From Rhodes' Twilight of the Bombs:
For the Senate subcommittee, Pavlov reviewed how the several levels of the Soviet control system worked together:
"...Let me describe … one possible scenario of attack under the conditions of the coup. The early warning system detects a missile attack and sends signals to the subsystems that assess the threat. It is a process that immediately involves the president of the country, the minister of defense, chief of the general staff and the commanders in chief of the three branches of strategic nuclear forces.
"Then the chief of the general staff and commanders in chief of strategic nuclear forces form a command and send it down to the subordinate units. In essence, this command is meant to inform troops and weapons systems about a possible nuclear attack, and this command is called a preliminary command.
"The preliminary command opens up access by the launch crews to the equipment directly controlling the use of nuclear weapons and also gives them access to the relevant special documentation. However, launch crews do not [yet] have the full right to use the equipment of direct control over the use of nuclear weapons.
"As a more accurate assessment of the situation is made, a message is received from the early warning systems confirming the fact of nuclear attack, and the decision to use nuclear weapons may be made at that point. It can be carried out according to a two-stage process."
The first stage of this two-stage process, Pavlov continued, once again involved the top leadership in a political decision—whether or not to generate a “permission command” that would be sent to the CICs. Then, during the second stage, the CICs and the chief of the general staff would decide as military leaders whether or not to generate a “direct command” ordering launch crews to fire their weapons. Even then, the direct command had to pass through an ordeal of what Pavlov called “special processing42 by technical and organizational means to verify its authenticity.” Each of these actions had time limits, and if the time for an action expired, the blocking system that normally prevented weapons from being launched automatically reactivated.
Cumbersome as the Soviet system seemed from their descriptions, Blair pointed out, it was “actually devised … to streamline43 the command system to ensure that they could release nuclear weapons within the time frame of a ballistic missile attack launched by the United States, that is to say, within 15 to 30 minutes.” And despite its complexity, Blair added, a nuclear launch by the coup leaders might still have been possible had they persuaded the general staff to issue Yanayev a Cheget and had one or more of the CICs gone along. “There is an important lesson44 here,” Blair concluded. “No system of safeguards can reliably guard against misbehavior at the very apex of government, in any government. There is no adequate answer to the question, ‘Who guards the guards?’”
More (#1) from Twilight of the Bombs:
...That November, the Republican Party won a landslide victory in the Clinton midterm elections, the first Republican legislative majority in forty years. Democrats lost fifty-four seats in the House of Representatives. Newt Gingrich, the new House speaker, announced his Contract with America. The new crowd of representatives brought a highly parochial perspective to government, Christian Alfonsi noted:
"Many of the new Republicans37 on Capitol Hill were young enough to have avoided Vietnam entirely; and most of thos
One open question in AI risk strategy is: Can we trust the world's elite decision-makers (hereafter "elites") to navigate the creation of human-level AI (and beyond) just fine, without the kinds of special efforts that e.g. Bostrom and Yudkowsky think are needed?
Some reasons for concern include:
But if you were trying to argue for hope, you might argue along these lines (presented for the sake of argument; I don't actually endorse this argument):
The basic structure of this 'argument for hope' is due to Carl Shulman, though he doesn't necessarily endorse the details. (Also, it's just a rough argument, and as stated is not deductively valid.)
Personally, I am not very comforted by this argument because:
Obviously, there's a lot more for me to spell out here, and some of it may be unclear. The reason I'm posting these thoughts in such a rough state is so that MIRI can get some help on our research into this question.
In particular, I'd like to know: