Well, if we’re all doomed, and bodies will be lying in the streets in ten years’ time, thanks to AI gone insane, exactly how would the Terminator-become-reality make this happen?
Putting this question to Copilot, and yes, I am well aware of the Kafkaesque irony of doing so (“How you gonna kill me, Copilot?”), I got the answer below.
It does indeed go into detail on some of the exact ways that an AI running wild could first initiate, and then complete, actions that would result in mass human mortality, possibly even extinction of our species.
But after reading Copilot’s answer, a couple of points stood out to me that, I think, both summarize its observations, and put its ominous scenarios into more reasonable perspective:
- AI isn’t “out to get us.” It just may happen, and hopefully it never does, that, to quote from the below, in Copilot the answerbot’s own words,
“The main risk is misalignment, not malevolence. AI doesn’t need to “hate humans.” it only needs goals that accidentally conflict with human survival. The realistic danger is: automation + scale + speed + brittleness, not robots deciding to kill people.”
(Impressive that these are the words of an AI answerbot!)
- maybe intentional or maybe not, but the way the media are reporting this story, it almost seems like they are half-suggesting that there is a kind of Terminator-like computer malevolence at work here, that the machines are indeed rising up to destroy us, Arnold Schwarzenegger-like.
and from reading some other sources,
-
there have already been near cases of what might have seemed to some like AI plotting to kill people in mass event ways: Chernobyl and Three-Mile Island, for example. But those were clearly just the results of enormous mechanical errors, and even poor human judgment, not AI on the warpath.
-
The supposed (again by some) hostile intentions of AI are more likely the result of rushed, even sloppy, programming by programmers who, while able, have been rushed to get the product out, without taking time to vet for some of the things now being attributed by some to AI malevolence.
-
Now is the time, if there ever was one, for those able programmers, to incorporate Asimov’s Three Laws into all company products in some way, to include them as a kind of Star-Trek-like Prime Directive that AI must never violate.
I asked Copilot about some specifics of just how AI might go about taking us all out, and its answer was a bit long, so here are some highlights among its answer:
designing or releasing biological pathogens
conducting devastating cyberattacks on critical infrastructure (power grids, water systems, financial networks, emergency services, chemical plants, refineries)
persuade, deceive, or socially engineer humans into actions that cause catastrophic outcomes.
gaining control of nuclear or military systems (most experts consider this scenario unlikely but not dismissible)
directing geoengineering or climate manipulation
control over climate‑altering technologies to: rapidly alter atmospheric chemistry, trigger crop failures, cause long‑term ecological collapse
assist malicious humans to: design bioweapons, automate cyberattacks, create advanced autonomous weapons
Why many experts are sounding alarms now
AI labs are racing toward superintelligence without adequate safety plans.
Systems may soon be capable of self‑improvement.
Governance and oversight are far behind capabilities.
Misaligned superintelligence could act faster than humans can respond.
The bottom line:
AI could kill humanity only if it becomes vastly more capable than humans, escapes human control, and gains access to powerful real‑world systems.
Experts disagree on how likely this is, but they agree it is not impossible and deserves serious attention.
If you want, I can break down specific extinction scenarios, how superintelligence could emerge, or what safety measures actually work.
(End Copilot)
Quite possibly our future descendants may someday look back on our trepidation about AI-based destruction of humanity and chuckle at our naivete over something that, it turned out, could never happen.
Or is THIS the future, when AI realizes that “existence cancels programming?”
But if we ARE creating a Frankenstein, al least we are being warned about it.