* This site is based on internet standards not supported by IE 6.
EC Container 5

Illustration: Hospital ER - Part 2 of 2

Hospital ER

The goal of this scenario is to introduce the concepts of efficiency levels, prediction, context space (the ER entry area), and network visibility (cannot see the parking lot).

There are two key characters, Janet - Director of Ecopoesis at a large health care organization and Richard - Janet's musician husband.

The scene takes place in the ER.


Warren Jones, Lana Rubalsky (2010) "Illustration: Hospital ER", wJones Research, January 18, 2010

Parts: << Previous 1 | 2

Janet and Richard talk about Stored purpose Computers

I am a musician and I specifically chose to leave the corporate world nearly a decade ago. So I admit, I first thought it all a bunch of bureaucratic horse mess. The whole idea of Stored purpose computers seemed another case of companies putting money into technology that should have gone into facilities, or people … that is, until I actually listened one day when Janet tried to explain what she did at work.

We were cleaning up the kitchen. Our son Trent was studying in his room and I had just finished by best Charlie Brown teacher imitation, “Wa wo waaaa wa wa wo” just to clarify how much I had understood of what she had said.

But she tried again and said, “All stored purpose systems work on the same basic principal, you define values that define Self for the client organization, i.e. the hospital, the intelligent systems then predict and measures actual values of Self, called Reality. When Reality and Self differ, the system pursues Goal Pursuit Routes to correct the discrepancy. For every definition of Self, also called a Goal, an intelligent system will have at least one Route for each possibility. So if a computer has a Goal of being a a comfortable temperature, it has a route formula for each possibility, i.e. one for “too cold” and one for “too hot”.

“It sounds impossible. How can a computer execute a formula for too cold, or more realistically, busy emergency room?” I asked.

“The system makes it simple through an idea called Context. You create a context like, I am an Emergency Room at Ashtin General Northwest. Then the hospital defines goals for itself like No unregistered patients and No people in dark corridors. So tonight, when an access camera sees a patient approaching the ER from the parking lot, it will predict a future state based upon the motion speed of the guest where Reality won’t match one or more Base Identity value for Self. The agent tracking ER will then process the highest value Routes it can take to resolve the divergence from Self for this and every other predicted divergence. In L3 systems, it does this just for ER. When we are full L4, valuation will take in consideration everything in the system, so if a Doctor is predicted to be needed for a procedure in another department like cardio in twenty minutes, an L4 system would know there is no value in assigning him to a 30 minute procedure in ER.” said Janet.

“But how can a computer pursue a route of no dark corridors? I thought you said the Metacomputer fabric made it possible for Stored purpose computer technology to work on regular computer hardware. I thought computers could only manipulate ones and zeros.” I asked, starting to get more interested.

“Yes, it’s close to standard computer hardware, just hardened so it won’t fry if you spill water on it. Think of an Metacomputer computer’s brain as a big spreadsheet. Each column represents a point in something called a Timespace. The spreadsheet constantly moves right, toward the future. The columns on the right are executed before the columns on the left so at some point each column hits a point in time that is Now and also known as Reality. Once a column has passed Reality it moves to the past and and becomes Memory. Columns to the left of Reality are still in the future. They are in what’s called Plan. To plan and pursue Goal Fulfillment Routes the agent simply places formulas in the spreadsheet.” Janet said.

Continuing, she said, “Imagine this is a pretty big spreadsheet with rows for every goal and object in the hospital. When the parking lot camera sees a patient park in the ER lot, it updates the spreadsheet column for where it thinks that patient will be a minute from now, five minutes from now. At some point it predicts that the patient will be in an entry area with no lights and in the ER without being registered. To ensure that future never happens, the system adds Pursuit formulas to correct the problem in a column on the right.” she said.

“But that means the formula that turns on the light actually comes in contact with Reality and is evaluated by the system before the situation of patient being in a dark area ever occurs … so that means the system actually achieves its Goals before the Goals become visible to you or me?” I asked.

“Well, visible to you.” she said.

“That’s right, you can see plan memory.” That must be weird, being able to know when a patient will walk through a door.” I said.

“The system isn’t always right and predictions are constantly being updated. But the ability to walk up to an Instrument and by touching it view it’s future activity is very useful.” she said.

“But doesn’t that cause a privacy or security problem.” I asked.

“Not at all. Not many people have access to plan memory and all I see is a lift or network address. Unless you have access to the address, you can’t see the data.” she said.

“I think I understand, but I still don’t get how a computer with an Intel or Texas Instruments processor executes a formula that greets a person or brings medication to the ER.” I asked.

“That’s where the lift and network address come in. Lift is nickname for local synthetic file system. The idea that Jones and Rubalsky had was that when we press a key on a keyboard or a grasshopper cuts a leaf, neither of our brains are actually giving detailed instructions to the limb. Instead we all have synthetic file systems that translate push or cut to another part of the brain that expands on the idea until eventually several muscles are moving in coordination with needed fluids and related balance adjustments of other muscles.” she said.
“So you are saying that your hospital is installing a lift as you call it?” I clarified.

“It’s a two step process. The first was to install adaptors for every display and medical instrument. That makes them Metacomputer compatible and also ties them in with our wireless communications. There are even very small adaptors called smartdust, that make the locks in pharma storage compatible. “ she said.

“So you can access anything in any hospital just by browsing a directory? Can you turn on the lights in your office?” I asked with more excitement than was probably warranted.

“Sure. A decade ago, when Metacomputer computers were first being invented by Jones and Rubalsky, they had one programmer named Priest, write stored-program code that could make anything a computer could see look like a file. They then had another programer named Herbert, write smartdust code that enable a computer to turn on a light or measure how much light was in the room. “

“So they then put the two together to enable a computer to think the light on.” I said.

“Precisely. So somewhere in plan memory, the computer inserted the formula /ER/entryway/lights = ON to turn on lights before the patient arrived in the entryway.”

“In a column to the right or in the future.” I added. “The idea that anything in the world could be represented in a synthetic file system is a pretty good mental leap.”

“I guess,” Janet added. “But it wasn’t an entirely new idea. Ritchie, Thompson, and Kernighan, the guys that wrote Unix were probably the first to try to make a filesystem represent any system capability. Jones and Rubalsky just extended the idea for the post Internet world. Did you know that Jones had been an intern at Bell Labs in the 70‘s and reportedly saw a demo of Unix and C?”

“Sort of like Steve Jobs staying in college just long enough to take a course on Typography and then a decade later creates the Mac.”

“Jones was a bit slower … Bell Labs at thirteen, nearly three decades before starting work on Stored purpose computer.

She looked left to suggest we take the conversation out of the kitchen since the dishes were dry and put away.

I followed and heard her say, “I think you are finally starting to understand all this.” After a pause she continued with a tone of restrained emotion, “Thanks for listening.”

I paused not knowing what to say. I was surprised by the depth of her feelings, and the complete lack of anger. I stopped walking for a moment. I knew at that moment, that over the last hour I had opened a slow drain on a pot of soured stew that I hadn’t known existed … a pot that probably wouldn’t have gone sour if my ear has been a little closer to her and not so close to the piano in recent months. So I said, “I’m sorry.”

After a brief pause, and without waiting for a response I asked, “Isn’t this all a little 1960‘s, aligning Self with Reality? Do we really want our computers wearing flowers on their heads and taking donations at the airport?” I remarked with gentle sarcasm and continued, “… and if a computer really doesn’t want to have unregistered patients in the ER how does it know to not just lock the door and call it a day?”

“A couple reasons. The First is that a system as large as ours has other values for Self, about 1.3 billion Base Values coming in active context on any day, including one that states the ER door is always open. “ she said.

“1.3 billion! I exclaimed, that’s a lot of doors and locks and instruments. And the system predicts future values for everything?”

“And selects the highest value paths and measures prediction against actual Experience for every dimension of Reality in every context space and learns from prediction failures by updating Route values and acquiring new Goal pursuit routes … and that’s just core network fabric objects. We are starting to connect with suppliers and the government which will add to our fabric. But the computers are very fast. Jones figured out in 2006 that CPUs had plenty of capacity to process Goals, but the large programs of stored-program architecture was an impediment. He calculated systems were so inefficient that the hundreds of servers his company employed could be replaced by just four. He was so incensed by the computed waste energy and hardware that he just did that.”

Janet continued, “So he called in three companies, Sun, VMWare and NetApp, told them his plans and two years later and a lot of hard work from his staff, two hundred old servers were in the trash. The cost of new servers from Sun was about the cost of maintenance on the hundreds of older computers they replaced.” she said. “The first Metacomputer fabric hosts were based upon the same Sun design.” Ashtin’s legacy data center had thousands of program-storage servers. We replaced them with just two Appliance Hosts, each kept in a separate location.” Janet said.

“So you are saying that the hardware people used for your video games was all that was needed to make smart machines?” I asked.

“No, not at all. Back in 2006 the inventors realized that the AMD and similar Intel CPUs would be able handle symbolic information such as image is patient, patient is in parking lot, patient is near entryway, light is on etc. But he also realized that CPUs weren’t very good at the kind of processing needed to run neural network algorithms on parking lot footage, predict the motion over time of a billion objects in a hospital or calculate the matrix math needed to determine highest value routes for Goal pursuit. At about the same time the head of technology at a company called NVIDIA realized that their graphics processors would someday make excellent platforms for the massively parallel calculations Jones realized stored-purpose computers would need. Almost for years later, Jones and Rubalsky first published their paper documenting a stored-purpose computer, coincidently the same month NVIDIA released the first Fermi GPU processor that would be used to build the first Appliance Host.“

“There were a few other major changes in the hardware that came after the chip manufacturers started to understand stored-purpose design. CPUs and GPUs were combined on chips with shared registers and texture caches, and new memory managers were designed to accommodate fabric time and route management.” said Janet, just before she paused to notice my eyes were glazing over.

“What was the second reason the agents wont just lock the ER doors?” I asked, more to wake myself up than to believe I could actually process another byte of information.

The Second is the system chooses Routes based upon the maximum Value it can achieve for the Purpose it is executing to preserve alignment based upon values and I can assure you there is no value associated with locking out sick people,” she said with much more patience and passion than I deserved. I realized how many times she had probably tried to explain this to me and resolved to muster my limited mental reserves to listen this time.

“So who defines these values, the Financial Director?” I asked.

“Actually, I do, with Board approval of course,” she said with a hint of pride.

More than a little embarrassed by the breadth of my own capacity for ignorance … not knowing that my wife defined the Values for the region’s largest healthcare provider. I tried to reply with more earnest and alert interest, “So can you explain these values to me?”

“Do you remember the healthcare automation conference I attended two years ago?” she asked.

I nodded and remembered a trip to Oakland where I had not seen her much, played at a corner club two nights and spent the better part of a week after the trip digesting cheap chicken and waffles.

Seeing my nod, she continued, “I talked to a lot of CIOs and EcoDirs and realized that most of the first wave hospitals that experienced difficult L4 uplifts had made the mistake of making valuation too complicated. So I came up with the simple concept of four H’s, Hygiene, Health, Happiness and Harmony. The last H was actually inspired by you.”

I didn’t know whether to feel honored that she had somehow thought of me when working on something this important and complicated or even more embarrassed. My facial expression probably showed the mix. I replied, “By me?”

“You remember that club you made me go to watch you play for just twenty minutes when I was dead tired after twelve hours at the convention center?” she asked.

I wasn’t a question a husband was really supposed to answer, so I just gave my best puppy dog look.

“You were playing backup to this singer that was all over the place ... fast then slow then high then low then fast again. But you kept it all in sync, shifting your gaze between the bass player and the singer, aligning harmony and pace. It was the seed that led me to enhance the assistant agents to align resource scheduling with client pacing and to integrate music into the environmental planning, ” she explained.

Janet continued, “Every doctor, nurse, procedure, hospital zone and situation has its own preferred pace and rhythm. Much of the time, the planners only have one or two options for scheduling, so the Harmony valuations don’t make much of a difference, but we’ve found they also correlate well with Happiness valuations. Together they affect which doctors and nurses work together, who is assigned which procedure, when a coffee and donut is placed on a desk, or most interesting, when people take vacation.”

“Why is that interesting?” I asked.

“It seems the agent assistants are nudging employees who might otherwise skip vacations, to take them by facilitating awareness of other employees who might make good travel companions,” Janet said.

“Cupid?” I asked.

“Platonic, I think,” she continued, pausing momentarily for reflection, “either way, Happiness and Harmony valuation is tracking upward which seems to be secondarily nudging up Health valuation for employees and also retention, something else we track.”

“But Happiness, Harmony?” I questioned gently. I couldn’t see a money making operation like Ashtin supporting a planning structure that hadn’t a C for cash or M for money. “A bit too warm and fuzzy, you think?” I added.

She immediately saw my gist and thankfully saved me from saying anything more.

“Oh, there’s plenty of thought about the bottom line, but cost efficiency is built into the system,” she said. “ICs can’t help but make a company money. It’s inherent in their DNA. While I know they are just machines, let yourself briefly think about Metacomputer systems in human terms. They literally feel bad when resources are idle. So they try not to feel bad by keeping people happy and things moving.”

“But don’t the doctors get bothered being ordered around by computers?” I asked.

“It’s not like that. Medicine has always been scheduled, so that’s not new. ” From the doctors perspective, they’ve gone from having to track a cell phone, laptop and various hospital equipment to carrying one very good gadget that can instantly project an image or medical record on any near HD display and takes care of half their paperwork. People already forget that agents solved the e-mail problem. Once simple agents learned to hold coupons for shopping, pass along messages from friends and inquire about the intentions of everyone else, e-mail stopped being the wasteful time sink it had become.” Janet said and continued ...

And agents understand what these doctors are doing in a way never before possible. If a doctor wants to know which other doctors on staff may have faced this issue or which articles might provide more information, the agent will have typically found the answer in the same time the doctor took to frame the question. And agents like to talk to other agents, so if one doctor’s agent can answer another doctor’s question, the answering doctor might not even know he helped until he’s received the Thankyou card … something else agents are really good at. An unlike chatty human assistants, it’s simply a matter of preference whether a doctor’s agent speaks only when spoken to, or always projects data in realtime to his notepad or headgear display.”

“It really can’t be that simple,” I said.

“Who said it’s simple? Those agents are incredibly complex works of science. But for the doctors, they make work easier and faster than any tool they’ve had before,” she said.

“Why is the Metacomputer tech so much faster?” I asked, a minute before my brain caught up and realized Janet had just told me an answer. But she was patient and answered the question again.

“When a doctor meets with a patient, an assistant agent predicts pretty much everything that might happen. The assistant agent’s purpose is to be prepared to assist with every possible need of it’s client.”

“Like Bebo,” I said, Bebo being the name I gave to my musical assistant agent.

“Yes. For the doctors, each agent will process all probable outcomes, as if they will happen, but only processing to the extent that they don’t cause unnecessary waste if possible. So whenever the doctor writes or mentions an indication, procedure, record or images, an agent begins working on it if it can be undone without penalty or cost. “ Janet said.

“Like Radar O’Reilly, on MASH,” I said referring the character of a super efficient Army supply clerk from the 1970s.

“I am sorry, do you see grey here?” Janet said, pointing to her hair, half joking. She continued, “If a doctor starts talking weight with a patient, the agent will retrieve weight history to a near display. If the doctor then wants to review the history either during or after the interview, the agent simply reveals the information that was already there, but hidden,” she said, with more excitement than I had seen in her for weeks.

“Sounds rather bizarre if you ask me … like lot’s of wasted effort.” I suggested.

“There is some waste, “ she said and continued, “but I’ve never seen you complain when your piano starts the perfect accompaniment whenever you need. You don’t think it actually waits for you to say, “hit me” before it computes what you hear?” she said, using a poor imitation of a male voice that I hope was not my own when she said ‘hit’.

“I guess you are right, who cares about wasted compute cycles.” I said.

“Certainly the agents don’t,” she said. “They are simply fulfilling their Purpose. Bots are a bit different in that they waste energy when they appear magically to deliver meds that aren’t needed. Typically, they have another delivery in the same corridor or section so the waste is minimal. They return the unused meds to storage but sometimes there will be waste, especially for critical cases where they are programmed to err on the side of caution and predict everything that doesn’t go beyond a certain price limit. But the agents will work that out. We have a new agent called Statistician, that will watch the performance of all other agents and suggest valuation and route changes when needed.” Janet said, and then paused for a moment giving me a chance to catch up.

Janet started again, “Do you realize that the principles behind Metacomputer technology so far have proven to be identical to the inner working of our brain … Catch!” she said, just after tossing a blue bean bag my way.

I caught it in my right hand.

“Did you feel the bag when it touched your hand?” she asked.

“Of course. I felt it as soon as I saw it reach my hand.” I answered, maybe too quickly.

“Wrong. What if I told you that the laws of chemistry and physics say there was no way your brain could receive and process the feeling of the bag, as soon as it reached your hand?” she said.
“Of course I would believe you … but no I wouldn’t,” I said, instantly contradicting myself.

Janet laughed and replied, “I think you get it, but you just aren’t ready to know what you get.” she said, pausing and smiling. “Your brain works just like the doctors’ agents. It predicted you would feel the sensation of catching the bag when it hit your hand, and you did feel the bag,” she said.

“But you are saying what I felt was some sort of mental construct, an approximation of the real feeling of catching a ball,” I said now getting a bit queasy.

“Yup. And here are a couple other things to think about when your fingers touch the piano tomorrow. Is there really a real feeling? How is it that you’ve always felt each key at the same exact moment you touched it, when you now know that’s not possible? How can it be that sensors in every one of your fingers are different, but every key on the piano feels the same, regardless of which finger touches it?“ she said.

Janet continued, “The answer is that our brains and sensory systems have layers. Low levels hide details from higher levels … your cerebral cortex tells you that it felt a key, but it doesn’t know any of the detailed electrical measures of each finger’s sensors … so everything you feel is a simulated feeling. It’s your synthetic file system, something all intelligent beings have, according Intelligence Theory,“ she said.

“Interesting and sort of freaky. But it sort of makes sense with that show we saw about the troops with lost limbs that still feel pain in arms they no longer have.” I paused and felt a wave of fatigue strike. I confided this with Janet but also let her know that I really had enjoyed the conversation. I felt bad it had taken so long to listen to her, to hear about the wonderful work she was doing in ways beyond pleasantries. I then found energy for one more question before standing and walking up the steps with her, “What’s the oddest thing you have noticed about this new technology?”

She stood, pulled me along toward the stairs and answered while walking. “Remember when you were into Einstein and told me nothing can travel faster than the speed of light?” she asked.

“Sure,” I said.

“Well, when you felt yourself catch the bean bag at the same time you saw it hit your hand, your brain effectively processed both transmission of information and processing of sensory input in the same time it should have taken just for the transmission to reach your eye. So you felt the “catch” faster than the speed of light.”

“So you are saying Metacomputer tech and people are FTL, able to process faster than the speed of light? But you said the brain cheats, by predicting the answer in advance and plugging in synthetic data. So Einstein's laws aren’t really broken and nothing really travels FTL, right?” I said.

“Your understanding of physics is right but cheating or not, Metacomputer systems still have a very real ability to create coordinated action at FTL speed. It won’t matter much when catching bean bags, but imagine when Metacomputer technology replaces stored-program in national systems such as communications, energy, emergency response.” she said.

“As the governments are already planning,” I quickly responded.

“Yes! And when someone tells an Metacomputer system to turn on the country’s lights or redirect Internet traffic, it will happen everywhere it’s supposed to, simultaneously.” she said.

“So no more light wave traveling across the county or state when lights go on or off, as in the movies?” I said, attempting to clarify for myself.
“Who knows? It’s probably not a big deal, but just another variable that needs to be taken into account during the transition from stored-program to stored-purpose systems.” she said.
. . . . .

So that’s how I learned nearly a year ago, a bit about what Janet does, sort of. And indeed it is a bit more complicated than “tweaking”.

Parts: << Previous 1 | 2

EC Container 6