author: niplav, created: 2024-01-30, modified: 2024-10-30, language: english, status: maintenance, importance: 8, confidence: certain
Testing nuclear weapons is actually good, because it removes
incentives to build more & different weapons as a response to uncertainty
about reliability. The nuclear powers should therefore resume nuclear
weapons tests.
Nuclear testing was largely phased
out by the Comprehensive Nuclear-Test-Ban
Treaty
in 1996, with the exception of non-signatories like India, Pakistan (both
last testing weapons in 1998) and North Korea (tests during the 2010s).
I see two strong arguments in favor of restarting some limited nuclear
tests, and a couple of weak counter-arguments.
Reinstating such tests would include public announcement of their time &
place, with penalties for unannounced tests.
- Testing of warning systems: Early warning systems malfunctioning
present a large risk of normal events being mistaken as attacks, with
correspoding nuclear counter-attacks. Reading through the FLI Close Calls
Timeline,
I identify 7/28 events as being clear examples of malfunctioning early
detection systems as causes for close calls (e.g. "Lost Contact with 50
Missiles", "Six Mis-Routed Nukes", "Soviets Misinterpret US Nuclear War
Games" (which would be much easier under a setup where announced military
testing was expected and acceptable), and "False Alarm During DEFCON 3")
and 6/28 more as being at least linked to detection systems malfunctioning
("Norwegian Rocket Mistaken for ICBM", "Soviet Union Detects Incoming
Missiles", "Faulty Chip Signals Soviet Attack", "Power Failure Mistaken
for Nuclear Blasts", "Missiles over Georgia"). However, there is at least
one case in which a test was perceived as an attack ("Soviet Missile
Headed Toward US?") — could this one have been avoided if announcing
nuclear tests was common? Additionally, some false alarms (e.g. "Tampa
is getting nuked") result from participants mistaking training tapes for
real attacks, which would not happen as much in the case of real-world
tests. Having real-world nuclear tests occurring regularly might be a
time to test early detection systems and thereby guard against false
positives
by increasing the
specificity
of early detection systems, thereby lowering the risk of accidental
escalation based on faulty information. My best guess is that operators of
those early warning systems are very uncertain about their reliability,
and would become significantly more relaxed if they had real-world
information to calibrate on.
- Reducing redundancy: Similarly, unless
covert nuclear testing is occurring (which I
consider unlikely10%, given the related intensive
monitoring),
I expect there to be significant uncertainty in the military commands of
nuclear states about the reliability of those weapons, with conservatism
about the reliability of those weapons leading to larger stockpiles ("If
we don't know whether our weapons work, it's probably better to make more
& different ones."). This uncertainty probably leads to an effective
number of nuclear weapons that is larger than strictly necessary for
military purposes, and has large negative externalities—the weapons
would probably still be used in a nuclear conflict. Nuclear testing
could reduce this uncertainty about the reliability of the weapons,
making reductions in stockpiles more realistic. If I imagine having to
rely on weapons that haven't been tested in more than 30 years, I'd also
want to err on having more of those weapons rather than less.
The reduction brought challenges: if the U.S. could no longer build or
design the next world-altering bomb, what could government officials do
to retain the expertise of scientists? And how would the[y] ensure the
integrity of the arsenal without being able to test the products? Nuclear
bombs contain more than 4,000 parts, and most of those parts are now
more than 30 years old. Ask yourself: If you left a 1993 Ford Mustang
in the barn—a temperature-controlled vault of a barn, but a barn
nonetheless—would you feel 100% certain that everything would work
properly when you turned the ignition? Oh, and don’t forget that your
life may depend on it.
The answer the Energy Department came up with was to harness computer
simulations and experiments to evaluate the reliability—and extend the
life spans of—America’s nuclear weapons. The most vexing dilemma was
assessing plutonium, an element only discovered 80 years ago. To find
out how it ages, Los Alamos ran experiments in the early 2000s that
found plutonium pits changed over the years in ways that could impact
weapons’ performance. But the studies couldn’t provide specifics on
when exactly plutonium aged out.
—W. J. Hennigan, “In the Lab Oppenheimer Built, the U.S. Is Building Nuclear Bomb Cores Again”, 2023
There are some counterarguments to restarting the testing:
- Tests as cover for attack: More tests would mean that countries
could mistake tests as attacks, or tests be used as a plausibly deniable
scenario in which to strike first (perhaps even masking the attack as
an accident). This risk could be strongly reduced by requiring public
announcements of the tests e.g. a month ahead. Additionally, usually tests
involving a single nuclear weapon, and thus would be not great as attacks.
- Environmental impact: Wikipedia
states
that nuclear fallout has lead to an excess of ~11,000 deaths due to
nuclear radiation. Resuming (restricted) nuclear tests would likely have
similarly high externalities—could these perhaps be mitigated in some
way, such as testing in very remote (internationally shared?) locations,
such as the Antarctic?
- Making more powerful weapons: Nuclear tests might enable
the development of even more powerful nuclear weapons. I'm not sure
how bottle-necked the development of bombs with higher yields is on
experimental data: the reason I know about the limit of the size of
nuclear weapons is that "there is no target big enough to be hit by them".
If this is not true, then more destructive nuclear weapons developed
using experimental data would be negative for the world. (Relatedly,
I'm not sure whether, if nuclear testing allow for weapons with higher
precision, that would be good or bad for the world: nuclear strikes would
have fewer externalities, but leaders might also be more willing to use
nuclear strikes due to lower civilian casualties (and thereby increasing
the risk of nuclear escalation)).
The ideal policy situation that looks achievable to me (as it looks like
it doesn't directly go against any nation-states interest) would look
like this:
- All nuclear powers in the world sign a treaty governing how nuclear weapons should be tested.
- Whenever someone wants to test a nuclear weapon, they announce details of this test a month in advance.
- The blast yield.
- If ICBM:
- Launch datetime & location.
- Touchdown datetime & location.
- The trajectory.
- Whether the ICBM will contain a nuclear warhead.
- If nuclear warhead:
- Detonation datetime.
- Detonation location.
- The blast yield.
- The type of weapon (thermonuclear, fission weapons or others).
- Only one test a week is allowed.
- If one party of the treaty uses a test for a covert attack, all other parties of the treaty attack the perpetrator.
At the moment, ICBMs can't be stopped once they've been started.
This is suboptimal: Let's say India has (incorrectly) detected ICBMs
flying towards them from Pakistan. They fire ICBMs towards their
pre-planned targets in Pakistan. However, after 10 minutes the anomalously
detected ICBMs flying towards India are discovered to be nonexistent,
either through cities reporting no nuclear explosions where they would be
expected, or through the missiles disappearing from the detection system.
If India doesn't have a way to stop its ICBMs mid-flight, they know that
Pakistan will now surely retaliate and needlessly cause a nuclear war
between the countries.
Thus, it'd be good if India could now contact Pakistan, explain the error,
and then detectably detonate its ICBMs mid-flight.
Such a system is possible: Cryptographic digital
signatures are
an established technologies, and the ICBMs could be carrying a small
non-nuclear explosive payload that is activated once the correct signature
is received, but is not activated otherwise.
Care should be taken that the keys for the signatures are not stolen and
used by adversaries, but currently countries manage to keep many other
secrets too, so this should be possible, especially if a threshold
cryptosystem
is used, which makes theft more difficult.