All of samath's Comments + Replies

Is this an accurate and helpful summary in layman's terms?

  • Training against an undesired behavior such as deception with a straightforward penalization approach is like giving the AI an instinctive aversion to it.
  • Such undesired behaviors would be useful in problems AIs will be asked to solve.
  • If an AI is smart enough, it will be able to translate some such problem to another domain where it lacks the instinct against deception, solve the problem there, and translate it back.
  • Once the AI notices this trick, it can overcome these aversions any time it wants.
2Viliam
Yes, but maybe with a bit more emphasis that this is also not a deception. Maybe:

Thanks for all the work! Can you turn those text lists into tables? It's hard to compare at a glance when the numbers aren't lined up.

Here's the relevant (if not directly analogous) Calvin and Hobbes story.

(The arc continues through the non-Sunday comics until February 1st, 1990.)

Thanks for reaching out! As a LW lurker, I've felt a bit of unease when I first heard about Intentional Insights or read the one HuffPo article, and it's taken me a while to discern where that unease came from.

One natural interpretation is that I'm not as comfortable with emotional appeals, and since that's expressly what you aim to give, it's going to rub someone like me the wrong way. If that's the case, InIn's community will just be a different subset of the population, probably bigger like you hope, and we should accept that.

A more concerning interpret... (read more)

1Gleb_Tsipursky
Appreciate your insightful commentary! I'm not comfortable with emotional appeal myself - that's not what I respond to personally. The same goes for the core participants of Intentional Insights. However, we're aware of the typical mind fallacy, and the large majority of the population is not like us. Like any science popularizers - think of Bill Nye or Neil DeGrasse Tyson - we're trying to speak to the broad population, and there's some evidence we are getting through (see the last sections of this document). Regarding the style of LW, I hear what you're saying. I think it's important to note that the goal of InIn is not primarily to cultivate Less Wrongers, but to raise the sanity waterline broadly. If we can spread good memes from Less Wrong, then we've done our primary job. We do aim, as a secondary component of our work, to cultivate people into aspiring rationalists and effective altruists, and in fact we have already had some accomplishments there, with people starting to read the Sequences, donate effectively, etc. Thanks for the suggestion on the magazine, will check it out!

Hmmm, I think a better word than "fantasy" here is "dystopia." Robertson is painting a bleak picture of a world where without moral authority, like the (much longer) bleak depiction of say, Fahrenheit 451 of a world without intellectual freedoms. Again, the natural reaction to reading Fahrenheit 451 or hearing Robertson isn't gleeful cackling, but shocked horror. "Something ain't right."

Sorry, but I'm guessing you don't spend much time around religious conservatives like Robertson. It's actually quite common among them to reason philosophically like this, mainly due to the emphasis on Christian apologetics. I'm sure Robertson has come across an argument of this form before and just reworked it for this.

Let me offer some more evidence. Listening to a recording of it, there are some chuckles in the audience at the beginning, but it grows silent by the end as most people grow more disgusted. The natural reaction, right in his last line, is, ... (read more)

9Jiro
The fantasy isn't mainly that Robertson likes torturing atheists or thinks his audience does. The fantasy is that their own atheism is responsible for them being tortured and that the awfulness of that demonstrates that atheism is awful. Whether his audience likes hearing about atheists suffering is a side issue. That's a bad comparison because Nazis did not believe that Jews could or should give up being Jews.

What I meant is that you could easily just define your ethics to include by definition "murder is bad" and it'd satisfy all of the other criteria (assuming you could coherently define murder). But if I imagine myself telling Robertson (or somone similar) that, they'd ask how I came up with that rule and why someone else couldn't just come up with the opposite rule "murder is good" and so it was just an arbitrary choice on my part.

8hairyfigment
When Lovecraft invented the blind idiot god Azathoth (as the human narrator calls it), he was likely just taking the Old Catholic/Aristotelian view of God and imagining what that might look like given the universe we live in. Azathoth maintains existence by sitting at its center surrounded by vast demonic dancers. There's a mediator, here called Nyarlathotep rather than Jesus or the Pope, who claims to somehow be doing Azathoth's will when he told humans to murder each other. I mention this because we would not consider N's commands morally binding, even in that scenario. We consider hypothetical deities moral or immoral based on whether or not they agree with "arbitrary" rules like not hurting people unnecessarily, not the other way around. Nothing else in the 'philosophical' account of God actually has moral significance. Nor can it provide a foundation for the claims that it sneakily assumes. So one big reason why I look down on Robertson's argument is that the charge he makes against atheists doesn't distinguish theism from atheism.

As someone who has spent a lot of time with religious conservatives, I've heard the sort of argument given by Robertson many times before. And they use it as an actual argument used against nihilism, which they tend to think follows directly from atheism. So Scott is completely right to address it as such.

I think Robertson conflates the two because he (and others like him) can't really imagine a coherent non-arbitrary atheist moral realist theory. Can anyone here give a good example of one that couldn't include what the murderer he depicts seems to believe?

0DanArmak
Why does Robertson, or anyone else, insist on moral realism? And what exactly does he mean by it? There seem to be different usages of "moral realism", which is confusing. The main two are: 1. Morals are an objective property of the universe, or possibly of mathematics (e.g. game theoretic cooperation), which can be deduced and agreed on, even separately from purely human concerns and attributes. So we can speak of objective morals. And if one believes that humans are typical of (evolved) intelligences, and that evolution removes behavior that is self-destructive or unstable, it's likely that common human morals are somewhat correlated with these universal morals. 2. Humans are very homogenous compared to all possible intelligent agents. Human moral beliefs, intuitions and actions are more alike than they are different. This shared core is objective or "real" in the sense that it is independent of any particular human or even any particular human culture. So we can speak of objective human!morals.
6hairyfigment
What does "non-arbitrary" mean, and why is it a virtue? More, why does Robertson's religion have this property, when plainly no moral claims can logically follow from the existence of some deity unless we start by assuming a connection?
-1TheAncientGeek
But a significant number of atheists are nihilists and relativist, and the reasons seem to be the same...they can't imagine naturalized objective ethics. And the common problem is stopping at personal incredulity rather than researching what anyone else has come up with. Show me the kind of atheists who is a moral nihilist, and I'll show you the kind of atheist who disdains philosophy.
2tog
How about moral realist consequentialism? Or a moral realist deontology with defeasible rules like a prohibition on murdering? These can certainly be coherent. I'm not sure what you require them to be non-arbitrary, but one case for consequentialism's being non-arbitrary would be that it is based on a direct acquaintance with or perception of the badness of pain and goodness of happiness. (I find this case plausible.) For a paper on this, see http://philpapers.org/archive/SINTEA-3.pdf
0seer
Well the fact that it appears to be impossible to get two LessWrongers to agree on whether a given moral theory is coherent and non-arbitrary is not encouraging in that regard.

In an article proclaiming the transcendent use of complicated, modern statistics in baseball, and in particular, one called "WAR" (wins above replacement):

I'm not a mathematician and I'm not a scientist. I'm a guy who tries to understand baseball with common sense. In this era, that means embracing advanced metrics that I don't really understand. That should make me a little uncomfortable, and it does. WAR is a crisscrossed mess of routes leading toward something that, basically, I have to take on faith.

And faith is irrational and anti-intellec

... (read more)

I downvoted for equivocating between faith and probability.

A doctor walking in with a syringe full of something that he says will prevent measles I would assign a much higher probability to being true than Bob from the car mechanic walking in with a syringe full of strange liquid that Bob says will prevent measles.

Essentially this seems like the fallacy of gray.

2MinibearRex
I'm not really sure that counts as faith. Faith usually implies something like "believing something without concern for evidence". And in fact, the evidence I have fairly strongly indicates is that when I step into an airplane, I'm not going to die.