“Mr. Mundine,” LAD said. “Can you hear me, Mr. Mundine? Please wake up.”
It was possible that the diagnostic had returned a false negative due to corrupted data. LAD triggered the voice command prompt fifteen more times before breaking the loop.
In the absence of direct commands from Mundine, LAD depended on stochastic behavior guidelines to assign and perform tasks. The current situation was not something LAD had been programmed to recognize. LAD needed information to select a course of action.
GPS was still unavailable. The antenna built into LAD’s necklace could transmit and receive on many different radio frequencies, but the only other bodytechs in range—Mundine’s PebbleX wristwatch, MetaboScan belt, and MateMatch ring—supplied no useful data. No other compatible devices responded to outbound pings.
The complete lack of broadband wireless reception suggested that LAD was inside a building. Mundine had installed an offline travel guide before departing Australia, and according to that data source, regular monsoon rains and frequent geological events (current surveys list 130 active volcanoes in Indonesia) led many in this region to use poured concrete for construction. Those locally composited materials often included dielectric insulators which interfered with radio transmissions. Weatherproofed glass windows would also have metallic coatings that deflected any wavelengths shorter than ultraviolet or longer than infrared. And the absence of satellite beacons like GPS implied a corrugated metal roof that scattered incoming signals. Perhaps without realizing it, the builders of this structure had made it a perfect cage for wireless Internet devices like LAD.
After 3,600 milliseconds of fruitless pinging, LAD re-prioritized the voice command UI and began processing input signals from boundary effect pickups in the necklace’s outer coating. It was sometimes possible to determine location characteristics from ambient sounds. The audio analysis software indicated human voices intermingled with music, and the stream included a digital watermark, indicating a television broadcast, but without Internet connectivity, LAD couldn’t look up the station identifier. However, the offline travel guide included Bahasa language translation software, so LAD was able to understand the words being spoken.
“See Indo-pop singing sensations Java Starship in their international cinema debut!” an announcer’s voice said over a bouncy pop music soundtrack. “When a diplomat’s daughter is abducted from a charity concert, and corrupt local authorities do nothing to find her, the boys of Java Starship take matters into their own hands…”
New voices overlapped the recorded audio stream. Audio analysis indicated live human speakers in the room, and LAD adjusted audio filters to emphasize the humans over the television. Based on pitch and rhythm, there were four separate voiceprints, speaking a pidgin of Bahasa and English.
“What are you showing us? What is all this?” said an adult female (Javanese accent, approximate age 35-40 years, label as H1: human voice, first distinct in new database). “Where did you get these things?”
“They’re from work,” said an adult male (Javanese, age 40-45 years, label H2). “A little bonus. You know.”
“(Untranslatable),” said the woman (H1). “You haven’t had a job for months. I know what you do, drinking with those gangsters—”
“You don’t know!” said the man (H2). “And you don’t complain when I pay for our food, our clothes—”
“Hey!” said a female child (13-15 years old, label H3). “That looks like graphene superconductor material. Can I see?”
“Which one?” asked the man (H2). “What are you pointing at?”
LAD took a chance and switched on the pendant’s external status lights. If the girl recognized graphene by sight, she might also know about other technologies—like the Internet.
“The necklace, there. Look, it’s blinking green!” said the girl (H3).
“You like that, Febby?” asked the man (H2). “Okay, here you go.”
LAD’s motion sensors spiked. 2,500 milliseconds later, the entire sensor panel lit up, and galvanic skin response (GSR) signal went positive. The girl must have put on the necklace. LAD’s battery began charging again.
“Cool,” said the girl (H3, assign username Febby).
“How about you, Jaya?” asked the man (H2). “You want something?”
“The wristwatch!” said a male child (14-17 years old, label H4, assign username Jaya). With all the voices cataloged, LAD decided this was likely a family: mother, father, daughter, and son.
“It’s too big for you, Jaya,” said the mother (H1).
“No way!” said the father (H2). LAD heard a clinking noise, metal on metal, likely the PebbleX watch strap being buckled. “Look at that. So fancy!”
“Pa, they have schoolwork to do.”
“It’s Friday, Nindya! They can have a little fun—”
“Arman!” said the mother (H1, assign username Nindya). “I want to talk to you. Children, go upstairs.”
“Yes, Ma,” Jaya and Febby replied in unison.
LAD’s motion sensors registered bouncing. The adults’ voices faded into the background as Febby’s feet slapped against a series of homogeneous hard surfaces (solid concrete, likely stairs). LAD was able to catch another 4,580 milliseconds of conversation before Febby moved too far away.
“…going to get us all killed,” Nindya said. “I can’t believe you brought him here!”
Arman muttered something, then said out loud, “They’ll pay, Nindya. I know what I’m doing…”
LAD kept hoping Febby would go outside the house to play, thus providing an opportunity to scan for nearby wireless networks, but she stayed in her room all day with the window closed. Incoming audio indicated writing (graphite/clay material in lateral contact with cellulose surface), which LAD guessed was the aforementioned schoolwork. There seemed to be an inordinately large amount of it for a 13- to 15-year-old child.
The good news was that Febby’s high GSR made for efficient charging, and LAD was back to 100 percent battery in less than an hour. With power to spare, LAD accelerated main CPU clock speed to maximum and unlocked the pendant’s onboard GPU for digital signal processing. Sound was the only currently available external signal, and LAD had to squeeze as much information out of that limited datastream as possible. The voice command UI package included a passive-sonar module which could be used for rangefinding. LAD loaded that into memory and began building a crude map of the house from echo patterns.
After the family ate a meal—likely dinner, based on internal clock time and local sunset time—LAD heard footsteps heading from the ground floor down a different set of concrete steps, likely into a basement or storm cellar. Febby stayed upstairs in her room. There was no way to adjust the directionality of the necklace microphones, but LAD increased the gain on the incoming audio and utilized all available noise reduction and bandpass filters.
When LAD isolated Willam Mundine’s voiceprint (91 percent confidence), system behavior overrides kicked in, and the Bluetooth radio drivers shot up in priority. As implied by earlier data, and now confirmed, Arman was holding Mundine captive in the basement of this house. But Mundine was too far away, and there was too much interference from the building structure, for a Bluetooth signal to reach Mundine’s bodyNet. The only thing LAD could do was listen.