Moonshot: AI avatar
A computerized brain with human intuition
In the same spirit that drove the engineers of the Apollo 11 mission to accomplish what had seemed impossible, Raytheon's innovators are working on world-changing technologies that push the limits of what people can do. Here, as part of the series we call "The Next Moonshots," we explore one of the many potential uses of artificial intelligence.
WHAT IT WILL BE
An artificially intelligent co-pilot with the human intuition of an actual, specific person.
"It's Jarvis, if we're being real. Iron Man's Jarvis," said Tayo Adedokun, a Raytheon systems engineer. "Maybe not as personable, (but) it's a very integrated operator assistant."
WHY WE NEED IT
Modern military systems take in and create enormous amounts of data – enough to overwhelm anyone. There's even a name for the phenomenon: Cognitive overload. One way to relieve the burden: a computerized co-pilot that processes data at superhuman speed, makes decisions on its own and even warns someone when their intuition is about to steer them wrong.
"It could say, 'Let's stop, let's backtrack,'" Adedokun said.
WHY IT'S A MOONSHOT
The first thing a computer needs to imitate a person is a brain – or at least something more human than a processor. Adedokun's mentor, former Raytheon engineer James Crowder, for example, built small robots with neurological structures similar to those of cockroaches and octopuses. Before his retirement, he created a computer model of the human prefrontal cortex.
Next, that computer would need huge amounts of biometric data to identify the patterns of a person's thought process – for example, how the person reacts to stressful situations and how emotions factor into their decision-making. Then, of course, there's the question of how to use it ethically – deciding what it can do on its own, and which decisions would remain with actual people. Such a system most likely would start with limited power that increases as people become more comfortable with the concept.
WHAT IT TAKES
Time and data, Adedokun said.
"Does anything need to be invented to research the possibility? No. All we need is funding and time," he said. "It boils down to a bunch of data capture in a bunch of different environments, a lot of operator studies and doing a lot of computational crunching to see if we can build a model accurate enough."
HOW IT WILL CHANGE THE WORLD
It would eliminate the lag between what a human wants a machine to do and what the machine does.
"Human-machine interfacing would be almost seamless. That can bring about a whole new future of how we act and how we're acted upon," Adedokun said. "Any place where a human does something, the human's ability is multiplied by a factor of X. It's basically cloning without cloning."