Moderation Questions

In theory an LLM digesting content from all sources should realize that if every group feels superior, only 1 worldwide can actually be right (at most) , so the normal situation is one of a *delusion*, maybe because it's useful to keep social cohesion, about your own group superiority.

So an LLM will NOT automatically develop any racist/cultural bias in the sense of preferring any one specific group, rather it will know it's normal for most human beings to feel their group is superior, even if almost always it objectively can't be.

I explained why it will give nazi responses when prompted with unusual language typical only of nazi groups. But that doesn't generalize to more common racism

You need to program AI with good values. Programing it without good values would be an existential risk.

And we are back at the problem of what actually good values are, given than even in western societies there is a deep, insormountable abyss about that topic among society.

We truly disagree deeply, it isn't just theater. It might be theater at the margin for some people, but on topics like trans in sports and immigration and how much redistribution (if any) to have in society and in favor of which groups, we do deeply disagree in polarized ways.

We don’t have to agree on good values to agree on what good values for humans are to agree what good values for AI are.

AI should not be deciding whether trans people should compete in sports for instance.

But anyway if you truly think we can’t agree then any attempt to bring about AI by anyone is completely unacceptable evil that is anti-human and they should be eradicated with as much force as legally possible.

^ was trying out your writing style

The whole anti-trans debate has been completely blown up by the Epstein cover up. The two realities are entirely incompatible.

If it was possible, it would be actually reasonable to fear AI as an existential potential threat down the line and regulate it as we do regulate nuclear or chemical weapons. Your seemingly hyperbolic take isn't inherently wrong.

But it's not possible because it's much easier to train and run AIs , everyone can do it, and you already have models around that people can download and improve upon.

Also, if you tried to strictly regulate/ban it somewhere, impeding access to oversea internet is only possible (and in a fauly way) with china-style oppressive controls of the population.

And also, if you hyper-regulate/ban it and others don't, all the top minds move there to develop it, and you get the AI anyway and it's other people who have it.

So you basically have to go allin and hope for the best, even if i am not minimizing the risks at all.

europe will do what it always do: exceed in regulations without achieving anything of value other than pushing development outside the EU, while keeping the totality of risks anyway.

We can hope the USA avoid that kind of mistake: the senate voted 99-1 to expunge a federal AI regulatory framework from the BBB so the hope is well-founded i think, for now.

i am not sure what you mean by this.

We do. For example, should AI value *all human lives the same* or not? and it not, which weights should it use?

Keep in mind that "all lives are worth the same" is false in every country, we do not legislate believing that. But a lot of people always claim that. So what do we do? and this is a very basic and fundamental element that will inform the totality of AI policy considerations in *any topic*.

Because you literally cannot giv any policy advice if you cannot weight human lives in a quantitative way. You cannot advice on speed limit, or food regulation, or anything else.

Then add the freedom element. Should AI be utilitarian or not in its assessements? and why? those are fundamental value questions that you need an objective, written down answer to before you can literally be of any use for policy.

The anti trans movement was framed as a war to protect our children and our women yet the reality of the Epstein cover up is that the current administration is either involved and complicit in the trafficking and abuse of untold young women and children, or they knowingly exploited the trafficking and abuse of untold young women and children to get elected

You can't pretend to be anti-trans to protect womens sports and not be ready to impeach Trump over the Epstein cover up. Its thoroughly incompatible

Well we can probably all agree that there are certain things AI should not be able to do autonomously, like physically kill people. And we probably prefer AI in the advisory role over being in the control seat. As far as weighing human lives equally, I’m not sure why we would want our own national culture to be weighed over others.

I get that it’s more complicated, but I don’t think there is as much wiggle room for values as we would have in humans, given the immense power AGI would have.

If both parties defended epstein and sent their leaders to the island as the conspiracy goes, then that doesn't matter. Whatever you vote, you vote for pedophile elite island, if you believe in the conspiracy. They are literally all in it according to the conspiracy.

So you can still do LESS DAMAGE to kids by being anti-trans in sports and anti mutilation of minors and anti sterilization of minors with drugs.

If everyone in politics is a monster, there is no incoherence to vote for the monster that "only" has fun personally on a pedo island than for the monster that does the same on the same island AND ALSO wants to mutilate and sterilize thousands of minors per year. + ruin women sports.

No incoherence at all even within their conspiratorial minds.

If we assessed that AI risk was high enough then it would be worth it to even enter nuclear war over. X risk from nuclear war is far lower than x risk from AI.

AI might be the biggest existential threat ever because it could potentially mean even just one sociopath could destroy the planet.

I was actually serious in thinking that AI risk is bad enough to justify some very violent responses, but I just think that trying to get AI as fast as possible is the best way to go about it.

That makes Trump’s attacks on science research all the more evil and dangerous.

Not necessarily about culture , i was thinking more broadly also within a country (young vs old, criminals vs normal people and so on).

But rest assured that most countries will value their own citizens much much more than foreigners, and not because of culture but because that's literally the reason in the first place countries exist.

Now you can write rules about complicated weaponry that only the military have and rule out autonomous AI being allowed to kill sure. But maybe the ukrainians won't comply if they need to send a drone that could lose coms and they would love the idea of the drone being able to do some damage anyway when coms fail, wouldn't they? actual real needs will always trump ideological positions when it's easy to do that in practice.

Woul you put Zelensky up to trial for war crimes if he disattends rules about autonomous AI weaponry and that's how he manages to take out russian generals 2000km within the border say in 2027? I wouldn't.

And surely Russia would disregard those rules as well as China would if it gains anything by disregarding them.

Why would anyone believe the "both sides" nonsense at this point?

Epstein died under Trump, they released nothing
Biden won the election and the conservatives stoked the conspiracy that the Dems were protecting themselves despite that Epstein happened under the Republican eye
Trump made Epstein a big part of his 2024 campaign
Now they are coving it up, again

There is no point Republicans haven't been exploiting this with lies and deception. You'd have to actually be ******ed to believe anything from the Trump administration at this point

You're not going to believe this, but...

Yeah I get it. Having the same people who pedaled a conspiracy to me cover it up twice and lie to my face just makes me want to shout "BOTH SIIIIIIIIDES"

Why would Obama do this?

who pedaled what in this forum? I just told you how people on the right who care about Epstein for real think because you asked.

only low IQ people ACTUALLY believed Epstein pedo island narratives.

the right exploited that for some Lols and some votes at the margin of society and like everything else, if it works it's fine, who gets the idiots wins the prize, that's universal democracy, the model you love, can't blame people who play to win the game whose rules your side wrote.

ofc Epstein might have provided escorts and drugs to elite clients here and there but that's very normal

we wrote the rules of what

But there's certainly more content online containing bias against Jews and blacks/Africans than against any other groups, so wouldn't it be likely to be biased against those groups?

Actually you can, if you believe everything Trump and his minions say, even though it completely contradicts what they said yesterday.
This seems true of most MAGAs.

"The Party told you to reject the evidence of your eyes and ears. It was their final, most essential command."

of universal democracy, expanding the franchise to every last person when only people with property voted at the beginning.

nah because there is a ton of content defending those groups as well

Do both sides have a photo like this one?

it’s not universal. non-citizens, felons, people under 18, all cannot vote.

and the people that expanded it most is the conservative party.

you’re just creating a grand historical narrative that doesn’t exist so you can draw neat little lines. what’s the point of that. it literally tells us nothing about history or modern day political struggles.

not even maga wants to change it back to that.