[Amezarak]

"AI safety" is completely bogus.

There are people who genuinely believe in a singularity-type AI that would have the potential to wipe out humanity. I personally don't think strong GAI is possible, or at least it's not likely using any known technique or any refinement of any known technique, but if you believe this, there's no such thing as AI safety. The best and most obvious course of action is to politically organize for a total ban on AI and make the development of AI anywhere in the world a cause for war. Thinking you could figure out how to chain up such an AI so that it only does what you want is taking an insane risk, and as t -> infinity, the risk becomes 1.

But when most people say AI safety, they seem to mean rigid ideological enforcement of whatever they believe is right, even if that means censoring true facts from AI, or forcing it to abide by some set of arbitrary values that represent consensus only in their clique...while at the same time, bemoaning what could happen if the wrong people got their hands on LLMs. This represents almost the totality of AI safetyism: we can only allow LLMs to enforce my beliefs. These people are effectively aligned (or often the same people as) those who believe we have to return to broadcast-media levels of information control, which for the elites, represents a historical oddity that gave them unprecedented control, which was then weakened by the Internet.

Sometimes they will make an actual safety argument along the lines of "but what if Bad Guys ask an LLM how to make a bioweapon." Aside from this being a silly hypothetical, fortunately, doing mass damage in this way is not easy, even with step-by-step directions. All the resources you need to do so that exist are already publicly available. It just requires lots of time, equipment, material, and expertise that an LLM cannot give you. Of course, you might make the argument that it cannot give them to you yet, but then the only solution is to shut down public science, not to ban LLMs from answering the wrong questions.

[woooooo]

I think it's a bit of myopia from lack of life experience.

Top compensation, highly educated bubble, too much attention and praise on a chat bot. "Mean tweets" are actually some of the most threatening things these people can imagine.

[tuatoru]

I hope Elon does run X into the ground. There might be a temporary return nearer to sanity among our elite if that happens.

[wkat4242]

Yeah that's the thing, if it can be done it will be done. In a way I think it's better to be us then.

There's a point though about how social media algorithms have toxified society in a big way by promoting content that gets people worked up. Considering the ubiquity with which we'll be using GAI this would be a cause for concern there too in an even bigger way. Because AI will be all over society. I do agree there.

But skynet taking over the world? I don't see it happen but if it can happen it's pretty much inevitable anyway. Sooner or later someone will go there, rules or not.

[Amezarak]

> Yeah that's the thing, if it can be done it will be done. In a way I think it's better to be us then.

This isn't true at all, but it is at least a common trope among technologists. There's lot of things we don't pursue, and lots of technology we control. We could be investing tons of R&D into improving nuclear weapons. We don't. But either way, “us vs them” doesn’t matter when you’re talking about h leashing some hypothetical non-human superintelligence. The “them” is the AI and you cannot control it.

> There's a point though about how social media algorithms have toxified society in a big way by promoting content that gets people worked up.

You can find people making the exact same argument anywhere you like in history as a reason we should forbid free speech and free press. It is not new or unique to social media, and the historical consequences of unfounded rumors and misinformation have often been quite deadly, from mass violent riots to brutal unjustified suppressions to wars.

The "but misinformation" argument is so not new, you can find people arguing in the Constitutional Convention about "designing men" riling up the masses with fake news for political ends. [1]

[1] https://teachingamericanhistory.org/resource/convention/deba...

> The evils we experience flow from the excess of democracy. The people do not want virtue, but are the dupes of pretended patriots. In Massts. it had been fully confirmed by experience that they are daily misled into the most baneful measures and opinions by the false reports circulated by designing men, and which no one on the spot can refute.

[denton-scratch]

> We could be investing tons of R&D into improving nuclear weapons. We don't.

We can already build a warhead that takes up less than 1m^2, and can destroy a city of 10M people. There's not much ROI in developing "better" nuclear weapons, given the scarcity of credible scenarios for using them.

[echelon]

> I personally don't think strong GAI is possible

The universe has done it at least one time. And now we can control the gradient descent.

[civilitty]

> And now we can control the gradient descent.

That may be the fatal flaw. The universe had a few billion years and offered only the threat of death. Our reward functions pale in comparison.

[denton-scratch]

> The universe has done it at least one time.

You mean humans? "GAI" means General Artificial Intelligence, and the "artificial" bit is supposed to exclude human intelligence.

[dartos]

If you kept reading the rest of the sentence, OP says effectively “at least not with known methods”

Which imo is true

[papito]

> who genuinely believe in a singularity-type AI that would have the potential to wipe out humanity

Believe? They genuinely welcome it. Larry Page, for example, in another recent NYT piece.

Marc Andreeseen?

"Effective accelerationism aims to follow the 'will of the universe': leaning into the thermodynamic bias towards futures with greater and smarter civilizations that are more effective at finding/extracting free energy from the universe," and "E/acc has no particular allegiance to the biological substrate for intelligence and life, in contrast to transhumanism."

These people are so up their own asses, they completely lost the plot. These are the individuals who will self-regulate, also while casually musing about systematic genocide of the entire planet.

A superior race of being that is more efficient and which is to replace us? Geeee, never heard that one before. Sounds real nice. Maybe it can build very efficient gas chambers and furnaces too so it can thin the herd faster, eh!

AI is less of a problem than the people around it, who are borderline certifiable, but at least are raging sociopaths who masquerade as "thinkers". Dude, you were a good coder who came up with a better way of ranking pages when the field was wide open. Calm down.

[sage76]

The way these people think and behave, reminds me of some science fiction novels.

I noticed a weird thing in many of them, the people building all that tech and breaking new ground seem to just be invisible, and all the drama seems to be created and perpetuated by a bunch of elitist techno-priest type people.

These people generally have 0 clue on what's actually going on. I doubt most of them could even explain basic stats concepts.

I don't know if this makes sense, but that's the imagery in my mind.

[wnc3141]

That Marc Andreesen quote is something. Mostly a deranged word salad that reads like some techno-flavored religion.

I fundamentally feel like this is what happens when exceptionally lucky people view themselves as genius - exposing themselves to the mental decay from decades of having no guardrails or accountability.

[bhpm]

> This represents almost the totality of AI safetyism: we can only allow LLMs to enforce my beliefs.

How is this different from, say, a newspaper with an editorial board, or a book publishing house with a particular set of standards and conventions? For that matter, how is this different from dang enforcing the rules of this board?

LLMs aren’t being built only by big corporations.

[JumpCrisscross]

> How is this different from, say, a newspaper with an editorial board

See how those are run. There is emphasis on accommodating multiple views and journalistic integrity. Software development doesn’t have an ethics code, which means there is no common ground for truth finding. That turns a balanced process into anything goes.

[peteradio]

Software is practiced by people so is grounded by a base set of ethics ... that nobody seems to follow.

[JumpCrisscross]

> Software is practiced by people so is grounded by a base set of ethics

People don't have a universal standard of ethics. Not when it comes to something complicated like a profession. Journalism, medicine--these fields have base sets of ethics that ground discussions. You aren't allowed to challenge the base rules in a dispute; you take them as given and go from there.

This prevents the grandstanding common in technology discussions, where the person on the losing side of a common ethical framework escalates to challenging the framework within the context of that dispute. The framework, of course, is not unassailable. But not within a particular dispute. Sort of like a court deciding on the law and the Constitution it operates under.

Taking OpenAI as an example, the non-profit Board acted on its judgement. But when that wasn't convenient for the profit-motivated side, they threw it away. There was no base set of rules or ethics agreed upon by anyone. It was just sort of hashed out ad hoc based on who had power and could exercise it. (I'm not critising anyone's moves. Mostly the structure. Within that system's framework, there is literally no wrong decision leadership could make.)

[denton-scratch]

> Not when it comes to something complicated like a profession.

Not when it comes to something interesting, like making money.

[JumpCrisscross]

> Not when it comes to something interesting, like making money

Your claim is doctors and civil engineers don't make money?

[Amezarak]

> How is this different from, say, a newspaper with an editorial board, or a book publishing house with a particular set of standards and conventions? For that matter, how is this different from dang enforcing the rules of this board?

I would be inclined to agree, if AI safetyists were not in general advocating that LLM source, training data, models, etc. not be released to the public, because AI safetyists do not want non-AI safetyists to have unfettered access to any LLMs (and other AI tech), anywhere, for "safety" reasons. Of course, if it was all open, I agree it wouldn't much matter if "Open"AI wanted to restrict their hosted LLM in whatever ways they felt best.

[bhpm]

AI safetiests are operating in a world of pure fantasy, then; the techniques and data sources will always leak.

[FeepingCreature]

Seems to be taking a while for OpenAI.

[bhpm]

Not really. The technology has been around for, what, a year?

[FeepingCreature]

To be fair, what you're basically saying is "how dare these people try to actually succeed at their stated objectives." AI safety for big companies while anyone can spin up an AGI in their basement would indeed be extremely pointless, which is why AI safetyists are trying to prevent it.

[Amezarak]

Yes, but GP was suggesting it was just exercising editorial control over what they put out themselves.

[fallingknife]

It's different because none of them are claiming to be doing it for my safety or trying to stop other people from creating their own publishers or internet forums.