Squonk’s Tears

Strategy notes for technology ventures during a “pseudo-jackpot” period subsequent to the COVID-19 pandemic and its economic fallout.

The following is a collection of personal notes. Most of these points compare personal reflections from previous economic crises w.r.t. the current pandemic. This is especially targeted at questions about leading an early stage technology venture.

Our world has changed. On the other side of a global crisis, where do we land? As an enormous range of possible risks and opportunities open up, which strategies are indicated? How do we prepare for a so-called “new normal” and its emerging business environments?

The term “jackpot” above makes reference to The Peripheral series of novels by William Gibson. The author describes vividly about an American community in a dystopian near-future where wars, pandemics, and other not-quite-cataclysmic events have killed much of the world’s population. Interestingly, these novels explore narratives and personal interactions across multiple futures which are interconnected.

How to position effectively

Based on earlier instances of global crises — to the extent that we may even make reasonable comparisons to current situations — we will likely experience some period of relative chaos, from which emerge new paradigms for economics, social norms, etc. The next “preeminent” paradigms we leverage will probably not become apparent for at least a few years, even among their current practitioners.

Consequently, a good approach for finding the leading signals is to sample others’ responses and experiences. Think of it as “crowdsourcing” to create a kind of rear-view mirror lens or suspended syntax used to examine what has happened and identify what is emerging. Attempting many strategies as an individual or even as a company would be impossible. Sampling from the experiences of others can lend an advantage. Look among your clientele, vendors, and other associates. Leverage each as a lens into how to respond more effectively.

In particular, work to identify maladaptive behaviors w.r.t. sea changes in progress. The people and organizations responding poorly are inherently simpler to measure. Look carefully at those who are using legacy approaches to respond, beating their dead horses into glue. Then, as a priority, impose distance from any and all bad actors. The remaining clients, vendors, and partners become your “region of optimality” with whom you build your subsequent business activities.

Any minor world that breaks apart falls together again
When the demon is at your door
In the morning it won’t be there no more
– Steely Dan, 1974

Of course this thinking contradicts normal business adages that “The customer is always right” or “Don’t burn bridges” — that’s why we’re considering a new normal. Now is an optimal time to burn bridges. Many people who were considered experts in late 2019 won’t have a podium or even an audience by early 2021. It’s an exceptionally bad idea in the current situation to maintain partnerships with individuals and organizations which are taking on existential risks. Not that I’m eschewing empathy and compassion: help people out where you have means to do so, but don’t you align your business futures with bad bets. Some very large organizations will go under suddenly. Many social norms will change forever. The risks in business now seem terrifyingly high.

If you have enough liquid assets, don’t rush: take time to get a good reading of the situation and formulate your strategy before reacting.

Social media can be useful as a lens+firewall: work on curation, aggressively unfriending and blocking the people who are noise (who become dangerous now), muting discussions that get derailed, making lists (e.g., on Twitter) to focus on those who are showing potential insights.

This initial chaotic period is when any given venture must practice operational closure (see Maturana and Varela, 1980) to create that kind of “firewall” effect: filter out the noise and the bad actors, focus less on external environment and more on priorities determined by your internal structure.

Earlier instances of pseudo-jackpot

We can reflect on earlier periods where global crises created economic chaos. Each resulted in transitions where technology adoption had enormous effect on the world overall. The aggregate events, responses, and their subsequent effects tend to follow a J-curve for GPT adoption, generally with a ~15 year expected value for latency. One may argue that the following discussions are more about “spillover” effects than actual GPTs, which is a fair point.

Translated: nature abhors a vacuum, so there’s room for the more recently introduced technologies— or older technologies that never quite gained traction — to advance and become preeminent.

In abstract:

T: new technology introduced
T+∂: large-scale external crisis
=> followed by paradigm shift
=> approx. T+15: new practices go mainstream

During the chaotic period which typically follows crisis, any paradigm shift that obtains will tend to have only limited recognition. The length of ∂ and the magnitude of the external crisis have implications on the extent and duration of the chaotic period.

Relatively recent examples of these transitions include:

1983: launch of TCP/IP
1987: Black Monday
=> followed by introduction of world-wide web
=> 1995 mainstream adoption of Internet use (leading to Dot Com Boom)

1997: inflection point for Amazon, eBay, Google, etc.
2000: Dot Com Bust (leading into 9/11)
=> followed by sea-change of ML use cases (see Breiman, 2001)
=> 2015 enterprise adoption of ML

200x: YouTube, eLance, and other origins of “gig economy”
2008: Global Financial Crisis
=> followed by disruption of traditional employer-employee relationship
=> ~2024 ??? new practices

201x: ??? new technology introduced
2020: COVID-19 pandemic and economic crash
=> followed by ??? paradigm shift
=> ~203x ??? new practices

Keep in mind that causal events and responses may overlap among the ongoing transitions. For example, the narrative arc in response to the 2008 market crash will be affected by events from the 2020 global pandemic.

Another interesting outcome from the 2008/2009 crash was how that time period marked the introduction of data science practices into industry. Clearly there had been earlier definitions: Cleveland’s suggested curriculum 2001 which ties into Breiman’s work noted above, and Tukey’s overall introduction of empirical data analytics in 1962. However, the 2009 date is when Patil (LinkedIn) and Hammerbacher (Facebook) and others (myself included) began being cited as leading data science teams. Most of us had those teams and practices in place by mid-2008. Also note that governments began instituting data science practices — EU, US, etc. — specifically in response to the 2008/2009 financial crisis events.

Analyses during 2017–19 (MIT Sloan, McKinsey, O’Reilly Media, etc.) indicated an accelerating gap between “haves” and “have nots” among enterprise firms. One cohort of enterprise firms had followed the lead of AI incumbents (aka, “hyperscalers”) and had invested in expensive transformations to:

1. Pay down tech debt in their data infrastructure.
2. Develop effective decision processes among executive ranks which embraced how to leverage uncertainty, data, and ML.
3. Hire or train staff for contemporary use of data science, ML, etc.

The accelerating gap indicated more than 50% were not investing yet, even though required transformations would take years to execute. Estimates placed ~2024 as roughly a “point of no return” for the laggards, with subsequent churn (bankruptcies, mergers, acquisitions) at very large scale. We can use that as an approximate marker for the upcoming “new practices” of the employer-employee relationship arc as well as a milestone for enterprise disruption and consolidation in response to AI applications, now gaining momentum due to the COVID-19 pandemic.

Which factors now change the most?

To paraphrase yet another popular science fiction franchise — specifically about the effects of climate change as root causes for what we’re experiencing now — “This changes everything.” Of course science fiction cannot predict the future any better than other speculative endeavors, people who read science fiction tend to be less surprised by abrupt changes. Also, there’s an uncanny historical tendency to imagine actual futures to a large extent. So we’ll keep moving forward with these references to science fiction.

A few truisms about Jackpot phenomenology — i.e., the dramatic shifts in business we’re facing now — can be stated even at this early point:

• Theory: Economists have no baseline data for the current environment and they tend to become even more of a groupthink cult than usual. In other words, the economists who hold tenaciously onto patterns of analysis from a different era become some of the most dangerous people around. Avoid.
• Tenure: Most small businesses get wiped out. Those which manage to survive will tend to have big opportunities to reshape their futures and gain significant advantages.
• Capital: Interest rates drop to historically low levels (for a while) which suggests that borrowing money may be much wiser than vying for (already scarce) VC funding.
• Staffing: Many talented and experienced people become suddenly unemployed, shifting into different career paths and reinventing themselves. This tends to create situations where people bring fresh perspectives to invent new approaches to solving old problems. On the other hand, reinvention can lead to tedious conditions, until the people who aren’t going to last long in a field get churned out.
• Hazards: Pandemic waves will return. Sometimes just when they are least expected and potentially most lethal.
• Climate: Interactions between extreme weather events, geographical regions at risk (especially the Southeast), the fragility of US healthcare, and supply chains in general have been understated in the news cycle — understandably, given everything that’s happening. However, the news cycle for extreme climate events in North America tends to run June through November, i.e., timed for a second or third wave of the pandemic. The impact of climate change is a much longer and more serious narrative than the pandemic, and likely one of its root causes.
• Tension: “The accelerated emergence of the new political spectrum: Networked Consensus vs. Networked Dissent” (Robb, 2020) Fortunately the rhetorical posture of any given actor w.r.t. this political spectrum becomes more straightforward to measure.

A major takeaway from the thesis by John Robb is how our long-term needs indicate an emerging rise of tools for decentralized, networked decision-making at scale:

A challenge of the 21st Century: Societies either learn to think like a network or the network slowly kills them.

The more effective approaches toward mitigating the impact of the pandemic have followed this general pattern. Note that this description dovetails well with the “accelerating gap” analysis of AI adoption in enterprise, as discussed above.

Other conditions are highly likely and should be stated explicitly even if they appear obvious:

• Organizations which ignore network security become risks as state-sponsored cyber attacks seize many enticing opportunities to gain asymmetric advantages. In other words, revulsion about attackers and attacks is probably less urgent than revulsion about those (vendors, partners, customers) who fail to pay attention to security concerns.
• Supply chains become more fragmented, chaotic, and carry higher risks.
• Travel becomes greatly reduced; move to business models and strategies that do not depend on travel.
• The oil economy becomes broken, perhaps permanently? The pandemic also makes the aging nuclear industry significantly more vulnerable. We confront the challenges and perils of climate change now. Arguably one of the clearest voices about this over the past two decades has been Mark Z. Jacobson at Stanford.
• Local manufacturing surges — in lieu of reliance on China and strategies that shave every last penny off the supply chains by burning lots of oil.
• Arguments for keeping legacy software in place diminish (e.g., COBOL+IMS on Z Series mainframes to compute unemployment insurance payments).
• Totalitarian control of media and pandemic responses don’t mix well. Russia and China will be found-out to have been exaggerating their numbers, and meanwhile launching cyber attacks against the rest of the world. Fuck them.
• Facebook, Apple, and Google have been eager to “come to the rescue” for contact tracing data, although they will be found-out for leveraging the situation to establish terrible privacy violations. Fuck them.
• Wild-caught live animals for food, yeah that’s over. Except for seafood, which is almost extinct anyway.
• People purchase more of their food through shorter supply chains — i.e., less processing — but also recognizing how the “farm-to-table” movement in high-end restaurants failed to grasp an important realization, which notable leaders in the food movement (e.g. Dan Barber, Danielle Nierenberg, et al.) are reconciling now. On the one hand, the amount of water and energy that currently goes into food processing (post-farm) is obscene and moreover it introduced critical vulnerabilities. On the other hand, while the more high-end restaurants had optimized for packing noisy crowds of people onto white-tablecloth spaces, in reality the essence of their “back-office operations” were based on highly efficient, highly localized food processing. In many cases, those operations are being reworked in response to the pandemic to to offer farm fresh products to communities in need. These inversions of supply chain related to the food-energy-water nexus are emblematic of what gets overhauled next.
• People work remotely. So many have experienced it now, and while some hate it, many will not return to working in an office setting. The trade-offs regarding commute time and costs, child-care, infectious risks, etc., will outweigh decisions about employers with legacy values, especially in the wake of widespread furloughs and layoffs.

The ascendance of bioregionalism

States on the West Coast were the first in the US to implement interventions for the COVID-19 pandemic. Based on experiences from the 1918 pandemic, these regions will be the first to experience economic recovery.

• Tech firms, which tend to be based in the West Coast already, may concentrate more there and will be hiring lots of unemployed people.
• Historically that’s created nightmare conditions for real estate and cost of living. However, now these firms (and their manager cadre) will be forced to allow employees to work remotely.
• One likely outcome is that more tech workers will choose to relocate their families to smaller communities (lower cost, lower risk) outside of the West Coast tech hubs.
• This will open opportunities for more high-paying work to people all over the world. NB: that runs contra to the notion of “outsourcing” as teams tend to become more far-flung in nature anyway.

Some points will likely emerge as insurers rework their risk management models and adjust criteria for assessing business insurance premiums:

• Employers will need to keep some percentage of their workforce working remotely at all times, as a “corporate lifeboat” for perpetuating operations.
• Suppliers will need to become more localized.

Two other observations beg some discussion even if they are relatively conceptual and abstract:

• The socioeconomic arc of deregulation => financialization appears to be encountering end-of-life events. If your business strategy depends on obscure/complex financialization tactics, you’re probably fucked.
• The socioeconomic arc of decentralization gains significant proof points. Those who know will understand.

The realities— the urgencies — of the on-the-ground interventions in response to the global crisis illustrate the practice of bioregionalism and its ascendence in public policy. At least three major pacts among state governments within the United States have formalized regional accords to shape their policies in response to the pandemic. As of this writing (2020–05–03) these regional accords account for nearly 60% of the 2019 estimates of GDP in the United States:

Note that the “blue state-red state divide” — which has become so emblematic of the clichéd neoliberal vs. populist/conservative political deadlock in the US elections — to a large extent depends on a regional axis of funding. In recent years, the combination of California and New York have served as economic powerhouses among the “blue states” and their policy endeavors, along with Illinois and other large-ish states which tend to lean more toward progressive approaches. These three states (CA, NY, IL) provide center-mass for each of the new regional accords, respectively, and that is most likely no coincidence.

During the same period of political deadlock we’ve seen the state economies of Texas and Florida counterbalance funding in support of the “red states” and their policy endeavors, along with North Carolina, Georgia, etc. In the wake of the pandemic, economies of the “red state” center-mass (TX, FL, NC, GA) are likely to get hit hard: oil industry, meat industry, cruise ships, commercialized megachurches, for-profit prisons, exploitive retirement communities, large scale amusement parks, etc., will be especially vulnerable to the effects of the pandemic. Moreover, they’re tracking in lockstep with White House rhetoric which is going exactly nowhere — other than to bury these regions even deeper into existential risks long-term. These regions are notable for having the highest per-capita rates of people with pre-existing health complications, as well as the highest per-capita rates of people who have no health insurance. While the medical research community grapples with the question of aerosols w.r.t. COVID-19, there is mounting evidence that PM 2.5 particulate has correlates strongly with the geographic regions that have experienced high death rates. Here again, the TX-FL-GA axis is where deregulation and other long-term conservative policies have led to horrible AQI conditions and chronic illness at scale which is almost purpose-built for a COVID-19 perfect storm. As the 2020 hurricane season in the Northern Hemisphere comes into view for its typical June-November run, these are also the regions which will be the most hard-hit by extreme weather events that create emergency crises within emergency crises while stressing the healthcare system, unemployment, debt, and overall logistics even further.

Pandemic deaths per-capita: 2020 regional conditioning for 2019 point estimates of per-state GDP

Meanwhile, the economies of the “blue state” center-mass in the regional accords tend to rely on business models and industry sectors which are significantly less vulnerable to the pandemic (e.g., emphasis on remote work, agriculture and food production, support of immigration and sanctuary cities) and may prove more adaptive as a result. In any case, the relative changes in per-state GDP contribution between 2019 and 2020 will almost certainly shift in favor of the “blue state” centers.

At Derwen we’ve created a list of recommended dashboards plus other custom data products specifically for monitoring that emerging landscape. Look especially at the economics dashboards regarding consumer trends, layoffs, strikes, and so on.

A scenario to consider

Circa early 2000s, many were reeling from economic and political chaos: DotCom Bust, 9/11, the years following 9/11. Rather quietly during the chaos, the world shifted. Prior to that time, advertising deals had typically been struck by a couple good ol’ boys having cocktails after a round of golf. However, a marriage of ML-based search engine with ML-based advertising network occurred roughly circa 2000, although it took a few years for the world (or perhaps even Google execs) to recognize the significance of that subtle change. Those former advertising execs out at the country club could no longer keep up with the world, and quickly became about as relevant as dinosaurs.

The global pandemic crisis has highlighted so many failures in both public policymaking and corporate decision-making. Those good ol’ boys sharing cocktails simply cannot keep up with viral infections, failing healthcare infrastructure, the extreme effects of climate change, etc., so they seek to sweep information under the rug by drowning it with noise.

Then the world shifts suddenly: https://blog.einstein.ai/the-ai-economist/

The signal emanating from the AI Economist reverberates in polyrhythms that, on the one hand, make intuitive sense, while on the other hand tend to fracture the foundations of oh so many assumptions of the fading 20th century. Now the armies of economists, public policymakers, corporate decision-makers, et al, who aren’t augmented by AI systems become about as relevant as an advertising executive playing golf. Alexis Richardson — one of London’s finest thinkers about technology pursuits in general— noted about this work from Richard Socher, et al:

It would be ironic if the thing that finally pushes innumeracy out of politics is AI, which delegates numeracy to a machine. I suppose you still need to ask the right questions and understand the answers. Maybe that will raise the bar for decision makers.

I couldn’t agree more.

Overall, it’s probably a good time for Boomers to retire and take the plague of their outmoded work traditions along with them! (says the aging GenX-er)

In late 1989, I was working on behalf of a large electronics manufacturer on a technology demo tour through Pacific Rim countries: taking the latest tech out of the lab and out to the field. One sales call to a customer site in Hong Kong was most poignant. We trekked out to a remote corner of HK (if any point in HK could even be called “remote) and up into a nondescript residential flat that was used for a tech start-up. The firm was importing old sewing machines from China, behemoths which could’ve been sold at a profit for scrap metal because they were industrial nightmares of iron and steel. Then they would retrofit these with microprocessors to produce “xyz tables” — in other words, creating industrial robots for pennies on the dollar. It was a brilliant response to the market dominance of Japanese industrial robots which were popular but complex and extremely expensive.

A similar market condition holds today w.r.t. AI applications, although few of the practitioners may have noticed it yet. While the general area of supervised learning approaches in machine learning have led to widespread enterprise adoption for predictive analytics, especially in deep learning since 2013-ish, there are other areas which have less commercial adoption. On the one hand, work on knowledge graph approaches has be growing since the 19th work by Charles Sanders Peirce, who is one of the most prolific scientists in US history. KG provides flexibility for representing new sources of real-word data along with context that supervised learning approaches generalize away. It also tends to support explainability in AI systems, which is crucial for ethical, robust technology adoption and overall trust. On the other hand, the field of reinforcement learning arguably traces back several decades to the field of optimal control theory and related work in robotics and optimization through agent-based simulations. RL leverages the use of one or more agents within a simulated environment to explore/exploit the gradients of solutions, then leverages deep learning to develop policies that provide robust responses to changing environments. Commercial applications are emerging, although much of this work has been within academic research teams. In particular, note the 2019 groundbreaking keynote talk by Michael Jordan about the intersection of markets and reinforcement learning, which essentially states that markets represent a form of machine learning.

RL researchers tend to build complex simulations to demonstrate evidence of their learning algorithms and their respective merits. Unfortunately, as I can attest based on client work, many of the RL researchers will struggle when pressed to articulate a real-world application other than playing video games. Ergo, note Jordan’s game-theoretic emphasis above. However, much like those behemoth Chinese sewing machines imported by the HK tech start-up building inexpensive industrial robots circa 1989, there are large quantities of extensive simulations already used through enterprise. Mostly definitely we see this for optimization work in supply chain analysis, factory optimization, and other areas of logistics — where my undergraduate advisor at Stanford focused, as did so much of my early academic formalized torture. To date, some of the better blending of machine learning and more traditional optimization work in enterprise has been focused in shops such as IBM (full disclosure: this work involves two Derwen clients).

In the wake of the pandemic, expect to see AI breakthroughs that tend to localize or regionalize supply chain optimizations which take into account more than one objectives to optimize. These will leverage KG and RL work, probably with as much more more momentum than what supervised ML experience in enterprise adoption following the DotCom Bust through post 9/11 crisis period. In other words, resilience and compliance may become just as important as objective functions as profit. Those polyrhythms resonating from the AI Economist tend to support this point.

A relatively upbeat note, if you haven’t read this article already I highly recommend the excellent “The Coronavirus Is Rewriting Our Imaginations” by Kim Stanley Robinson (New Yorker, 2020–05–01):

The virus is rewriting our imaginations. What felt impossible has become thinkable. We’re getting a different sense of our place in history. We know we’re entering a new world, a new era. We seem to be learning our way into a new structure of feeling … The spring of 2020 is suggestive of how much, and how quickly, we can change as a civilization.

Meanwhile we wish you and yours to keep safe and be well,

Derwen.