ChristianKl

Karma: 35,363

ChristianKl Apr 23, 2025, 2:18 PM
2 points
0
in reply to: cherrvak’s comment on: Please don’t throw your mind away
Are people intrinsically drawn to having the eye fooled by abstracted drawings?

If you want to apply this to alignment, the next question would be, is there something in the human nature that causes this, or would an AGI likely be drawn to similar effects?
If you gather knowledge about what’s intrinsically motivating AGI’s as well, that would be valuable for alignment research, because it’s about creating motivations for AGI’s to do things.
Did these sinister implications go over the heads of audiences, or were they key to the masses’ enjoyment of the songs?
You can reframe that question as, “Is this aspect of the Beatles songs aligned with the desires of the audience?”
Both of your examples about what people or agents in general value. AI alignment is about how to align what humans and AGIs value. Understanding something about the nature of value, seems applicable to AI alignment.

ChristianKl Apr 23, 2025, 10:52 AM
4 points
0
on: Are we “being poisoned”?
Asking “Are we being poisoned?” sets up the issue is an supotimal way. It frames very complex interaction as if they would be simple and with agency of an hostile agent.

I want to know what, specifically, is driving this problem (doesn’t everyone) and
Part of being a rationalist is accepting that there are questions where the available evidence doesn’t clearly answer the question.
David Chapman’s Nutrition: the Emperor has no clothes is worth reading to be aware of how irrationally the issue of nutrition is commonly approached. If you want to do better, understand it is valuable.
what an ordinary person can do to minimize the risk of serious illness.
I don’t know what you mean with the term “ordinary person”. Solution for someone with bad body awareness and who has a program salary are quite different then for someone with awesome body awareness who lives on minimum wage.
For many people, internal drives get them to eat iron if they are deficient at it. The dynamic is strong enough that Alicorn wrote. Experiential Pica. If you pay good attention for the internal drives and understand your drives you are going to eat enough iron.
For myself, I have effervescent tablets for Magnesium, Calcium and Iron. With Magnesium in particular there are days where I clearly feel like a Magnesium tablet would be tasty and days I don’t. Since I switched my diet, I don’t find them appealing anymore.
On the other hand, for someone who doesn’t really trust their impulses and has the cash to throw around (or healthcare that provides them with free access to testing), doing blood testing and supplementing iron if you need it is a great idea, as you can also do blood testing as Elizabeth lays out.
There are both ways that leverage high self awareness and ways that leverage having big financial resources.

ChristianKl Apr 22, 2025, 6:16 PM
2 points
0
on: The Uses of Complacency
It wouldn’t surprise me if people having a narrative like this make it less likely to suffer from akrasia.

ChristianKl Apr 16, 2025, 12:13 PM
2 points
0
on: Should AIs be Encouraged to Cooperate?
I would expect by their natures of how AI gets deployed that a lot of cooperation happens pretty soon.
Let’s say I want my agent to book me a doctor’s appointment because I have an issue. I would expect that it’s fairly soon that my AI agent is able to autonomously send out emails to book a doctor’s appointment. On the other side, it makes a lot of sense for the doctor’s office to have an AI that runs doctors’ appointments on their side.
In Germany, where I live, how soon the appointment is can depend on factors like how urgent the appointment is and the type of insurance the patient is using.
This is a simple case of two AIs cooperating with each other so that the doctors’ appointment gets scheduled.
Interestingly, the AI of the patient has the choice whether or not to defect with the AI of the doctor’s office. The AI of the patient can lie about how critical the condition of the patient happens to be to get an appointment that’s sooner.
There’s no need for AI that is near human capabilities to be around for the ability of AI to negotiate with each other to be relevant. All the major companies will need training data to optimize how their AI negotiates fairly soon.

ChristianKl Apr 13, 2025, 9:47 AM
10 points
8
in reply to: beyarkay’s comment on: Experts have it easy
If you go to Yudkowski and say “Hey, I want to build an AGI, can you please tell me how to do so safely?”, he will answer “I can’t tell you how to build an AGI safely. If you are going to successfully build an AGI you are likely to kill us all. Don’t build an AGI”.
Yudkoswky is self-aware that he doesn’t have the expertise that would be necessary to build a safe AGI. If you then search for an expert who gives you some advice about how to build your AGI and you succeed in building your AGI and kill everyone, that was a trap of your own making.
If you want a past example, David Chapman´s article about Domestic Science is great. More recently, you had a flatlining of childhood obesity rates in the second Bush administration. Afterwards, Michelle Obama started a huge program to improve childhood nutrition through various measures recommended by nutrition scientists. After those programs childhood obesity rates were rising again.
Scientists studying nutrition do believe themselves to be experts, but seem to give advice that produces more problems than it solves.

Reality doesn´t just reward you for having put in a lot of effort.

ChristianKl Apr 13, 2025, 8:54 AM
3 points
1
on: Experts have it easy
Before asking for help, they made a wrong decision, which maybe solved their short term problems but shot them in the foot in the long term. This bad decision made other bad decisions more likely until they were helplessly stuck in a trap of their own making.
It’s worth noting that there are domains where there are no experts. In those domains people who explore them still tie themselves up.
A class of people who thinks of themselves as experts but doesn’t really have a clue is the most dangerous when it comes to trapping themselves in traps of their own making.
AGI development is the obvious example for LessWrong and other examples are better left as exercise to the reader.

ChristianKl Apr 12, 2025, 8:05 PM
5 points
0
in reply to: samuelshadrach’s comment on: Distributed whistleblowing
The US does not have laws that forbid people who don’t have a security clearance from publishing classified material. The UK is a country that has such laws but the first amendment prevents that.
I don’t think that chosing jurisdiction in the hope that they will protect you is a a good strategy. If you want to host leaks from the US in China, it’s possible that China’s offers to surpress that information as part of a deal.
4chan has a single point of failure. If the NSA would be motivated enough to burn some of their 0-days, taking it offline wouldn’t be hard.
Taking a decentralized system with an incentive structure like ArDrive down is significantly harder.
Attacking ArDrive is likely also politically more costly as it breaks other usages of it. The people with NFT that store data on ArDrive can pay lobbyists to defend it.
Just convincing the developers is not enough. You also need the patch they created to be accepted by the network, and it’s possible for the system to be forked if different network participants want different things.
Torrents are also bad for privacy everybody can see the IP addresses of all the other people who subscribe to a torrent.
For privacy onion routing is great. Tor uses that. Tor however doesn’t have a data storage layer.
Veiled and the network on which Session runs use onion routing as well and have a data storage layer.
In the case of Veiled you get the nice property that the more people want to download a certain piece of content the more notes in the network store the information.
As far as creating public knowledge goes, I do think that Discord, servers and Telegram chats serve currently as social media.

ChristianKl Apr 12, 2025, 11:07 AM
9 points
0
on: Distributed whistleblowing
When it comes to meta data and plain text extraction it’s worth noting that meta data can both be used to verify documents and to expose whistleblowers. If a journalist can verify authenticity of emails because they have access to the meta data that’s useful.
- guidelines for latest hard-to-censor social media
  to publish torrent link, maybe raw docs, and social media discussions
  guidelines must be country-wise and include legal considerations. always use a social media of a country different from the country where leak happened.
The Session messenger is probably better than country-specific social media.
country-wise torrents (not sure if this is needed)
- torrents proposed above are illegal in all countries. instead we can legally circulate country A’s secrets via torrent within country B and legally circulate country B’s secrets via torrent within country A. only getting the info past the border is illegal, for that again need securedrop or hard disk dead drop if any country or org seals their geographic borders.
The US does have the first amendment. That currently means that all the relevant information about AI labs is legal to share. It’s possible to have a legal regime where sharing model weights of AI gets legally restricted but for the sake of AI safety I don’t think we want Open AI researchers to leak model weights of powerful models.
The main information that’s currently forbidden from being shared legally in the US is child pornography, but whistleblowing is not about intentionally sharing child pornography. When it comes to child pornography, the right thought isn’t “How can we host it through a jurisdiction where it’s legal”, but to try to avoid sharing it.
While sharing all the bitcoin blocks involves sharing child pornography, nobody went after bitcoin minors for child pornography. People who develop cryptography who don’t intend to share child pornography generally has not been prosecuted.
Torrents are not a good technology for censorship-resistant hosting. Technology like veilid, where a data set that gets queried by a lot of people automatically gets distributed over more of the network is better because it prevents the people who hosts the torrents from being DDoSed.
If you just want to host plaintext, blockchain technology like ArDrive also exists. You need to pay ~12$ per GB but if you do so, you get permanent storage that’s nearly impossible to censor.

ChristianKl Apr 12, 2025, 9:54 AM
2 points
−2
in reply to: ryan_greenblatt’s comment on: Why do misalignment risks increase as AIs get more capable?
I don’t think it’s a spectrum. A spectrum is something one-dimensional. The problem with your distinction is that someone might think that there are safe from problems arising from power seeking (in a more broad definition) if they prevent the AI from doing things that they don’t desire.
There are probably three variables that matter:
1. How much agency has the human in the interaction.
2. How much agency has the AI agent in the interaction.
3. Does the AI cooperate or defect in the game theoretic sense.
If you have a low agency CEO and a high agency very smart middle manager that always cooperates, that middle manager can still acquire more and more power over the organization.

ChristianKl Apr 11, 2025, 4:45 PM
2 points
0
in reply to: ryan_greenblatt’s comment on: Why do misalignment risks increase as AIs get more capable?
Plenty of things I desire happen without me intending for them to happen.
In general, the reference class is misalignment of agents and AIs aren’t the only agents. We can look at how the terms works in a corporation.
There are certain powers that a CEO of a big corporation intentionally delegates to a mid-level manager. I think there’s plenty that a CEO appreciates his mid-level manager to do that the CEO does not explicitly task the mid-level manager to do. The CEO likely appreciates if the mid-level manager autonomously takes charge of solving problems without bothering the CEO about it.
On the other hand, there are also ways where the mid-level manager does company politics and engineers a situation so that the CEO giving the mid-level manager certain powers that the CEO doesn’t desire to give the mid-level manager. The CEO feels forced to do so because of how the company politics play out, so the CEO does intentionally give the powers out to the mid-level manager.

ChristianKl Apr 11, 2025, 3:19 PM
2 points
0
on: Why do misalignment risks increase as AIs get more capable?
What’s the difference between being given affordances and getting power? If you are given more affordances you have more power. However seeking is about doing things that increase the affordances you have.

ChristianKl Apr 10, 2025, 8:07 PM
7 points
0
in reply to: Julian Bradshaw’s comment on: birds and mammals independently evolved intelligence
I searched a bit more and it seems they don’t have personal relationships with other members of the same species the way mammals and birds can.
Personal relationships seem to something that needs intelligence and that birds and mammals evolved separately.

ChristianKl Apr 10, 2025, 5:03 PM
5 points
0
in reply to: SoerenMind’s comment on: Learned pain as a leading cause of chronic pain
The 2019 update add many codes that orthodox Western medicine disagrees with.
If someone wants Chinese medicine codes they got it in the update. Ayurveda got codes. Activists for Chronic Lyme got their codes as well.
The philosophy of that update seemed to be “If there anything that doctors want to diagnose, it should get a code so that it can go into standardized electronic health records.”

ChristianKl Apr 10, 2025, 2:52 PM
6 points
0
in reply to: Julian Bradshaw’s comment on: birds and mammals independently evolved intelligence
The Humboldt squid is an octupus that can coordinate to hunt together.

ChristianKl Apr 10, 2025, 2:51 PM
7 points
3
on: birds and mammals independently evolved intelligence
It’s important to be clear what we mean with intelligence in this context. The last common ancestors of birds and mammals lived 325 million years ago or roughly 200 million years after the Cambrian revolution. It was likely a lizard-like creature of maybe 20-30cm length.
It was likely intelligent enough to coordinate the movement of its four legs and make some decisions about what to eat. It had eyes. It probably had a brain with neurons. Maybe capable of a freeze response to avoid getting eaten by a predator but no fight/flight. It was likely much dumber than today’s mammals and birds.
Given the same building blocks of neurons and axions that you connect together to form a brain birds and mammals seem to have found different strategies of how to create higher intelligence.
Octopi who diverged before the Cambrian evolution would likely make a much better argument of independent development of intelligence.

ChristianKl Apr 10, 2025, 9:41 AM
5 points
0
in reply to: Knight Lee’s comment on: Short Timelines don’t Devalue Long Horizon Research
The key is still to distinguish good from bad ideas.
In the linked post, you essentially make the argument that “Whole brain emulation artificial intelligence is safer than LLM-based artificial superintelligence”. That’s a claim that might be true or not true. On aspect of spending more time with that idea would be to think more critically about whether that’s true.
However, even if it would be true, it wouldn’t help in a scenario where we already have LLM-based artificial superintelligence.

ChristianKl Apr 10, 2025, 9:17 AM
14 points
0
on: Learned pain as a leading cause of chronic pain
In 2019, the WHO’s added “nociplastic pain” (another word for neuroplastic pain) as an official new category of pain, alongside the long established nociceptic and neuropathic pain categories
It’s worth noting that in 2019 the WHO also added various diagnosis from Chinese traditional medicine. The process that the WHO uses is not about finding truth but to provide codes that allow healthcare providers to talk with each other and store diagnoses.

ChristianKl Apr 8, 2025, 6:35 PM
4 points
2
on: Misinformation is the default, and information is the government telling you your tap water is safe to drink
I did agree with the framing of the problem at the beginning but it’s unclear to me why the conclusion is something besides “focus on how we can produce more reliable sources of information”.
If you take an issue like ‘people don’t believe in CDC guidance’ it’s possible to reform the CDC in a way that forces it to give explicit reasoning that’s backed for all their recommendations.
Instead of fighting misinformation it would be possible to focus on improving the processes inside institutions to make those more reliable.

ChristianKl Apr 7, 2025, 9:19 PM
2 points
0
in reply to: RussellThor’s comment on: How much progress actually happens in theoretical physics?
You made statements that you know why the field is stuck. That both contains an assertion of the field being stuck and you knowing the explanation is due to diminishing returns. I made no claim that I know the answer to either of those questions.
#1 There is not a significant group of young capable researchers saying we are taking things in the wrong direction, but a smaller number of older ones.
If you think that theoretical physics is going in the wrong direction and are unlikely going to get to research in a way you think will make progress, there are good reasons for not being not making your PHD in theoretical physics. The strongest disbelievers filter themselves out.
But even among the people who are actually in the field, I don’t see a good reason why you would publically see signs.
You might ask PHD students: “If you wouldn’t need to seek grants and would get a lab with 5 million dollar per year, would you pursue the same research agendas as you are currently do or would you pursue research for which you wouldn’t get grants in the current academic environment?”
It might be interesting to ask that question for researchers in every field and see the responses, but unfortunately I don’t know of any source that asks such a question in a good way.
I followed Hinton et all before they became famous. I saw the paradigm thing play out in front of my eyes there.
It’s certainly possible for their to be fields where it can be obvious to outsiders that a particular paradigm changing approach will work, but that’s not necessary for there to be superior paradigms that could be persued.

ChristianKl Apr 7, 2025, 7:43 PM
2 points
0
in reply to: NormanPerlmutter’s comment on: NormanPerlmutter’s Shortform
While the Trump administration campaigned on deporting very many people till now they haven’t. Instead they seem to pursue a policy of taking actions that produce fear I potential immigrants and managed to drastically reduce the inflow illegal immigrants that way.
Both mainstream media and the Trump administration like it if the policies look scary.
If you want to decide whether it makes sense to be personally scared it’s important to think through your personal threat models.