HOW MUCH RISK CAN WE HUMANS TOLERATE?

I’m not cranking out many blogposts lately.   I tell myself that, having turned the big 8-0 in age by years now, I have a right to slow down.   I’m not doing as much as I used to in many different areas.   But for me the most noticeable area of all in that of risk-taking.   That’s where I’ve slowed waaaaaay down over the years.

As an 80 year old dude looking back over my life, I’ve taken some risks in the past that now seem absolutely insane, not just ill-advised.    I’ve survived more high-risk moves than most cats going on life #10.    Some on the water, some on the highway, and the rest scattered across an adolescence and adulthood of simply stupid decisions.   

Which brings me to a question about the risk to human survival if Artificial Intelligence (AI models already in use and more forthcoming) goes unchecked or deregulated.    Obviously, I’m not the only one asking such a question.   Most folks, if paying attention, are wondering about the future risks associated with AI.   Perhaps even wondering with me about the very risk of human survival itself if we don’t get AI right in our design.   

Which then brings me to the question of whether our Actual (human) Ignorance itself (the other AI) is not at its own highest risk level ever.   Whether by selfish will or pure accident, our human ignorance carries a past track record of death and destruction brought upon ourselves as a human species.    If we’re equally ignorant about our design of Artificial Intelligence models, how risky will that be?  

Are you still with me here or have I already lost you?

My experience of “old age” thus far is that I calculate risk rather differently than before.    For example, in areas of medical testing………something 80 y/o adults are most familiar with……the risk of NOT testing is considered far higher than the risk of testing and finding out bad news.    The risk of early detection is VERY low when compared to the risk of late detection.   The risk of doing something vs. nothing is not always straight forward, but there are patterns worthy of our notice should we hope to live until, say, 90 y/o.   

And so I recently wondered within my own mind which risk is higher where human survival is concerned, that of Artificial Intelligence if not adequately regulated or that of Actual (human) Ignorance if also unchecked?     Which is another way of asking, will we kill ourselves at a higher or faster rate by means of our own human ignorance (willful or accidental) or by means of our artificial intelligence models now in design or production?   

I decided to ask such questions to ChatGPT out of curiosity.     I wanted to learn its own calculation of risk for human destruction when comparing our own actual ignorance to its own artificial intelligence.  Here is a condensed sampling of its answer:     

“If by ’human ignorance’ you mean lack of knowledge, misunderstanding, or failure to recognize relevant evidence—not simply stupidity—there are many historically measurable negative consequences. The strongest examples are cases where we can connect an information deficit to an observable outcome:

·         Disease and preventable deaths. Before germ theory, physicians often did not understand that microorganisms could transmit disease. Handwashing and antiseptic practices subsequently produced large reductions in mortality in some medical settings. Ignorance of disease mechanisms therefore had measurable consequences in morbidity and mortality.

·         Famine from misunderstanding ecological and agricultural systems. Historical famines have sometimes been worsened by inadequate understanding of crop disease, soil depletion, food storage, or ecological interactions. The Irish Great Famine, for example, involved a potato blight whose biological mechanism was poorly understood at the time, compounded by economic and political factors. The measurable consequence was roughly a million deaths and mass emigration.

·         Industrial and occupational deaths. Workers historically encountered asbestos, lead, mercury, radiation, and other hazards before their long-term effects were adequately understood. Once epidemiological evidence accumulated, occupational regulations reduced exposure. The consequences of earlier ignorance can be measured through excess mortality and disease incidence.

·         Nuclear and radiation exposure. Early researchers and workers sometimes underestimated radiation's biological hazards because the science was still developing. The resulting exposures—including among some early radiation workers and populations affected by nuclear testing—produced measurable increases in particular cancers and other health effects.

·         Environmental pollution. The widespread use of lead in gasoline is a particularly striking example. Atmospheric lead exposure was eventually linked to neurological and developmental harm, and removing lead from gasoline produced large reductions in population exposure. Retrospective studies have estimated substantial effects on childhood cognitive development and later-life outcomes.

·         Financial and economic disasters. Ignorance of complex systemic relationships can produce enormous losses. The Great Depression, for example, occurred partly in an era when economists and policymakers had a much poorer understanding of monetary policy, banking-system fragility, and macroeconomic stabilization than exists today. The measurable consequences included massive unemployment, business failures, and declines in output. Importantly, this wasn't simply “people didn't know enough”; competing theories, institutions, incentives, and political decisions also mattered.

·         Engineering and technological catastrophes. The causes of disasters such as the Chernobyl disaster included inadequate understanding of reactor behavior, flawed design assumptions, deficient safety culture, and operator/institutional failures. The consequences can be quantified in deaths, radiation exposure, displacement, economic costs, and contaminated land.

·         Natural disasters amplified by ignorance. Lack of understanding about earthquakes, hurricanes, tsunamis, and volcanic hazards historically resulted in much greater mortality than comparable events can cause today. Improved hazard mapping, building codes, early-warning systems, and evacuation procedures have substantially reduced vulnerability in many places.

·         Agricultural catastrophes caused by mistaken biological assumptions. The Great Famine of Ireland illustrates how monoculture, pathogen susceptibility, and inadequate understanding of plant disease can interact with human institutions. Again, ignorance was one component rather than a sufficient explanation.

·         Medical treatments that were ineffective or harmful. Bloodletting, certain mercury-containing treatments, and other historical medical practices were sometimes maintained because their underlying assumptions about physiology were incorrect. Modern clinical trials provide measurable evidence of the difference between some of these practices and effective treatments.

·         Species extinctions and ecosystem damage. Humans have repeatedly failed to understand ecological feedbacks and the long-term effects of introducing species, overhunting, or destroying habitats. The extinction of the dodo and numerous other species illustrates the irreversible consequences. Modern ecological science has also demonstrated measurable effects of biodiversity loss on ecosystem functioning.

·         Discrimination reinforced by false beliefs. Some historical systems of racial, ethnic, or social hierarchy were justified using claims that were later shown to be scientifically unsupported. The consequences can be measured in disparities in mortality, wealth, education, incarceration, displacement, and political rights. Here, however, ’ignorance’ alone is insufficient: deliberate exploitation and institutional incentives can coexist with, or even survive despite, accurate knowledge.”

I found myself wondering about such other human fatalities as related to military warfare or the genocidal patterns of human tribalism that, in all fairness, relate more to human immorality than to mere human ignorance.  Or about human fatalities past, present, future from human-induced global warming.   And I started to sink deeper into the thought of high risk human behaviors (versus those of artificial intelligence) when going on to read the following ChatGPT answer:  

“We almost certainly underestimate the historical cost of ignorance.  Deaths from an unknown disease, nutritional deficiency, environmental exposure, or poor engineering practice often weren't recognized as belonging to a common cause. Once the cause was discovered, the historical record could be reconstructed—but many cases remain invisible because nobody knew what to measure.

That makes the comparison to future AI risk particularly interesting: with ordinary ignorance, humans often discover the missing variable after causing damage. With sufficiently capable AI, one of the central concerns is that the system could encounter a domain where humans themselves don't know what they don't know, while operating at a scale and speed that makes the traditional 'experiment → disaster → learn → correct cycle' much more dangerous.

If you want, I can build a quantitative historical ignorance vs. AI risk framework, including approximate deaths/$ damage, reversibility, speed of onset, ability to detect the mistake beforehand, and whether the ignorance was human ignorance, institutional ignorance, or unknown-unknowns.”

Okay, so this is where I left it.  

At my age, why take the risk of going any further?   I’m not a care-free kid anymore! 

Maybe it’s better to apply the brakes on managing the risk of Artificial Intelligence in areas of human destruction until after we’ve managed the risk of human ignorance and human immorality in areas of our own self-destruction.    What’s to stop us from already becoming our own worst enemy?

Now there’s a task for ChatGPT to figure out!