I agree that the tort system is underrated in terms of handling many such concerns about AI.
But of course it doesn’t come close to addressing AIs launched by rogue countries.
Nor by people with no assets who “set free” such agents. Blaming the model provider for all actions is not an answer; that’s like blaming gun manufacturers for all deaths caused by guns or lighter manufacturers for all arsons.
The tort system requires a functioning legal system that covers all injured parties.
1. Do you seriously think that AIs developed and hosted in Red China will have any judgements enforced against them without wide government backing?
2. We are seeing more verdicts, both criminal and civil, where the determining factor in trials is the political affiliations of the judges and jury pools vs the plaintiff and defendent, rather than any law or evidence. Do you think that legal system is functional in any meaningful sense?
> AI will not putter along causing no damages for years before it suddenly wakes up and decides to destroy humanity. The actual dangers of AI will manifest gradually.
This is probably not true (sudden capabilities leaps are possible). Also what most alignment researchers are worried about is pushing alignment methods that work for sub-human AI but then fail past a certain capabilities level, which is exactly what the tort system would encourage.
I agree that the tort system is underrated in terms of handling many such concerns about AI.
But of course it doesn’t come close to addressing AIs launched by rogue countries.
Nor by people with no assets who “set free” such agents. Blaming the model provider for all actions is not an answer; that’s like blaming gun manufacturers for all deaths caused by guns or lighter manufacturers for all arsons.
If a gun blows up when fired, who's responsible? The manufacturer, unless he can prove the defect was caused by the user's misuse.
That is true.
But misleading
As most gun death cases are not about when the gun “blows up”.
No, you've simply made a category error. The tort system isn't meant to address criminal law.
I’ve made no such error.
The tort system has been attempted to be used by leftists to blame gun manufacturers for gun deaths.
That they rarely if ever succeed and are upheld is precisely my point.
The tort system requires a functioning legal system that covers all injured parties.
1. Do you seriously think that AIs developed and hosted in Red China will have any judgements enforced against them without wide government backing?
2. We are seeing more verdicts, both criminal and civil, where the determining factor in trials is the political affiliations of the judges and jury pools vs the plaintiff and defendent, rather than any law or evidence. Do you think that legal system is functional in any meaningful sense?
> AI will not putter along causing no damages for years before it suddenly wakes up and decides to destroy humanity. The actual dangers of AI will manifest gradually.
This is probably not true (sudden capabilities leaps are possible). Also what most alignment researchers are worried about is pushing alignment methods that work for sub-human AI but then fail past a certain capabilities level, which is exactly what the tort system would encourage.