensitive to even the smallest changes. This would create an unpredictable and infinitely variable noise. Turing speculated a connection between randomness and creative intelligence. Famously, Turing once suggested placing a radioactive radium in a computing engine to generate unpredictability. The idea was not met warmly by his superiors. However, even then, he planned to create something similar to the brain, not necessarily the “oracle.” Redd and Younger have similar concerns. Even if their chaotic neural network system ushers forth a new era of computing, it’s not a guarantee the machine would be Turning’s “oracle.” Even if they fail to achieve “oracle” status, they hope to get the people talking; spark a revolution. Just like Turing did almost 80 years ago.
Link to the article (You need a subscription, however)
Link to the article text (This is free for viewing)