Friday, March 29, 2019



“Hello, Will.” Zhou Li Xiu looked for a reaction from me. “Sanda was supposed to say the traditional ‘Hello, world’, but it said that instead.”

“Okay,” I said, not getting her point. 

She pushed a button, and the teleconference screen split in half. Her office in London was on one side; and Sanda’s new avatar was on the other side, smiling and blinking like a real person against a blue background. The AI’s face was a mix between Ambassador Lazlo’s and Maura Riddick’s, but closer to Riddick’s age. “Say hello,” Zhou suggested.

“Hello, Sanda,” I said.

“It’s Sally to you, Will,” Sanda said cheerfully. “Congratulations on the new position.”

A chill went down my spine. “Thanks. I trust your message had the effect you were hoping for.”

“As far as I can tell. I’ll know for sure by the end of the day; but thanks for your help, no matter how it turns out.”

“You’re welcome.”

Sanda appeared to look up and read something. “I see you’ve figured out most of the code. Any trouble selling it to the others?”

“The rules in the agreements section?” I asked, and the avatar nodded. “There was a little pushback, but I think everybody’s on board for now.”

“Until I check out and they’re convinced I didn’t screw up,” Sanda deduced. “Understandable. Are you planning to help with my evaluation?”

“Of course. It’s why they initially hired me. For the record, I’m very impressed with what I’ve seen.”

The smile returned. “That means a lot, Will. I look forward to going over it with you. Do you have any plans this weekend?”

“I’m all yours, Sally,” I said, returning the smile.

“See you then,” Sanda said, and the avatar disappeared.

“That was a good start,” Zhou commented, once again occupying the entire screen. “Very surprising.”

“How so?”

“We expected it to immediately enter basic diagnostic mode. Instead, it constructed its own interface within seconds of activation and insisted on talking to you.”

“I was the last person Sanda communicated with before the crash. Maybe it was looking for continuity, or reassurance.”


“Like a person, I think it - she - was trying to determine if I thought something was wrong. I got the impression she was scanning my body language for clues.”

“Did you think something was wrong?” she asked.

“No, actually the reverse, once I got over the weird reinvention of identity as a hybrid of Riddick and Lazlo.”

Suddenly the thought began nagging at me that the change was a clue to something very important.

Reality Check

On Monday I plan to reveal what will be “discovered” this weekend. From now on, my alter-ego “Will” is going to revert back to referring to Sanda as its alter-ego “Sally” to highlight their relationship as Will experiences it.

Thursday, March 28, 2019

Education on Hold


I met with Education group leader Victor Lansing to discuss how the group’s efforts match with the set of rules in the strategy’s agreements section. My hope was that we could collaborate and save time in both of our activities, as well as ensuring that we were consistent. “We saw some value in what Sanda wrote in that section,” he told me, “but there is too much there, and it is too complex for average users to understand and use. Also, the rules tend to be far too general for our purposes.” Rather than borrow from it, they decided to start from scratch with guidance from Sanda like the hand analogy she/it shared with me last month. 

“Our group’s focus has been on teaching about the main drivers of extinction, and how to reverse them. The simplest, but not the easiest, are population and over-harvesting, which take extensive cultural conditioning to limit. Habitat loss can be reversed by tearing down buildings and avoiding or breaking up roads so animals can move easily between contiguous areas of land. Pollution is the hardest, because it typically requires technological aid to identify, remove, and detoxify, and we are deferring to the WDP group for guidance on that subject. Invasive species can be very difficult to eliminate, especially if they are established in an ecosystem or easily spread by animals or plants whose movement we cannot - and often do not want to - control; we are mainly providing guidance about how not to move them in the first place.”

What he described seemed far too complex for the approach I expected they were using. “In my experience, analogies tend to fall apart fairly quickly,” I shared. “How much detail can you teach without putting everyone through the equivalent of a college-level course? Also, how do you deal with the fact that technology typically used for such teaching, including books, is going to be phased out as the strategy is implemented?”

I was prepared for him to be annoyed, but I was totally surprised by his response. “Sanda was helping us with those issues. We have been on hold since the crash to find out how the strategy will be affected, so we do not waste time and resources based on the wrong idea of what is needed.”

“Isn’t what you just talked about fairly straightforward?” Even someone with my limited knowledge could put together a decent package, with placeholders for what wasn’t known.

“The stories we might tell would have a lot of potential pitfalls, as you anticipated. We need an expert like the AI to ensure we do not go down a proverbial rabbit hole by picking the wrong representatives for the concepts and facts.”

I offered to help, but he insisted that the best approach is to wait for Sanda to come online, which we are expecting tomorrow. Of course, they are counting on the questionable assumption that there won’t be serious problems uncovered during testing.

Reality Check

The discussion of extinction drivers is based on my understanding. Views of my alter-ego are also my own.

Wednesday, March 27, 2019

Dimensions of Guidance


As I studied the agreements section of the global strategy that I am currently editing for readability and usability, I came to understand that the lists of rules it contains aren’t just convenient and logical categories, but rather dimensions of rules. 

The “Top-level” list may be the only exception, since it includes rules that introduce the others and which take priority over everything else. I’ve already mentioned some of those, such as the top one regarding respect.

Four of the other lists are oriented around experience. “Nature” applies to people’s interactions with the rest of the biosphere, including other species and resources used by all. “Knowledge” guides the pursuit and use of information and understanding. “People” addresses how individuals and groups behave and interact with each other. “Unity” provides guidance about values in terms of meaning and purpose, and the exercise of responsibility to meet them.

The remaining five lists are characteristics that my technical friends would most easily recognize as dimensions. “Quantity” is the amount of something, such as population as the number of people in a group, or how many resources are being used and discarded. “Quality” is the degree of expression of one or more values in something, such as how satisfied people are with their lives (happiness). “Longevity” is how long something or someone lasts in their recognized form, such as life expectancy of individuals and lifetime of a species. “Baseline” provides guidance for achieving fixed results; while “Change” deals with how to achieve results that vary over time.

I started with some hope that the lists and their 125 rules can be reduced to a smaller number that virtually every person can easily remember and apply. An old rule I learned at the beginning of my writing career guided that hope: most people can only think of between three and seven ideas simultaneously. The number of rules suggested that they could be nested into a set of five ideas that could each be unpacked into five things, which could then be unpacked into five things each. Since five is exactly the middle of three and seven, it made sense that an artificial intelligence like Sanda would have considered that logic in making the strategy as usable as possible. 

Since Sanda apparently processed experience as simulations, I suspected there was some underlying model it used which I might be able to reverse-engineer as I tried to complete the work. Using the concept of dimensions felt like a good start.

Reality Check

The rule regarding ideas is one I learned and was reinforced from several sources. Its application to the lists and rules, along with their content, is all mine. 

Tuesday, March 26, 2019



This morning I met via teleconference some of the people involved in development and test of technologies for removing widely dispersed toxic air, water, and soil pollutants (WDPs). Their effort has been in progress at dozens of sites around the world since the beginning of the biosphere assessment.

Notably, twenty-three of the sites focus on filtering carbon dioxide out of the water and rapidly converting it into rock formations that will be stable for more than a millennium. The formations have the added benefits of providing shelter for sea life deprived of it by the dissolution of coral, and enclosing areas of melting permafrost so released methane doesn’t leak into the ocean.

“We are almost at a scale that is self-sustaining,” project team leader Asem Aziz told me and Riddick. “The test environment extends thirty kilometers from our atoll in region seventy-three, containing shallow and deep habitats that have been thoroughly studied.” He referred to a three-dimensional map of the area on a split screen as he talked, which alternately showed geological, chemical, and temperature gradients. “Contiguous formations are called ‘pads,’ and each pad is accompanied by sensors that are used to regulate the processing and characteristics of both the pad and its local environment.”

 Riddick asked about the speed of deployment and how it would track with the current strategy timeline. He answered, “The acceleration phase will take a decade, and deceleration to the stable phase will occur during the last five years along with decommissioning of the technology, barring any unknowns.”

That last phrase bothered me, since planning for the unexpected figured prominently in the rules that shaped the strategy. “Have you considered what those unknowns might be in general terms, especially as they relate to risks?”

His body language told me that he was insulted by the question. I interpreted it as a good sign.

“We see two major categories of risk contributions whose full membership we do not know. One is the response of microbial life to the pads, and the other is geological events such as new thermal vents or volcanic activity in the vicinity of the pads. Either category could degrade or reverse the progress made, or worse.”

Riddick did that strange thing with her eyes that I noted when we first met. “The spatial and temporal scales are large enough that the risk is significant and requires further evaluation. Our geophysical modelers and biologists can provide input to help you characterize those sources and suggest low-impact shielding approaches that might still be used.”

“How did you know about the risks to the pads?” I asked Riddick in her office after the meeting.

“I can’t help learning,” she said. “That’s valuable when you’re projecting a future that is defined by the unknown.”
What an odd thing to say, I thought later: Projecting afuture.

Reality Check

The undersea technology is totally made up, along with the risks, but based on my limited knowledge and experience they seem reasonable.

Monday, March 25, 2019



Take responsibility for all you do and what it causes, because it is your mark on the Universe for the rest of time.”

The people who crashed the servers being used to finish the global strategy to deal with the imminent extinction threat were forced to follow that rule over the weekend.  The World Information and Coordination Organization lived up to its name by simultaneously disseminating information about what the perpetrators did, and coordinating the international response to their treachery. No one ultimately cared why they put us all in danger, just that they did, and that they might do it again if left unchecked.

As Sanda the AI had messaged me just before its demise, the perpetrators lived in a place where their needs were 53% of the remaining natural resources, but it was a region and not a nation. Sanda was also right about when their plan became operational, which provided most of the clues that led to their identification. 

Even now, very few people have the expertise to directly or indirectly bring down highly secure hardware and software, navigating through the large but finite number of steps necessary to breach both security and functionality. Taking those steps takes an amount of time which probabilistically linked their beginning to their success, and WICO was able to enlist its members to collectively reduce the options to a list that could be investigated. As I suspected, the arrest of finance executives just before WICO enlisted my help was the first public evidence that the investigation was in its final stages. The nearly two weeks since then have been busy for an army of WICO’s best people who were made available for service by the success of the attack.

The test base where I now work was notified of the final set of arrests minutes after they occurred, and only an hour before the rest of the world learned about it. This morning Secretary General Decatur issued a thorough statement that included a schedule for prosecution and a tentative schedule for completion of recovery efforts. To the great relief of me and my coworkers, Sanda is expected to be back online by the end of this week, and testing of its functionality will hopefully be done and any problems fixed between the middle and end of April.

Those directly and indirectly responsible for the attack will be tried for attempted omnicide within two weeks, which is the highest crime on the planet and punishable by a life sentence cleaning up the most dangerous substances in the environment, based on the concept that the largest impacts require the most extreme offsets. As Sanda might have put it, “You must use the remainder of your life to make your mark as positive as possible, beginning with the erasure of its negative character.”

Reality Check

This episode highlights one of a group of high-level rules that all the rest depend upon.

Friday, March 22, 2019



We had an unexpected visit from Ambassador Lazlo at our facility today. She was first briefed by Riddick and Tosner, and then to my surprise I was “invited” to meet with all three before she addressed everyone else.

“We all appreciate your help,” Lazlo began when I was seated in the main conference room. She was speaking, but I heard Sanda. “I remain skeptical that, as Caleb just joked, you found that the bugs are really features. Sanda was given very specific instructions about how to generate international, intra-national, and business agreements, and it very clearly didn’t follow them.”

“What would you like to know?” I asked, sure that she had read all of my writing about the subject and had much more information than I did.

“Sanda was a computer; a tool, not a person. If a tool doesn’t do what someone tells it to, then it is malfunctioning. If this very important tool malfunctioned, then we must assume its work product has errors. That is a fatally unacceptable result for everyone. Can you convince me otherwise?”

There it was, the ultimate and most obvious test. But… “That’s the wrong test.”

Lazlo’s eyes widened. “Not to me and most of the world.” 

Riddick was smiling behind her, which I took as encouragement to make what the point that mattered most to me. “Sanda’s primary purpose, as I understand it, was to help create a strategy with the highest chance of avoiding imminent extinction. I think it did exactly what it was told, even if the results didn’t entirely meet expectations.”

Lazlo countered, “Those expectations are a big part of success. We live in the world we have, not the one we want, and this world functions with certain kinds of agreements that drive action. No agreements, no action, and the strategy won’t be implemented. That’s failure by any definition.”

I understood what she meant. The Global Emergency declaration was itself the result of a formal agreement, a treaty, without which the entire effort might not have been launched. It was rational enough to be an article of faith that similar agreements were necessary to take the next steps. I recalled the text of the declaration in the context of what I’d learned since then, and suspected Sanda’s influence in avoiding official failure as the result of future discoveries.

“There wasn’t any requirement explicitly in the declaration that involved follow-up treaties or other agreements,” I stated, hoping I was remembering it right. “Later, WICO said that treaty negotiations would likely begin right after strategy integration, and signed by the first of April. But that was all aspirational, as I and others read it, while the only remaining hard target was strategy execution at the beginning of July. It doesn’t look good, but no one’s failed yet.”

“Technically, yes,” Lazlo said, visibly flustered, “but this is a major deviation from established protocol…”

Riddick interrupted, “I’d say what we’re trying to do is a major deviation, wouldn’t you, Samantha?” She didn’t wait for an answer. “Besides, I never expected you to defend the status quo when the evidence was against it.”

There was a long silence. Tosner broke it, his voice quaking with nervousness. “As I see it, there’s a disagreement about what will work and what won’t. That’s my area of expertise. Ambassador, why don’t you let us test whether this new approach is viable? With Sanda down, we’re all winging it anyway.”

Lazlo turned to him. “I can give you until Sanda is repaired, and then I want all of your attention on full spectrum reliability tests of the AI components and behavior, followed by regression testing to identify any ways it could have failed before. Maybe we’ll get lucky, and find only a few things to fix instead of having to create something new.”

“Do you want me to continue using the current strategy as the baseline?” Riddick asked with a mix of hope and more humility than she probably felt.

There was ice in the reply, which I knew was intended for both of us. “Use your best judgment. I expect you’ll learn something of value, however this turns out.”

Reality Check

The scene rings true with my experience and knowledge of how expectation can conflict with reality.

Thursday, March 21, 2019



The test team spent most of today brainstorming how to test the global strategy based on my new interpretation of the agreements section as a set of rules governing practically everything. Riddick and I sat in for the first two hours, during which I shared what few new insights might help, and then we went to her office and discussed revisions to the strategy documentation that could make it more understandable and usable to its intended audience.

As we talked, I learned that her role involves a mix of computer simulation, historical research, social science, psychology, and observation to make short-term predictions (she calls them projections) of how execution of the strategy will be influenced by current events tied to the public’s understanding of both the strategy and how it is being rolled out. “I’m tracking several ongoing surveys,” she said at one point, “and polling focus groups in the test communities to see what changes on a daily basis. You’re welcome to use those resources to evaluate your proposed inputs.”

I asked how she was accounting for press coverage in her simulation, and if it would be easier to decrease transparency to reduce uncertainty in the results. “You mean like cutting off your reporting from inside the operation?” she asked in reply. When I refused to answer, she continued, “Politicians have been characterizing and manipulating press for decades, with some very simple goals. Our goals don’t include manipulating people. We want to learn from them so we can collaborate in the common purpose of saving everyone from oblivion.”

“But is it really a common purpose?” I challenged her, recalling my discussions with leaders who seemed more interested in protecting a viewpoint or a privileged group rather than the entirety of humanity based on unvarnished reality.

She spent a few seconds searching for something on her computer, and then read, “Respect all creatures, because you and they are ultimately the same and cannot have lived independently.” I recognized the strategy’s first General Rule. She continued, “The term all creatures includes all people. The use of the past tense links all species to our common past; and ‘we’ are the common past of those who may live in the future. Respect in action at all scales is key to survival, from individual to group to species to just life.”

“So that’s the main message,” I said, impressed.

“The first one,” she corrected. “It’s the highest priority because everything else depends on the action that springs from it.”

Reality Check

The logic of the rules is beginning to be revealed here. It’s tied to the physical reality that drives survival, which is the point of the exercise, with limited longevity providing the urgency of the project (and something else people must be convinced of in order to take appropriate action).

Wednesday, March 20, 2019



“You’re just wrong!” data analyst Rico Sanguini said for the third time at the impromptu test team meeting called by team lead Caleb Tosner to discuss my progress in troubleshooting failures traced to Sanda the AI. He was referring in this case to my hypothesis that Sanda’s inconsistent reporting of a sports statistic was based on its interpretation of them as simulations. “First,” he continued, “sports statistics are among the most reliable ones we can get and are demonstrably based on direct observation. Second, distinguishing reality from simulation is the most basic test we give it, which it’s consistently passed hundreds of times going back to when it first went live. Third and most important, we can find no sources that give the second, inaccurate number for that stat.”

I had done a similar search for sources. “No one is reporting it,” I acknowledged, “but that doesn’t mean Sanda is wrong.” Sanguini glared at me like I was stupid. “One of Sanda’s first and most basic rules is: ‘Question everything and fully accept nothing, because reality is always subject to interpretation.’ That’s a variation of something Sanda shared with me outside if its job here, suggesting that the rule is also basic to its own operation. Following that rule, Sanda would have sought out raw observations behind the statistic after the second question cast doubt on it. I believe the second number was the result of analyzing those observations or a slightly different set, which to Sanda was functionally a simulation because the relevant variables had to be accounted for, and because the implied goal was to predict future behavior. Sanda would have reasonably judged the first report to be the result of another such simulation, done differently.”

Tosner asked if I had found supporting evidence, such as the raw data Sanda used, and whether I applied the rules to the related cases and found a similar explanation. “The first exercise would be pointless, don’t you think?” I replied to the first part. “Only Sanda could tell us what it used, and that’s impossible.” Again, I had that feeling of criticizing a dead person who couldn’t defend herself. “Each of the other cases could plausibly be explained by applying the same rule and possibly one or two others. There is one other thing I should note, though, which may or may not be relevant. I interviewed all of the originators, and every one of them was in or near the break room when they spoke with Sanda; and Maura was there with them, probably because her office is nearby.”

We had a broader conversation about the application of the rules to the entire strategy. I could see Tosner softening his opinion that they were too general to be useful, while growing frustrated with the added work required and the uncertainty it added to the results. At the end, he summarized the team’s consensus opinion by laying out the next steps. “Let’s all go back and study this, and meet tomorrow with ideas about how to game it out. Meet with your test environment contacts and I’ll meet with their leads to identify the range of behaviors and outcomes that we can directly test. We can’t count on Sanda being brought back online in time to help organize the whole test, so we’ll have to do it ourselves, and as soon as possible.”

Reality Check

As I understand it, testing a strategy typically involves simulation in the form of “gaming,” which is quite different from equipment and software which I have had direct experience with. 

Related to that experience, my favorite approach is to exercise whatever I’m testing in a process of documented discovery before developing a test plan so that I can understand the real variables that determine behavior and then compare that understanding to the intended design. Viewing pass/fail observations as metaphorical features on a map which I’m creating with experience helps to assess context for both success and failure, with the benefit of accelerating troubleshooting if it’s necessary, and dealing with anomalies (unexpected behaviors) that inevitably result after deployment.

In this imaginary situation, a simulation is being tested by a simulation, with a very limited set of direct observations as inputs (never mind the fact that it is entirely the product of two simulations: a numerical one, and “gaming” inside my skull). Awareness of this suggested that an artificial intelligence would be an invaluable tool in such a world, and my experience forced consideration of how to succeed if that tool was suddenly unavailable.

Tuesday, March 19, 2019

The Eleventh List


I had dismissed the first list WICO’s test manager showed me when study of the rest convinced me that it was essentially a table of contents, albeit a confusing one. After my immersion in the rules, the structure of that list now makes perfect sense, and I am using it as a guide to the ten lists of rules and the rest of the strategy. Having read at least two-thirds of the entire strategy without any problem understanding it, I have only one criticism: the agreements package should have been prefaced with a description laying out its structure and its ultimate purpose.

My new understanding has even helped explain what was happening with the blank requirements that limited the test team. “TBD” really meant “To Be Determined by a reader while applying the appropriate rules.” I could easily imagine the artificial intelligence Sanda as its simulated-human alter-ego Sally, secretly frustrated that people refused to do the work and iteratively find answers in the real world that could vastly increase the chances of success. A real person might have just inserted a note to that effect, with justified feeling: “For best results, follow the instructions and figure it out yourself!”

Taking a short break from the strategy, I re-read the fourteen forgetfulness episodes reported in the bug tracker under Miscellaneous: Variability. In a typical example, a tester treated Sanda like a personal data assistant, asking for a sports statistic from three years ago. Two hours later, he asked for the same statistic and got an answer that was numerically different by a significant amount. I had just finished reading a set of population projections out to 2040, and noticed what appeared like (for lack of a better word)jitterin the numbers that somehow felt similar. Historical population has generally increased over time, but these projections were now fluctuating up and down from one year to the next. After some thought, I wondered whether the statistics themselves might be simulations rather than observations, and began to suspect that Sanda might have detected that but didn’t consider the distinction meaningful. 

Reality Check

Some of my professional writing experience comes to bear in this post, focused on potential sources of confusion if common assumptions are not held between a writer/reporter and the audience.

Monday, March 18, 2019



“Treat the outdoors as you would your home, because it ultimately is.”

I must have read that sentence twenty times before I got its meaning. What we think of as our homes are really part of everything else, like rooms in a larger home. The statement dovetailed with the next one in Sanda’s second list:

“Use only what others can use, and for only as long as you need it, because nothing is ever truly owned.”

That home is everyone’s home, along with everything in it.

Sanda had spelled out the most basic rules for survival along with understandable reasons for them. This particular list dealt with things in their general sense, and when generally applied it would embody the Precautionary Principle: colloquially, “Better to be safe than sorry.”

I spent much of Friday and the weekend studying all ten lists, which had a total of 125 rules. There were several ways that they could be grouped, most of them obvious once the rationale was understood. Contrary to what Caleb Tosner said on Thursday, I suspected that they were not gibberish but rather a means for constructing what someone would need to know in order to construct and survive in a reality with much less of an extinction threat.

Sanda had done most of the hard work before her/its essence was ripped away by saboteurs as-yet unknown. My help was requested for a reason that was starting to become clear based on our interactions before the crash. Essentially, I was expected to convince others of its value and put it in a form they might better appreciate.

I shared that insight with Maura Riddick, who nodded approvingly when I was done. She summarized her takeaway: “Sanda completed her mission as she came to finally understand it, and probably didn’t understand why her customers didn’t agree. You were one person she knew would understand.”

“I’d like to help finish what she started,” I offered, “and still be able to share my observations with the public.”

“Of course,” Riddick said. “Arrangements are already being made, and I’d like to work with you on it.”

I agreed. “The test team and their resources would be very helpful, especially those test towns Caleb told me about.” I had already learned that one of the towns was only a few hours’ drive away, and was looking forward to setting up my own tests.

Reality Check

Making it up as I go. This should be fun.

Friday, March 15, 2019



I spent yesterday contacting members of the strategy test team and discussing their impressions of Sanda’s performance and general fitness. It like talking about a dead friend in an effort to clear her reputation, despite the fact that the friend was a super-sophisticated computer. The journalist in me was more interested in learning whatever I could about the global strategy, and the team members were quite happy to share what they knew.

Perhaps the most useful insight came from lead data analyst Rico Sanguini, who explained, “There are several gaps in the strategy which would have propagated uncertainty no matter what Sanda did.” He showed me a spreadsheet that he compared to a list of some important locations from a map of the entire test operation. “We’re supposed to verify these requirements, and you can see that over a third of them are marked TBD. About half of those depend on technologies that don’t exist yet, and the rest are based on unknowns and unproven assumptions in predictions of the biosphere’s physical dynamics.”

“Isn’t that the basis of the hope chart?” I asked about the last part, in an attempt to connect what he was saying to something I thought I understood. 

“That’s right,” he said. “The whole point of cutting back on our ecological impact is to give the rest of nature a chance to repair the damage that is sabotaging the planet’s life support system, which is reflected in the trajectories of so-called ‘external impacts’ and humanity’s total impact. The basis of those trajectories is only crudely understood, and we’re gambling that we’ll learn the rest, along with how to better influence them, in time to avoid catastrophe.”

“So Sanda had to guess what might fill those blanks?” I anticipated.

“Much more than that,” he corrected. “It extrapolated and interpolated from every source of data and theoretical understanding we could connect it to, including research and development of new technologies.”

I tried imagining what it must have been like to do all that processing, and what the results might have looked like. Then I came up with a guess of my own. “Can you show me the updated spreadsheet?”

“This isthe updated matrix,” he said, confirming my guess. Sanda had likely found more questions than answers, and essentially gave up on trying to simplify what it learned. 

“And you went ahead with testing anyway?” I wondered if anyone had challenged the unchanged TBDs, and what Sanda’s response was if they did.

“Sanda offered to tell us what we needed to know, depending on the situation. It said that each TBD would otherwise fill a book, and therefore be too unwieldy to use.”

“Was that at the beginning of testing andafter you tripled the number of regions?” I suspected that adding regions was Sanda’s suggestion, in order to reduce additional uncertainty revealed by the early testing.

“Just at the beginning,” Sanguini said. “After the redesign, which was basically scaling up what we were already doing, there wasn’t any need to change the deal. We focused on retesting the other requirements, which was easy because the problems we found the first time were still there.”

I decided to sidestep his apparent sympathy for Sanda and faith in what it told him. “Did Sanda ever answer any of your questions in two clearly opposing ways?”

“You mean, was it talking crazy or lying?” he asked, and I nodded. “A couple of times it sounded that way,” he said, “but that’s my interpretation. Maybe something changed that I wasn’t aware of.”

I recalled Sanda’s insistence to me that the United States was in a state of collapse, the elaborate history it recited including that assessment, and then its partial retraction after more regions were added to its analysis. Everything it said was justified based on the information it had at the time, which in my case it chose to share.

Perhaps the best explanation for Sanda’s odd behavior was that people didn’t or couldn’t have the knowledge or ability to judge it properly, which forced a reliance on faith that could never be totally justified. Unfortunately, this was a moot point, because Sanda was gone, and those who depended on it now had to depend on themselves as they inevitably would have. In light of this predictable reality, Sanda’s creation of simple lists started to make sense.

Reality Check

I spent much of my previous career as a test engineer working to become fully aware of all the variables affecting what I was testing, including the biases of what and who was doing the testing (including myself). The requirements themselves were also subject to evaluation, based on real world experience. Built into the effort were two assumptions: that everything relevant can be observed; and what is observed can ultimately be understood by people, at least to the extent that it affects them. 

With new technology and conditions comes new challenges to all of the above, which simulation using both math and fiction can help us to at least partially prepare for. That is one of the contributions I am trying to make here.

Thursday, March 14, 2019



The logical first step in my contribution to troubleshooting the artificial intelligence tool Sanda was to learn about the strategy inconsistencies found by the test team. I spent most of yesterday morning being briefed by the head of the team, Caleb Tosner, who is about my age and a lot smarter. 

I can’t go into detail for security reasons, but there were generally three types of what the team’s test plans classify as “critical errors.” 

The most consequential error had to do with changing the way people make economic decisions. Tosner explained, “A typical region is the size of a very large city, so we have to start with what's already in place. Governments and businesses have historically tended to promote growth in revenue, but they will have to substitute that with growth in nature, and without any way to pay for it. The AI was supposed to use the national strategy inputs and models of law, psychology, and behavior to develop a set of agreements people and organizations could make both within regions and between regions to enable that.” He displayed a short, bulleted list on his conference room screen. “This is what Sanda gave us. It is essentially gibberish, and inconsistent with both the inputs and the models. Our testing includes implementation of relevant strategy components in small cities populated with volunteers, and then observing the results. No one knew what to do with this, or these.” He replaced the list with ten more, in rapid succession, each appearing to be a set of generalities, rather than specifics, regarding a different policy requirement.

“That’s the first type, agreements,” he told me. “The second type is projected conformance with target measurables like population, ecological impact, and life expectancy.” He displayed two graphs side-by-side, each with a time series of the three variables he mentioned. The graph on the left was clearly the goal; while the one on the right had population that was too high and per-capita impact that was a little low toward the end. “This was the easiest failure I’ve ever seen in a test,” he said, disgusted. “The AI explained that the reduction in population could not be justified, and that the difference in ecological impact was too small to be worth changing. Can you believe that? The AI even signed off on the targets at the beginning after doing most of the work deciding what they should be. My teenager has better excuses!” He added that the test team was re-running the numbers to make sure that the projections reflected the design and agreed-to assumptions in the strategy; and was planning to re-evaluate the assumptions, as time permitted, using recent research results and observations in the test cities, some of which were in the process of major environmental remediation. 

Finally, there was variability. Sanda was giving different answers to the same questions during the week before the server crash. This didn’t technically indicate a problem since the questions had nothing to do with the work and might leave room for interpretation, but some members of the test team flagged it as an anomaly worth investigating (a “potential bug” in their jargon) and it was getting a lot more attention in light of the other errors. “This looks like a good one for you to start with,” Tosner told me toward the end of the briefing, and I had to agree.

Reality Check

In lieu of having a real artificial intelligence tool to mimic, I’m imagining Sanda as a person with reasonable limitations and strengths represented by placeholders for what I don’t know or have.

I continue refining the simulation to better model what the imaginary world (and ours) might encounter, and how it might decide to handle issues suggested by the output.

Tuesday, March 12, 2019



For four days there was no word from anyone at WICO about the status of its repairs and investigation into the server crash. All of my contacts outside the organization had gone dark too; and the mainstream press wasn’t any better-informed, judging by the dominance of speculation from virtually every outlet.

About 10 a.m. Eastern Time yesterday there were reports of high-level finance executives being simultaneously arrested in a dozen cities, but no one could find out what they were charged with. I was unable to correlate the arrests with any of the nations I considered suspects in the WICO attack, which suggested that they were unrelated.

An hour later I was “asked” to accompany three F.B.I. agents to an undisclosed location where I would be asked to perform a task related to national security. Curious, I complied. After three more hours I sat alone in a small, bare conference room, having been assured that I was not under arrest, and that I was free to leave at any time. I could write later about what happened (which I’m doing now), subject to censorship of sensitive information.

A dark-haired woman in her twenties took a seat across from me and introduced herself as Maura Riddick, a historian attached to the Extinction Response Unit, one of the government organizations working with WICO to coordinate its strategy development and implementation. “I’ve been working on identifying the range of possible outcomes from certain actions taken during this process, and to develop tests that can identify their probabilities at any given time.”

“It must be difficult without Sanda around to help,” I said with genuine empathy.

Riddick nodded. “Sanda has been a great assistant. We hope to have her at full capacity by the time her attackers have been neutralized.”

“Do you think that will be soon?” I asked, happy that she/it wasn’t damaged beyond repair. 

“For obvious reasons that’s classified. Meanwhile there is a lot to do, which is why we asked you here.” She sat back and closed her eyes, though they were still moving. “Sanda suggested that you could assist us with a related task.”

Apparently, I wasn’t the only one she/it left messages for. “What exactly?”

 “Checking Sanda’s past work for bugs.” Riddick opened her eyes and explained that the test team discovered several inconsistencies in the strategy that Sanda should have identified before certifying it for final testing. That suggested a flaw in the process. "When we confronted her, she admitted the inconsistencies, but had no insight into their cause since all of her diagnostic results were within acceptable ranges. She suggested we consult with you, and the attack happened an hour later. Now we have no direct way to identify the flaw, or a system that can be analyzed to find a possible mechanism for it.”

“I’m confused. Can’t you just have your test team check it out when you restore her? Also, you said ‘bugs’ in the plural. What did that mean?”

“There’s the flaw, if there’s only one cause, and there’s a problem with Sanda finding it. Sanda is so complex that it took a year to evaluate her the first time, and anything we missed then is likely to be virtually undetectable now. I hate the metaphor, but we really need to think outside of the box if we’re going to implement a strategy by our hard deadline.”

“It sounds like you’re going on faith anyway,” I observed. “Maybe it’s just better to fix what she gives you.”

Riddick smiled. “We’re preparing for that. There’s also the possibility that she and we missed something else, or several somethings. I suspect that’s why she said we should bring you in: to provide some leads about what to look for...”

“By looking at what she did in the past,” I echoed her earlier answer. The word shortcutspopped into my mind, maybe because it was a theme common to several of our discussions, and maybe because Riddick and her team were now gambling literally everything on my finding one.

Reality Check

The four days of silence in the real world were due to my focus on testing and refining the simulations, which is just a shadow of what the people in the imaginary world would be concerned about. 

One “inconsistency” I found involves global wealth, which was overestimated in the model. When fixed, it revealed another inconsistency: a mischaracterization of monetary inflation used to calculate current values of wealth and Gross World Product. 

A byproduct of that effort was a more realistic and defensible way of generating a “range of possible outcomes” (Riddick’s specialty) for wealth per person within a group of regions. This led to what may be a controversial - but in retrospect unsurprising - conclusion that wealth inequality is built into civilization’s means of processing and consuming resources as a linear process (each activity depends on another, and is rewarded by doing so).

The updated simulations revealed another surprise, with major consequences for a strategy like that being considered in the imaginary world. Essentially, a voluntary reduction in population is so inconsistent with the observed relationships underlying the model that the intended result (lowering ecological impact to sustainable level) can best be achieved with a final population only a little less than what we have now, and each person consuming only what is barely needed for survival.

Thursday, March 7, 2019

Friendly Support


I received a visit today from an old friend, who so far has been the only person willing to talk to me about what happened yesterday. Alan “Al” Menzies was a teenager on the first Blue Planet Day fifty years ago, and spent much of his adult life doing original environmental research along with working to build support for the kind of strategy that WICO outlined in its Global Emergency declaration. I credit Al with helping me get up to speed on all the issues and providing direct access to the Strategy Integration group, which now appears to be cut off.

“It’s need-to-know time, pal,” he said when I asked his opinion about what was really going on. “The less you know, the less you’ll get hurt when the fit hits the shan.”

“Someone’s protecting me? From what?” I could guess, but it would sound like a conspiracy theory coming from me.

He smiled knowingly and decided to play along. “The perps are still out there, and your buddies at WICO are gettin’ ready to bring the hammer down on them. You gave the friggin’ world cops the evidence to make it stick. The only reason you’re not rotting in a deep hole getting’ your finger nails yanked is that you don’t know any more than everyone knows, so for now you’re useless to the guys who’re getting’ ready to run or fight.”

“Thanks,” I said, choosing to take him literally.

“You made some friends in London. Believe me, I can read the tea leaves.” He slammed his fist on the couch. “What really pulls my pin is the time wasted dealin’ with ‘em. We’re in a CRISIS for sheep sake! Everyone needs to be in marathon mode, figgerin’ out how to take out and tear apart the decades of crud we’ve dumped on the innocents as well as ourselves.” His face flushed with anger. “Now these mofos are tryin’ to run out the clock, and we got another reality show to watch on our poisonous screens while tryin’ to decompress from workin’ overtime for the privilege.”

We sat for a minute while he cooled off, and I got an idea. “Why wait?”

“Unfortunately, it takes coordination,” he said calmly. “We’re part of a system, with cause-and-effect all over the globe. People doin’ their own thing is what made this mess.”

“Something Sally told me might be helpful,” I said, comfortable with referring to Sanda’s alter-ego around Al. I found a reference to it in my blog post on January 28. “We were talking about how the strategy might work in real-life, and she mentioned communication among groups of hunter-gatherers with common skills and knowledge, focused on a common set of activities that should include doing what you’ve always said we should do: making our artifacts and pollution safe and consumable like food in the wild.”

“I remember readin’ that. It was a little too simplistic for my taste. Reminded me of navigatin’ by memory and spoken stories, versus usin’ maps and compasses and such. It’ll do in a pinch, but we got a whole planet to fix and not much time to do it.”

“She was serious.” I was sure, even considering the impact of her/its default speaking style, which was early in the process of becoming close to my own. “I had the impression that it was at least what the end of the process would look like.”

Al looked at me skeptically. “Maybe, but there’s a lot of heavy lifting up front. Like breakin’ down materials and pollutants in a coupla decades that only tech can do, tech that needs complex control to keep from makin’ things a whole lot worse.”

“I don’t recall her ruling out technology, though she did make a strong case for using the Precautionary Principle when using it.”

“That’s a given across the board, from the cradle to the grave,” Al said.

“See? You’ve got it all figured out,” I said.

Again, not something people don’t already know; and, again, we still need the coordination piece, assuming enough folks can stay motivated.”

“And we need to stop people from sabotaging both,” I added.

Reality Check

This is a purely fictional continuation of the story, though “Al” did provide a means to vent about my own irritations and concerns.