Silicon Valley’s myths and realities of existential risk

It's not AI that's going to kill us, it's inaction

Existential risk has been on many of our minds the past year. Sales of survival goods from food kits to nuclear-proof bunkers are way up, we doomscroll on Twitter all day, and it seems like there isn’t a week that goes by where civilizational collapse isn’t at least a possibility on the agenda.

Yet, if you hang out in tech circles long enough, there remains an astonishing divergence between the realities of existential risk and the speculative nature this subject tends to push us toward.

In Silicon Valley, the fun topics here are scenarios like custom-designed pathogens that assassinate individuals or the entire human population constructed in a small biolab by an irascible bio Ph.D. (throw in CRISPR as an acronym to make it sound interesting). Coronal mass ejections or some sort of electromagnetic bomb comes up frequently, events that could knock out all power as we know it. You’ll also frequently run into some sort of hacking scenario where all the chips in the world are vulnerable to the same line of code (perhaps inspired in the vein of Meltdown or Spectre).

What the last year has shown is that we have a very bad cognitive bias where we think about the speculative dangers far too much and the mundane civilizational dangers far too little.

These scenarios are fun and imaginative, and it’s a great Zoom drinking game in an otherwise “what do you do for a living” conversation.

What the last year has shown, however, is that we have a very bad cognitive bias here where we think about the speculative dangers far too much and the mundane civilizational dangers far too little.

COVID-19 is the too-obvious example, a pandemic that has been predicted in some form or another for literally decades. But it’s hardly the only “boring” disaster that’s befallen us. The failure last month of much of Texas’ energy grid knocked out power for days for millions of people at some of the coldest temperatures experienced locally — no electromagnetic bomb required. California’s wildfire season has expanded, leading to lives lost, wide-scale power outages, billions of dollars in damages and that iconic orange air — no Hollywood special effects required. A “software issue” led to a swath of the East Coast losing the internet in January, while a lone bomber in Nashville knocked out much of the telecommunications in that metropolitan area over the Christmas holidays.

Here, then, is the divergence: we have had forms of civilizational collapse, but so far, they’ve been limited in duration and limited in scope. We didn’t all lose power: just Texas last month and California last year. We didn’t all lose internet: just the East Coast and Nashville at different times. We didn’t lose civilization, we just had a pandemic that has forced much of the world to regularly shut down schools and stores to limit viral spread. It’s almost like a disaster doesn’t count if Netflix still operates for 10% of the population.

So we have folks talking about artificial general intelligence and the singularity and fleeing to Mars, when the immediate reality is that there are thousands of dams under incredible strain where millions of people could perish in the coming years if a number of these fail. Not just fail hypothetically, but fail as predicted when these structures reach the end of their usable lives and become increasingly vulnerable to collapse.

What’s strange is how much we as people have begun to cognitively route around these quotidian catastrophes. AWS outages that used to elicit extreme opprobrium a few years ago are now a snow day from Zoom calls. Power outages are just the new normal. Pandemics — well, why bother wearing a mask at this point anyway, even as variants start to threaten the recovery we all expect? As one colleague put it me, just buy a battery-powered radio — you know you’ll need it here soon, as if we should just expect to lose connectivity at any time.

The solutionism that makes Silicon Valley an entrepreneurial spectacle never seems to migrate over to the solvable world of most daily existential threats. Power grid failures are preventable. The internet was supposed to be designed to route around damage, not be based in a couple of central data centers and exchanges where one rogue patch or saboteur brings down the global GDP. Healthcare systems are capable of managing outbreaks — we know the playbook, if only we could execute on one.

Perhaps more frustrating than the lack of resilience and planning, which use precisely the sort of analytical skills that the Valley loves, is the sheer lack of action during these catastrophes. If the last year has shown us anything, it’s the complete lethargy from government to organizations to everyday citizens, all of which are apparently completely unprepared to do anything if disaster strikes.

Now, I don’t want to cast aspersions on the crowdsourced projects that sprung out of COVID-19 to direct people to hospitals, or to track data, or to guide people to finding vaccines today or finding masks in the early, hectic days. These projects are important despite their fumbles, and represent a fresh and flourishing civil society. It’s key though not to assume that keyboard actions can somehow compensate entirely for a lack of action in the field. The tech industry loves to code up a web app to solve all problems, when most disasters really and truly can’t be responded to with Python code.

Within the tech community, the one exception I have been able to find is Google co-founder Sergey Brin, who seems to have put his time and resources behind building out global capacity for disaster response with an organization called Global Support and Development, which Mark Harris described at length in a piece last year. Per his article:

For the past five years, GSD has been quietly using high-tech systems to rapidly deliver humanitarian assistance during high-profile disasters, including the COVID-19 pandemic. These range from drones and super-yachts to a gigantic new airship that the outfit apparently hopes will make it easier to get aid supplies into disaster zones.

We need more of this, and stat.

Existential risk has always been at the heart of the tech industry. From a radio broadcast of “The War of the Worlds” and the Manhattan Project to AI and cybernetics in the 1960s, cyberpunk and climate punk and all the other punks in the 80s, continuing right up to artificial general intelligence and the singularity today, we’ve always known that the technological progress we make could have massive consequences for the world as we know it.

Yet, it’s time to turn our attention away from the crazy and speculative and future, and more toward the chaos and challenges facing our world right now, using tools that we have available to us today. Most of our major challenges aren’t just solvable, they’re eminently solvable if we put the time into them. But that requires us to actively thwart our cognitive bias away from the scary and improbable to the mundane and normal — the boring events that almost certainly will actually kill us in the end.