That dreaded “just after the middle of the month” point, where the initial thrill is gone, the “big push to the finish line at the end of the month” is still over the horizon, and even that little “middle of the month rededication bump” has crapped out. Success lies balanced on the knife edge between “banzai!” and “I’ve had it.”
And I think last night I blew it and labeled Day Eighteen’s work as Chapter Nine. It’s actually the beginning of Chapter Ten. It’s tough getting good help. Once we move out beyond my ability to count on my fingers, we’re really going to be in trouble.
While I normally put in a lot of internal links to previous, related posts here, I won’t be doing that for what I hope will be this year’s thirty NaNoWriMo posts. If you have jumped into or stumbled onto this story in mid-adventure, there are plenty of other ways to navigate around the site to find previous installments. Actually doing so is left as an exercise to the student.
CHAPTER TEN (continued)
Crystal had to admit, the setup and the work being done at Homolacrum was impressive. As a consultant and a prominent columnist in the computer and data management worlds for over two decades, she had seen most everything. She had also hacked into most everything.
“Let’s see what this button does,” she told the small group of assembled scientists and software engineers, many of whom hadn’t been born when she was doing her doctorate work.
“Please don’t,” Pete asked.
“Then can someone explain what we’re looking at here?”
“This is a real-time schematic of the CPU and storage being used by Sherman,” said the technician giving the demonstration. “While we of course have automated procedures which monitor the system state, we have found that with a proper visualization matrix, a trained observer does very well in looking for patterns and structure in the output. These patterns and structures in turn can give us insight into the connections between data sets that Sherman is making as he gains experience in dealing with the real world.”
“You make it sound like he’s actually alive,” Crystal said. “What you’re describing is learning, isn’t it?”
“Only in the broadest terms,” the technician said. She manipulated the image hanging in the air above the conference room table, freezing it and zooming in on one section. “You see here, we saw this over the last week. These are the various data sets regarding different animals and biology in general. Each of these nodules represents a class, while each of the different dimples and specks within the nodules are different species.”
“Is this something that Sherman worked out or is it something pre-programmed in?”
“A bit of both at this point.” Again she zoomed in on one of the nodules. “Here for example you can see a representation of all of the data on canines. It’s fractal in nature, so we can keep zooming in and keep getting more and more structures. At this level, the individual kernels you see are different breeds. This is all just raw data which we started with in Sherman’s data library. It gets constantly updated and expanded, but it’s just data.”
“How much data are we talking about?” Crystal asked.
“The exact amount would be proprietary information, but I can say that it’s well in excess of a zettabyte.”
“Okay, so you primed the pump with a gazillion tons of data. Hopefully it’s been carefully verified, cross referenced, and doesn’t include Wikipedia. Where does Sherman do what others can’t?”
“Here,” the technician said, adjusting the image to highlight a web of fine lines. “These are the points where Sherman is gaining enough experience to draw new conclusions from existing data. Let me show you. Sherman?”
“Yes, Lois?” answered Sherman.
“Please show a dog.” A video of a Golden Retriever puppy playing with a toy appeared.
“Show a dog that I would like.” The retriever was replaced with a picture of a toy poodle.
“Show a dog that Soichi would like.” An English bulldog appeared.
“Show a dog that Pete would like.” A large, Persian cat hung over the table.
“Not to be critical, but that’s not a dog,” said Crystal.
“No, because Sherman knows that Pete’s highly allergic to dogs. This demonstrates in part how Sherman can cross reference data between vastly different databases, given the proper permissions, of course. It also is a good example of Sherman’s functional snark and humor parameters. The various personality parameters are tuned to each individual using Sherman, and as Sherman gains experience with a given user, it ‘gets to know them’ in a sense and can extrapolate for other situations.”
“Thanks for the heads up on Pete and dogs.” Crystal turned to look at Pete with a hint of a wicked smile. “That must be why you were sneezing all night, dander or hair from my dogs carried over on my clothing.” She turned back to the group, please to see that Pete was turning a suitable shade of red next to her.
“How many personality parameters are there?” Crystal asked.
“Seven primary, but a given user can define others and teach Sherman what they are and how they are expressed. Each user creates their own virtual Sherman. In the early days of personal electronic devices, users had the option to change the synthesized voice from male to female or to vary the accent or language. This is like that, but several orders of magnitude more.”
“What are the seven primary personality parameters that Peter Piper picks?” asked Crystal.
“Snark, humor, empathy, responsibility, maturity, aggressiveness, and nagging,” said Soichi.
“Ah, I see, S.H.E.R.M.A.N. Very clever. But snark? Humor? You programmed your system to talk back and give you grief? If that doesn’t sound like a prescription for the apocalypse, I don’t know what does.”
“You’re not the first to suggest that,” Pete said.
“The key is in the ability for each user to set those parameters and to define circumstances for each set to be used,” said Soichi. “For example, I prefer my default setting to be serious and factual, where I believe Pete prefers a default that is more playful and funny. However, I have my Sherman trained to be very bland and robotic when in a social or business setting such as this, as to most of us. Sherman can change modes autonomously as circumstances warrant, such as when someone else enters the room.”
“I see,” said Crystal. “Sherman, tell me a joke.”
“I’m sorry, Ms. Reasoner, but I do not yet have permission to allow you access to the system,” said Sherman.
“Let me show you,” said Pete. “While you can give explicit instructions to change a given parameter, most often the changes are keyed to a phrase. Sherman, can you give me the weather for Los Angeles today?”
“Of course, Pete, it’s currently 78° in Venice, 82° in downtown, and 89° in Sherman Oaks. It’s clear with no rain in the forecast. Would you like more detailed or additional information?”
“No thanks, Sherman. Why don’t you be yourself for a while? What’s the weather in San Francisco?”
“What do you care?” said Sherman. “You aren’t going there any time soon, and even if you did, you probably wouldn’t go outside and enjoy the beautiful 68° temps and evening fog.”
“Get serious, Sherman. What’s the weather in New York City for next week?”
“It’s predicted to only be in the upper fifties at the beginning of the week, with showers and possible snow flurries coming in toward the end of the week.”
“Thank you, Sherman. As you were.” Pete looked at Crystal. “See?”
“Oh, if you can’t have fun in San Francisco, we’re going to have issues,” said Crystal. “I am going to…never mind. Doctor Doi, earlier Lois mentioned permissions within a group. Do the Shermans talk amongst themselves?”
“First of all, while it may appear that there are multiple Shermans being active, in fact it’s just one system. It’s extremely nimble and competent in its ability to monitor dozens of individuals in a location and to be different things to different users, depending on their needs. However, that is only based on training and a large accumulated data set which defines each user’s Sherman to their needs. The permissions arise in determining which data can be used and accessed for each individual interface transaction. For each task, question, or command, Sherman can only use private data from other Sherman data sets where it has permission. In an office environment such as this, there is naturally a great deal of exchange in terms of professional interactions and a fair number of social interactions. To a stranger on the street, Sherman would say much, much less.”
“This is all fascinating, and I would love to learn more,” said Crystal. “How can I get a Sherman to use for evaluation purposes?”
(Chapter Ten will be continued)