>Altman's death would change nothing about that fundamental calculus. You'd have to kill probably tens of thousands of people to really put a dent in AI development
Your analysis seems to assume that people will remain more afraid of being "outcompeted" than of being murdered, even after a campaign of terrorism that would make 9/11 look minor.
>it often also creates new [problems], often surprising ones
Let's reframe this to remove the negative bias: murder has the obvious direct first-order effect of removing the target from existence, but also a host of non-obvious higher-order effects resulting from people's response to that violence. These can be counterproductive to the goals of the murderer, but they can also work in favor of it. That is why "terrorism" is a real thing - the higher-order effects are essentially a force multiplier, and if you have nothing to lose then the calculus of causing a major disruption begins to look favorable; any disruption, because regression to the mean is good if you're at the shitty end of the bell curve.
>Your analysis seems to assume that people will remain more afraid of being "outcompeted" than of being murdered, even after a campaign of terrorism that would make 9/11 look minor.
AI is such an important technology that in the face of such a campaign of terrorism, governments would bring the development of the technology directly under the protection of the state security forces, largely outside the reach of terrorists. If not in the US, then in China or other places. At that point the terrorists would have to attain a level of power where they could feasibly overthrow the government in order to stop the development of the technology. Now, some scientists would be uncomfortable in such conditions and would stop working on the technology, but enough would remain that the technology would continue to progress, albeit more slowly.
>and if you have nothing to lose then the calculus of causing a major disruption begins to look favorable; any disruption, because regression to the mean is good if you're at the shitty end of the bell curve.
Very true, if the status quo feels shitty enough one becomes extremely willing to just roll the dice.
Your analysis seems to assume that people will remain more afraid of being "outcompeted" than of being murdered, even after a campaign of terrorism that would make 9/11 look minor.
>it often also creates new [problems], often surprising ones
Let's reframe this to remove the negative bias: murder has the obvious direct first-order effect of removing the target from existence, but also a host of non-obvious higher-order effects resulting from people's response to that violence. These can be counterproductive to the goals of the murderer, but they can also work in favor of it. That is why "terrorism" is a real thing - the higher-order effects are essentially a force multiplier, and if you have nothing to lose then the calculus of causing a major disruption begins to look favorable; any disruption, because regression to the mean is good if you're at the shitty end of the bell curve.