Wild Fyre (11 page)

Read Wild Fyre Online

Authors: Ike Hamill

BOOK: Wild Fyre
5.22Mb size Format: txt, pdf, ePub

Jim was interrupted by the sounds of cellphones. Most people had their phones muted, but they still vibrated on the tables. Ed looked at his as the other people picked up their phones to see the alert. It was an email from Jim. The subject was “Description of node cluster extrapolation technique.” The email contained an attachment which was a document about the technique.

“So you scheduled this email to be sent during your presentation?” Kevin asked.

“No,” Jim said. “Fyre sent it.”

“Who is Flier?” Dale asked.

“It’s Fyre,” Jim said, “F-Y-R-E. That’s the name I’ve given to the system. It’s monitoring my phone. It runs speech recognition on the microphone. It’s not a hundred percent, but it often picks up tasks that I voice and executes them for me. A couple of minutes ago I said I would email you the document. Fyre heard that and sent you all the document.”

“Yeah,” Kevin said with a smirk. Ed looked around the room. They all seemed to be waiting for a punch line, as if Jim’s statement was a joke. It seemed that Ed was the only person in the room who knew that Jim didn’t joke about his projects.
 

“That’s pretty good,” Maco said. “Ask Fyre to email me a picture of a naked woman.”

Jim frowned. “That’s not a good idea,” he said. “Lister, name a domesticated animal.”

“Dog,” Lister said.

“Harry, name a type of hat—baseball cap, bowler...” Jim said.

“Top hat,” Harry said.

“Fyre, please email a picture of a dog wearing a top hat to everyone,” Jim said.

The room was quiet for several seconds. It felt like everyone was holding their breath.

“Fyre,” Jim said with a big pause. “Please… email… everyone…”

He was interrupted by the sounds of the phones. Some smiled, some nodded, and some gaped at their phones. Ed felt the hair on the back of his neck rise as he looked at the picture of an English bulldog wearing a fancy top hat with the Union Jack on the side. The dog had binoculars hanging around its neck.

“That’s pretty good,” Kevin said.

“So your program listened to your command and then sent the email?” Maco said.

“With the help of some public services, yes,” Jim said. “There’s a web service where you can send an audio clip and it will translate it to text. Fyre just deciphered the text into an actionable command. But that’s the least of what she does. The better example was the first one—I said I was going to send you all a document and Fyre figured out the task and executed it for me.”

“She?” Maco asked.

Jim shrugged. “She strikes me as feminine.”

“So instead of reminding you of tasks, it just nags you?” Maco asked.

Dale laughed.

“What other applications do you see for this technology?” Ed asked.

“Endless,” Jim said. “That’s why I wanted to talk to you all. I’ve got the system controlling a lot of my electronics and communications, but I suspect that with the right user interface, it could do much more.”

“You need to give it one useful chore to gain acceptance if you want to market it,” Kevin said.

“Wait,” Ed said. “Aren’t there already a dozen devices out there where you voice a command and it does your bidding? Isn’t this a copycat?”

Jim shook his head.

Maco turned. “Yes and no,” he said. He looked to Jim for permission to explain. “From Jim’s explanation, those devices just respond. This one sees an opportunity to help and it takes action. Right?”

Jim nodded.

“You said it was controlling your communications—what did you mean?” Lister asked.

“I think we’re approaching an overload of notifications,” Jim said. “Think of how many times your phone goes off during the day. You’ve got texts, alerts, emails, and phone calls. You can set your phone to silent, but what if a critical call comes in. On my phone I can specify that I always want to hear the ring if a call is from my mother’s cell. She only calls me from her cell phone if there’s an emergency. Otherwise, she hardly seems to remember to answer it. But Fyre can broker your alerts based on the context. If a message is important, the system makes sure you receive it immediately. Otherwise, it might stay silent until you check.”

“That sounds dangerous. What if you miss something important?” Kevin asked.

Jim shrugged.

“It could err on the side of letting it through,” Lister said.

“Or it could always vibrate. There’s nothing wrong with that,” Harry said.

“Wait,” Maco said, “what’s this running on? There’s not enough processor on a phone to make all these decisions.”

“No, it’s hosted on a server,” Jim said.

“Well there’s a problem then,” Maco said. “I don’t want all my communication sent up to a server to be evaluated. My important email traffic is encrypted and I don’t want anyone spying on who I get calls from.”

“So it won’t work for the security nuts among us,” Dale said.

Harry and Kevin laughed, but Maco looked serious.

“No, it’s not for everyone,” Jim said.

“What kind of server? Where’s it hosted?” Kevin asked.

“Fyre has a half-rack at Demon Den,” Jim said.

Maco whistled and said, “Wow.”

“Demon Den charges a fortune. Where’d you get that kind of money?” Kevin asked. “I’m sorry, forget I asked.”

“No, that’s okay,” Jim said. “Actually, Fyre pays for herself. Fyre was making investment choices for me as she moved my cash into savings. The system picked so well that I split off another account and turned over control. Since then, Fyre has been self-sufficient. In fact, Fyre even manages the contract with the vendor. If the system needs more processors, or network, or storage, she can add resources within her budget.”

“Hold on,” Lister said. “Are you suggesting that you accidentally created a money machine that can pull profits from the stock market?”

“No,” Jim said. “I created a general system to optimize events by predicting trends. I’m sure it wouldn’t scale if it were investing more funds, but for the amounts we’re dealing with, it’s able to find enough revenue to sustain itself.”

“That’s still unfathomable,” Lister said. “I’ve worked for three companies in the last decade who have had dozens and dozens of the brightest mathematical and financial minds who were trying to do no less.”

Jim shrugged.

Maco laughed. “And Einstein was trying to figure out why the train was late when he unraveled the mysteries of the universe.”

“Not exactly,” Lister said.

“Yeah,” Harry said. “Forget about brokering emails and phone calls. How about you share the code that so I can invest better? I lost twenty percent last year.”

“It’s not the code, it’s the database,” Jim said. “And I’m not profiting.”

“What? Why aren’t you profiting?” Kevin asked.
 

Jim paused. “This is not really what I wanted to discuss. Can you guys think of another use case that I can apply to this system?”

Dale leaned over and said something to Harry. Kevin was looking up at the ceiling. Lister and Maco were both looking at their phones.
 

“Hey,” Ed said, leaning to the side to get Kevin’s attention. “I forgot to have you collect phones. Do you think we should do it now?”

“Seems like it might be crucial to the demo, you know?” Kevin said.

The waiter came in right as Dale began a suggestion.

“What about a natural language query tool?” Dale asked. “People have been trying to solve that problem for years. You just ask it a question and it finds the answer and reads it back to you?”

“It doesn’t work that way,” Jim said.

The waiter moved discretely from person to person, taking orders.

“It can interpret and then execute tasks, but it’s not geared to answer general knowledge questions,” Jim said.

“It found the picture of the golden retriever in the top hat,” Dale said.

Ed looked at Dale. Maco asked the question before Ed had a chance.

“You got a golden retriever? My picture was of a little terrier with a top hat and a monacle.”

“Wait, they were different?” Dale asked. He got up and edged around the waiter so he could compare emails with Maco. “Hey, Jim, how come we got different emails? Your program sent a terrier to Maco and I got a golden.”

“Mine was a bulldog,” Ed said.

Jim shrugged. “I’d have to go through the logs.”

“Can you do it from here?” Maco asked.

“Yes,” Jim said. “I can show you the control panel.”

He turned on the projector and opened a browser. He connected to a page of information and clicked through to a status log.

“Here you can see an entry for all the decisions Fyre has made recently. I’ll make this page refresh automatically in the next version. Right now you have to reload to see each new entry.”

“Are those timestamps correct? It’s making decisions each millisecond?” Dale asked.

“More than that,” Jim said. “Let me turn down the resolution to just show decisions based on voice detection.”

Jim clicked the control and the list changed. Now, each line began with the tag, “Voice analysis,” and then a quote with the first few words of Jim’s sentences. It read like a reverse-order transcript of Jim’s demonstration. He scrolled down through the entries until he found the one that began, “Fyre, please email a picture…” He clicked and the page refreshed. At the top, the page read, “Fyre, please email a picture of a dog wearing a top hat to everyone.” Below that, each word was listed with a description of the program’s interpretation.

Jim read aloud as he traced through the entry.

“You can see here that she pulls the command of send email. She begins to construct what the email will contain. I got a lot of this parsing logic from other projects. Okay, here’s where she decides what picture to pull,” Jim said.

“Wait, a few lines above that, why does it say there’s an ambiguity with the command to email?” Kevin asked.

“Because the way I stated it, I wasn’t clear if I wanted one email to all of you, or an individual email to each of you. As long as everyone received an email, my command was satisfied,” Jim said.

Jim kept clicking until he found what he was looking for.

“Okay, here you can see the image query she ran. She created a query for each person. the inputs are ‘dog,’ and ‘top hat.’ It looks like the same query. I’m not sure why the pictures would be different,” Jim said.

“Well all of our pictures come back with that query, some are just lower in the returned results,” Maco said.

“So she’s using logic to sort the returned results as well,” Jim said. “Interesting. I don’t have an interface to show that. I’ve got an idea though. If we just wait a few seconds.”

Kevin looked confused. “What are we waiting for?”

“To see if we’ll get an interface to see how she filters the results,” Jim said.

“Get an interface from whom? You just said you don’t have one,” Kevin said.

Jim refreshed the page. It was the same. He hit refresh several more times and scratched the side of his face.
 

“Jim? Who’s going to give us the interface?” Kevin asked.

“It would be simple to program,” Jim said. “There must be a filter query generated based on the recipient. This control panel would just need an expandable area where it could display that query.” He hit refresh again and the word “filter” appeared. Jim smiled and clicked it. “There—now we can see the filter logic.”

“Wait, Jim, who put that link there?” Kevin asked. He moved forward to the edge of his seat and leaned to get Jim’s attention.

“Pardon?” Jim asked.
 

“You just theorized a button, waited a minute, and the button appeared. Where did it come from?” Kevin asked.

The rest of the group waited to hear the answer to Kevin’s question.

“It was trivial,” Jim said. “So Fyre added it.”

“Your program modified itself to add a button?” Kevin asked.

“No,” Jim said, smiling. “No, no. The program just altered the control panel. It’s simple—she had one more criterion for producing the email and I suggested that I wanted to view that criterion. Fyre only modified the control panel.”

Kevin looked around at the others to see if he was the only person confused. Everyone was focused on Jim.

“But still, you’re saying the program itself made the update?” Kevin asked.

“Yes. That’s not the interesting thing though,” Jim said. He was speaking quickly now. Ed recognized this version of Jim—this was the version that was just barely connected to his surroundings. Most of Jim’s brain was now preoccupied with some other notion. Jim continued, “The interesting thing is that she decided to add filter criteria to the image results. I only built the query logic, and I set it up to return the very first result. Each of you should have received the first result for ‘dog top hat.’ Instead, she took those results and decided to refine the answer.” Jim clicked on the new filter button next to Ed’s name.
 

“Nope, this is bullshit,” Kevin said.

“Kevin...” Ed said.

“No, I know we’re supposed to be respectful and non-threatening to each other, but this is crazy. You’ve got someone coding in the background. Someone’s listening to you talk and they’re trying to convince us you’ve created a sentient application, right? What’s your goal? Do you want us to all invest in your wonderful new technology?”

Jim wasn’t listening. He was too deep into his own thoughts.

“This isn’t about money, Kevin,” Ed said. “Jim just wanted your feedback on his demonstration. I’m sure he would be willing to prove to you that there’s no trickery going on.”

“And how would he do that?” Kevin asked. “We have to communicate with this program, right? It’s a Mechanical Turk.”

“I’m sure it’s not,” Ed said. “Jim wouldn’t do that.”

“You can’t expect me to believe that in order to turn his lights on Jim built a sentient program,” Kevin said.

Dale leaned forward. Jim continued his investigation, oblivious to the debate that was heating up.

“Nobody said it’s sentient. It just does a few things. He said he reused code for the speech-to-text and for the sentence parsing. And there are Business Intelligence tools that let you add dimensions to the reporting on the fly. The control panel up there is really just a pivot table of the command transactions. What we’ve seen is just a clever combination of tools,” Dale said. Lister and Maco were nodding along with him as he spoke.

Other books

The Last Romanov by Dora Levy Mossanen
The Loveliest Dead by Ray Garton
Delayed Penalty by Stahl, Shey
The People of the Eye: Deaf Ethnicity and Ancestry by Harlan Lane, Richard C. Pillard, Ulf Hedberg
Daughter of Texas by Terri Reed
Dante's Poison by Lynne Raimondo
Forbidden by Rachel van Dyken, Kelly Martin, Nadine Millard, Kristin Vayden