Previous posts in series:
2023,
2024,
2025.
Mostly-non-AI-related
Freedom of speech and anonymity online.
Lazy-loaded section (will type it out if there is demand for it).
A reminder about the theory of protest.
Lazy-loaded section.
Do you hate your government?
Wait, let me rephrase: do you hate your government more than average? Have no one who represents your interests in there? Do you see ridiculous and unjust laws passed one by one? Do you feel powerless, forced to choose between the piss and the poop every election cycle? Do you want to submit your feedback, but don't know what to do? There is a way!
>How?
Pay less taxes.
>What?
PAY LESS TAXES.
>But won't I, like, go to prison for that and stuff?
Not if you're smart about it. Here is the trick: you are under no legal obligation to pay income tax on income you did not earn! Work less and earn less. Not only you will pay less in absolute terms, but (in cunts with progressive taxation, which is the majority of them) you might also land in a lower tax bracket and have to pay even less, relatively speaking.
>But I would have less money then.
As long as money is instrumental and not an end goal for you this fact will be more than offset by you having more free time. After a certain very low threshold it's easier to cut your spending than to increase your earnings. Don't buy bullshit goods. Don't buy bullshit services. In fact, ideally don't buy any services period, with the exception of something you completely can't avoid, like internet. Embrace DIY. Buying something on the market hits you with heavy losses TWICE: first you exchange your time+skills for money, but lose a significant chunk (typically over 50%) as state taxes and company's revenue, then you exchange your money for goods/services, but lose a significant chunk (typically over 50%) as state taxes and company's revenue. By comparison, doing it yourself is 100% efficient: you are directly spending your time and effort on whatever you desire and become your sole beneficiary. Even stuff that you can't do yourself can be optimized. You cannot exactly etch and package a 5nm-process chip in your backyard, but if you really need latest hardware, then you can sell your previous generation machines directly to your friends at a reasonable price and recover a part of their cost, and if not, then you can buy slightly outdated but cheap hardware from your early adopter friends, all bypassing usual retail channels (which tax you twice).
>My company won't let me work less even if I offer to get paid less.
If they won't let you continuously lower your output, then you can always do it in discrete steps. Earn a comfortable buffer and resign, then find another job once the buffer starts running low.
>But muh gap on the resume!
If you are so worried about that (even though you probably shouldn't be: places with the strictest workaholic-like requirements will try to wring you out the most), then there's always the most extreme version: do all your working in one span and then retire early.
>Won't the government just shrug and carry on with lower spending?
Did you know that it was a crime not to be officially employed for more than 3 consecutive months in commiestans? Do you think they crafted that law just for shits and giggles? Paying less taxes is probably the second best thing you can do to hurt the slave owners. And it's totally safe for you, making it the best strategy overall. And you get more free time. It's a win-win-win situation.
>Won't the government just raise the taxes if many people start paying less?
Raising taxes is not a free action. Start taking away by force larger share of people's earnings and they will begin noticing. Then citizens start asking questions, "how much are we paying and what exactly are we getting in return?". This costs rulers their legitimacy. Once they fully run out of it... in the past it used to trigger revolutions, today in developed countries that's neither desirable nor possible (I think?). We'll see when we get there.
>My friends will have lower opinion of me if I become poorer.
Get better friends.
>My friends have a strong work ethic and will have lower opinion of me if I work less.
Talk to them, explain your reasons, chances are they will understand. If not... then it's up to you whether you want to spend your limited time trying to please others or pursuing your own interests.
>I'm waist-deep in debt, I'm paying off several mortgages, I have many dependents, I absolutely cannot cut my spending.
Well, you ended in your position as a consequence of your own choices, so I (genuinely) hope you're happy with the results. If not, then try to do better next time.
>It can't be all roses, what are the downsides?
1. Yes, it is a tradeoff. Now you can't be braindead about your purchases. You'll be paying for more efficient use of your decreased pool of money with more planning and thinking. Shouldn't be a problem if you're even slightly used to it.
2. It is not a feedback mechanism, it is a "fuck you" mechanism. Whatever functional remnants of the government cannot look at the new numbers and correctly interpret them as "policies X and Y are hated as expected and we need to revert at least one of them, voters are indifferent about policy Z, and policy T is surprisingly well-liked". They just see less money to embezzle redistribute.
>I tried not working and now I can't think of what to do!
You can't be serious. Spend time with your family, play a computer game, chat with your friends, exercise, take a walk in the park, watch a movie, listen to some music. Entertainment in this century is either the cheapest it has ever been, or even cheaper, depending on your piracy stance. If all else fails you can just lie on a couch, relax and take it easy. Even that is better than doing your best like an idiot so that the parasites can thrive.
>Why are you telling me this?
I don't really expect anyone to change their ways, although it'd be amusing if someone does. I just find it pleasant to write down my thoughts and observations and I also find it pleasant to read someone else's thoughts that I agree with and nod along. Maybe the one doing the nodding here will be you. Or maybe you will find a mistake and comment on that and that will improve my understanding.
AI-related
Last year in AI.
Lazy-loaded section.
Should we try to pause AI development?
I think by now plenty has been said about what happens if humanity fails to solve the AI control problem. So, what if we succeed?
>Isn't it glorious post-scarcity future?
Now hold on. There are two distinct problems in alignment. First is making sure AI will not go and do or want to do its own thing, the control problem (currently not solved, not even in concept form). Another is making sure everyone's interests are taken into account, the "who controls" problem (currently not solved, not even in concept form). You could bundle both together and say that alignment is not complete as long as both of them remain unsolved, but I think the distinction/separation is important because it's possible to have one without the other. Assume for a moment that the first one is somehow solved and then a superintelligence is created. What do you think happens next? Maybe I'll ask an even easier question, given all your knowledge about human psychology/history, which outcome do you think is more likely?
Whoever controls ASI at the moment it is created will ask it to:
a) grant everyone's true wishes where they agree, resolve conflicts and find a compromise where they disagree and create a peaceful and prosperous society
or
b) make him the god-emperor of the universe and delete anyone who opposes that
>Uh-oh.
Exactly.
>But there will be safeguards against that, right?
Yeah, we have an example from history. In 2015 when OpenAI was created its founders declared in its mission "The specific purpose of this corporation is to ensure that artificial general intelligence benefits all of humanity" and "The corporation is not organized for the private gain of any person". The company was incorporated as a non-profit, which was a perfectly reasonable choice given its mission, and also generated good publicity and helped attract talent. Later when its CEO started doing some questionable things the board of directors fired him and all was well.
>Cool.
Lol, just kidding. In real life its CEO threatened to take most employees with him to another company, then got reinstated, kicked inconvenient members from the board of directors and later looted stole restructured ~1 trillion of potential future valuation of the OpenAI non-profit (now a closed for-profit company) mostly into his and his cronies' pockets. It turned out legal documents are not a sufficiently good commitment.
>Not cool.
I know.
>But it's possible that next time will be different!
It certainly is. Maybe the future board of the first company which builds AGI will add some hardcore transparency measures like permanent audio-video records of every meeting with instant public dissemination, even though no company is trying to do that now, not even thinking about it. Maybe that board will consist mostly of risk-averse individuals, who won't try to grab more that their fair share, even though the C-suites are among the most risk-seeking people in the world. Maybe the government will step in and send some military/three-letter-agency types, who... uh, I don't know how that is supposed to help at all. Notice how many maybes there are. Do you still want to roll the dice, knowing what's at stake?
>So, what do you suggest?
Those who are advocating for a pause/stop in AI development say/think that it is extremely irresponsible, reckless and dangerous to develop AI capabilities without appropriate effort in alignment (and I agree with them). I also think it is even more irresponsible, reckless and dangerous to develop generic AI control capabilities without appropriate effort in controlling humans in control of AI. Without ironclad guarantees that no one will be able to usurp the power it is safer to aim for the most broad category, the paperclipper, instead of a tiny o(n^2) true win surrounded by an o(n) hellscape. It would be terribly rude to any civ caught in its near-lightspeed bubble of expansion, but what other choices do we have now?
>So you're saying we should accept status quo and do nothing? How convenient.
No. If I'm correct about all this, then the best strategy is not to do nothing, but rather to ally with the most unhinged accelerationists and work towards helping AI labs create the superintelligence as fast as possible, thwarting any attempts to pause ot stop the development of more powerful systems, so that the labs have less time to accidentally stumble on a solution to the control problem (without solving the "who controls" problem, because that's even harder). Seeing AI-concerned crowd split into two camps and push in opposite directions would be kind of sad, but that should not stop you from trying to push in the right direction.
>"Let's make sure everyone dies because the alternative might be even worse" doesn't make a good rallying cry.
True. But I think it's possible to reframe this statement into a version that many people would endorse while remaining completely honest.
>Weren't you advocating for a pause just a year or two ago?
I was, but I have three excuses: I'm not that smart in absolute sense, I'm not thinking about these issues every day and the easily digestible answer - the pause - seemed like a good idea. So, in total I have zero valid excuses. Still, it's better to correct yourself earlier than later.
>You can't be the first one to think of this. Where are the others who are worried about the same thing?
Asking real questions here. I don't know - while I saw others mentioning ASI-assisted dictatorship in passing as one potential problem among the others I am not aware of people who worry about this to the same extent as the loss of control. Maybe I'm just not that well-read on this subject. If there was some discussion on this topic that I missed which arrived at a reasonable answer, please inform me.
>How certain are you in your conclusions?
Less than fully certain, but still I don't see a flaw, which is why I'm posting, so that others may look at it.
>Dictatorship might be better than extinction.
First, go read history books about lives in Nazi Germany, Maoist China or Stalinist Russia. Second, that provides only a weak lower bound on how miserable it can get (i.e. it might get worse). Dictatorships of the past were not omnipresent. Superintelligence can help with that. Dictatorships of the past were not omniscient. Superintelligence can help with that. Dictatorships of the past were not nearly omnipotent. Superintelligence can help with that too. Finally, even though it would provide little solace to those living under dictatorships of the past, they were not eternal. I wonder if there's something which can help with that? And even if you're hoping to become one of the secret police members, here's a reminder: with ASI dictators don't need henchmen.
Personally, if I were to choose between Scenario A, 100% guaranteed extinction, and Scenario B, 3% chance of a single person or a small group of people obtaining control of unaligned to humanity, but otherwise steerable ASI plus the rest of the probability of any other outcome of your choice, I'd choose A any day. On the lower side I'd gamble with B at 1 ppm level and I didn't think hard for now about where my neutral point lies.
>What can we do about it?
On a societal level, the same thing as before. Talk to others, provide your honest thoughts and objections, listen to what they have to say and respond to that, in short raise awareness.
On a personal level, not much. Once it looks like creation of superintelligence is imminent (hard or impossible to tell even if you're lab insider, never mind a casual outside observer) if your risk estimate of a worse-than-extinction scenario remains unacceptably high you can opt out by unironically killing yourself, which is exactly as bad of an option as it sounds and I would recommend requiring some consensus at least in the fringe circles of AI-concerned that it's not just you being crazy here - there is no undo after all. If you expect the ASI to become extra malicious, then you would also need to simultaneously ensure complete destruction of every single copy of your DNA in the environment (shed skin cells, hairs, etc) to prevent creation of you-like entities (humans with your exact psychology, but different memories) by the ASI, and that is even more problematic in multiple ways. And don't even get me started on how you're supposed to sell this to your normalfriend relatives and acquaintances (protip: you can't).
>Any good news?
Humanity is not that competent. I still think it's highly likely we'll get the default outcome.
No comments:
Post a Comment