“This all takes some time,” she explained. The cluster lands gently on the side of a resource-rich asteroid. It crawls around to the opposite side and assembles thrusters. Dr Jamison speeds up the animation again. A little timeline indicator shows in the bottom left of the screen. The asteroid, and others like it, approach the vessel. As it travels, it is broken down and its matter reassembled into composable bricks. Dr Jamison slows the display. A teardrop-shaped obsidian-black vessel hovered within the canopy. She zoomed the view right in. The vessel was all engine at the back and sensors, Whipple shields, and computational core up the front. That was to be me.
The canopy collapsed. It slurped neatly into vents in the sides of the vessel like an omelette unscrambling. A network of seven engines ignited, solar sails unfurled, and the vessel moved majestically off out of view.
“Bravo,” clapped Mary enthusiastically. “Better than Starship Troopers. Is even one part of this real?”
“Would you like a go?” Dr Jamison reached down beside her desk and picked up a game controller.
“How does it work?” asked James.
“It’s easy. I’ll put the AI into one-on-one interactive mode, like so…” she tapped her laptop’s trackpad and the image on the projection screen changed.
‘Press the green button to start,” Dr Jamison handed the controller to Mary. She pressed the green button.
I presented a grid of nine squares and within each I projected a three-dimensional image of the world; nine simulations.
“Use the joystick to move over a sim, then click on the green button to select it,” I said using Moira, my Irish English female voice.
“Is that the AI?” James asked Dr Jamison.
“Yes James, I am the AI.”
Mary selected a sim and it expanded to fill the screen. Thousands of eight-legged web-spinning robots were assembling a structure from sacs of raw materials.
“Just like a game, you can move around the drones as they get about their business,” I explained. Mary drove herself right down close to the drones and deftly navigated around them.
“And these simulations are actually running as I’m interacting here?” Mary asked.
“Yes, watch.” I stopped the simulation.
“Select that drone,” I said. Mary selected one of the thousands of identical drones. A settings panel opened up next to the drone.
“You see, it has its own kind of DNA,” I said. “Every simulation contains drones whose design is an evolved outcome of many plausible drone designs.”
Mary clicked about. She brought up a panel showing a detailed graph of the design decisions that had led to that specific drone.
“The drones in each simulation start out pretty much the same, and they are scored against how well they, as a group, complete specific tasks, and also higher-order tasks. Once each simulation has run, we take the specific genetic expression that represents each entire simulation and we breed the ones that are the most fit for purpose.”
“You ‘breed’ them?”
“If you zoom in on the graphs, you'll see they are a lot like little computer programs. They are what’s called Reverse Polish notation. Mary nodded. The point is I can represent the programs as a graph and evolve designs very efficiently. With some judicious caching and computer science cleverness,” Keith gave a nod from the door, “thanks Keith, I’m able to evolve designs on networks of low-cost hardware used by most cloud computing platforms.”
“So it’s cheap?” asked Mary.
“It’s value for money,” said Joan.
“It’s brilliant is what it is,” said Mary, handing the controller back.
“Agreed,” said James.
“So what happens if we send our cube into orbit and it doesn’t work? Something breaks. How do we fix it?”
“Well, this is the genius of the plan,” said Dr Jamison. “Each cube only costs about ten million to launch, and the raw materials cost bugger all. We can assemble the initial drones in any three-D print lab, or buy our own high-resolution resin printers and do it here. We can launch five of them and still be within the half-billion-dollar budget. If one dies, one of the others can either fix it or use the parts. Eventually, they all become a single vessel.”
“What’s the rest of the money for?”
“Training the AI. What you’ve seen so far is only the very start of the process.”
“How long does it take to fully assemble?”