LessWrong 2.0 Reader

The current number for slaves in the world in something like 50 million. It's a significant amount of people. In prostitution, you unfortunately have women that are enslaved with violence and other forms of dominance that are not just about the poor bargaining power of the prostitutes.

Slavery was abolished and remains abolished through dominance:

• first by getting outlawed by the Northern US and Great Britain, who drew strong economic benefit from higher labor prices due to them industrializing earlier for geographic reasons,
• secondly by leveraging state dominance during the great depression to demand massive increases in quality and quantity of production, to make it feasible to maintain a non-slave-holding society without having excess labor forces being forced to starve,
• thirdly, endless policies that use state violence and reduce the total fertility rate as a side-effect,

Throughout most of history, there has been excess labor, making the value of work fall down close to the cost of subsistence, being only sustainable because landowners see natural fluctuations in their production and therefore desire to keep people around even if it doesn't make short-term economic sense. This naturally creates serfdom and indentured servitude.

It's only really prisoners of war (e.g. African-American chattel slaves) who are slaves due to dominance; ordinary slavery is just poor bargaining power.

Death in Damascus is easy, but boring.

Doomsday Argument is not a decision theory problem... but it can be turned into one... I think the trolley version would be too big and complicated, though.

Obviously, only problems with discrete choice can be expressed as a Trolley problems.

Wow, XOR-trolley really made me think

Dominance underlies the things that can be done most efficiently with dominance. The moment dominance is no longer the most efficient force, it collapses, because in the vast majority of cases, dominating others takes a lot of time, energy and effort. This is actually how and why slavery (pretty much the most powerful example of dominance) was abolished: it started to make less economic sense than Bargaining (paid employment of freemen) and just Getting Things Done (through better tools and ultimately machines), so even its most ardent supporters became dispirited.

Two things to note.

First - I feel like putting every occupation in the same pile and deciding are you for or against licensing isn't helpful? I personally don't need licensed lawnmower, but I would very much prefer licensed doctor. The cost of mistake in two occupations differs a lot and can be used for a threshold which jobs should require a license.

Second - there should be a difference between doing a thing to yourself (argument can be made even that here we shouldn't have any limits), doing things for free to your friends/relatives with their full knowledge of your skill level and experience (most of the non life-threatening things can probably be allowed here) and selling your craft for money.

Interesting, I really hope TMS gains more acceptance. By the way, according to studies, ECT (the precursor of TMS) is even more effective, though it does have more side effects, due to the required anesthesia, and it is gatekept even more strongly. In my youth I suffered from depression for several years, and all of this likely would have been avoidable with a few ECT sessions (TMS wasn't a thing yet), if it wasn't for the medical system's irrational bias in favor of exclusively using SSRIs and CBT. I think this happens because most medical staff have no idea how terrible depression can be, so they don't get the sense of urgency they'd get from more visible diseases.

Pre-registering bcbe94590cf03bb7fbcb1ef04e0f779b1e39037d247ebb83b82ee18d5bcbd59f

A related thought: an intelligence can only work on the information that it has, regardless of its veracity, and it can only work on information that actually exists.

My hunch is that the plan of "AI boostraps itself to superintelligence, then superpower, then wipes out humanity" relies on it having access to information that is too well hidden to divine through sheer calculation and infogathering, regardless of its intelligence (ex: the location of all the military bunkers, and nuclear submarines humanity has), or simply does not exist (ex: future Human strategic choices based on coin-flips).

Most AI Apocalypse scenarios depend not only on the AI being superhumanly smart, but being inexplicably Omniscient about things that nobody could be plausibly Omniscient about. 

Thanks for the insight on the locality check experiment.

For inducing refusal, I used the code from the demo notebook provided in your post. It doesn't have a section on inducing refusal but I just invert the difference-in-means vector and set the intervention layer to the single layer where said vector was extracted. I believe this has the same effect as what you described, which is to apply the intervention to every token at a single layer. Will checkout your repo to see if I missed something. Thank you for the discussion.