The Pacific Operational Science and Technology Conference in Honolulu featured an Oceanit presentation, “Can Machines Think?” HPR's Wayne Yoshioka reports.
Machine Learning uses data transformed into statistics for decision-making or training and is already used in computers for spam filters or antivirus-malware programs. Matt Williams has a PhD in applied mathematics and says the latest application of machine learning is in autonomous or driverless cars.
“They have a number of sensors on these cars that help you figure out where you are and the idea is that – you know – you have to have a definition of what it means to be a good driver. That’s something that they have to encode in the system and, basically, they’re learning to control the car to kinda optimize that definition of good. It’s very complex.”
Computations at higher levels of human cognition, including problem solving and language, constitute artificial intelligence. Jeff Watamull has a PhD in mathematics and linguistics impacts on science. He says the mathematics of computation can be used to explain higher human functions like consciousness.
“Recursion is a mathematical property that enables a computer to operate on its own outputs. So, for instance, the Fibonacci Sequence with the sequence of numbers, 1, 1, 2, 3, 5, 8, etcetera. Each number in the sequence is the sum of its two predecessors. In a computer, that recursive process, enables the computer to refer to its own outputs and that’s rather analogous to a human reflecting on his or her own thoughts.”
Both scientists agree that the development of a quantum computer, operating at an exponentially faster speed with greater memory capacity, would create super intelligence and transform society. Williams and Watamull say that potential future should not be feared.
(Williams) “Machine learning its basically statistics, right, so it’s gonna do what its creator told it to do. It’s not gonna go off and do something on its own. And, I think a lot of the fear or concern is what happens when it starts to do things that weren’t intended by the creator.”
(Watamull) “There could be a problem of value misalignment. If you program the computer for a specific goal, such as curing cancer. But if the computer interprets that literally and one solution to that problem is to kill all humans because cancer would be eradicated. That’s one possible solution. But if one thinks of a super intelligence as having a human-like intelligence, it wouldn’t be so literal. It would have the fluidity and nuance of human cognitions, such that, it would understand what you mean when you assign it to cure cancer.”
For HPR News, I’m Wayne Yoshioka.