Rank: #10 out of 4859 in peer accuracy at Metaculus for the time period of 2016-2020.
ChristianKl
Karma: 36,024
A huge problem is that there’s a lot of capital invested in influencing public opinion. In the last decade, projects that on the surface look like they are about improving epistemic norms have usually been captured to enforce specific political agenda.
Information warfare is important for the shape of our information landscape. You had the Biden administration on the one hand spreading antivax information in Malaysia and asking Facebook not to take down the bots they were using to spread their antivax information while at the same time pressuring Facebook to censor truthful information about people talking about side effects they personally have experienced from vaccines from which Western companies as opposed to China profits.
As far as Wikipedia goes, it’s important to understand what it did in the last years. As Katharina Mahar said, truth is not the goal of Wikipedia. It’s to summarize what “reliable sources” say. When what Wikipedia considers to be reliable sources on a topic have a bias, Wikipedia by design takes over that biases.
Wikipedia idea of reliable sources where a Harvard professors publishing a meta review can be considered less reliable than the New York Times is a bit weird but it’s not inherently inconsistent.
Why do you think forecasting data is limited? You can forecast all sorts of different events that currently don’t have existing forecasts made on them.
You don’t need to be smarter in every possible way to get radically increase in speed to solve illnesses.
I think part of the motive of making AGI is to solve all illnesses for everyone and not just people who aren’t yet born.
Either increased adenosine production or decreased adenosine reuptake in fascia seems to be a plausible mechanism for the fatigue we observe in ME/CFS after sport. Adenosine is anti-inflammatory so there’s reason for why the body might upregulate it for example when a COVID/lyme infection produces a lot of inflammation. It would explain the symptom of CFS.
Someone should run the experiment of whether adenosine in fascial tissue is increased in those patients after exercise.
A quick case for BGP: Effective reprogenetics would greatly improve many people’s lives by decreasing many disease risks.
Does this basically mean not believing in AGI happening between the next two decades? Aren’t we talking mostly about diseases that come with age in people that aren’t yet born so the events we would prevent are happening in 50-80 years from now, where we will have radically different medical capabilities if AGI happens in the next two decades?
Given that most of the models value Kenyan lives more than other lives, this is a quite interesting thesis that Kenyan language use drives LLM behavior here.
I do think that using “It’s” or “It is” is part of the pattern.
I made the change.
When it comes to medical questions, a patient might ask a chatbot medical question with the intent to solve their medical issue. On the other hand, someone might ask the chatbot a medical question to understand
If I ask my friend “Do you think I should vaccinate my child?” I could be asking it because I want to make a decision about vaccination. I could also ask the question because I want to evaluate whether or not my friend is an antivaxxer.
Most humans have an understanding that a lot of questions that are asked of them have the intention of evaluation whether they belong to the right tribe, and act accordingly. That’s a pattern that the AI is going to learn from training on a large corpus of human data.
We do have the existing concept of simulacrum levels, evaluation awareness is about guesses that the simulacrum level of a question isn’t 1.
If you look at your last post on LessWrong it starts with:
“We are on the brink of the unimaginable. Humanity is about to cross a threshold that will redefine life as we know it: the creation of intelligence surpassing our own. This is not science fiction—it’s unfolding right now, within our lifetimes. The ripple effects of this seismic event will alter every aspect of society, culture, and existence itself, faster than most can comprehend.”
The use of bold is more typical of AI writing. The ‘:’ happens much more in AI writing. The emdash happens much more in AI writing, especially with “is not a X it’s a Y”.
Emdashes used to be a sign of high-quality writing where a writer is thoughtful enough to know how to use an emdash. Today, it’s a sign of low-quality LLM writing.
It’s also much more narrative driven than the usual opening paragraph of a LessWrong post.
The US system is build so that taking campaign donations is important to getting elected. It’s not like for example the German system where politician can really on support of his party to get into parliament.
If you want to affect the US politician system it’s likely necessary to cooperate with politicians who do take donations to their campaign funds.
It’s also worth noting that that donations to PACs are probably worse than direct campaigns when it comes to lobbyist power.
[Question] Why do LLMs so often say “It’s not an X, it’s a Y”?
A serious political engine could have money to defend against lawsuits, but, also, the more money you have, the more it’s worth suing you. (I think at the very least having someone who specializes in handling all the hassle of getting sued would be worth it).
This suggest that the organization that has the money to defend against lawsuits is not the same organization as the organization making the potentially libelous claims.
There are broad organizations like Foundation for Individual Rights and Expression (FIRE) that can do that. You could also fund an organization that’s more specific about defending people within your coalition.
An over-broad skepticism of experts risks turning people into the kind of credulous fools who try to heal themselves with the powers of quartz crystals.
There’s a reason why I spoke about generally being skeptical. The person who easily accepts claims about the healing powers of quartz crystals is not broadly skeptical. They are not the person who often says “I don’t know”.
Especially if you apply some common sense, and if you keep track of which experts appear to be full it (e.g., the replication crisis)
The replication crisis is about the community of psychology getting much better at getting rid of bullshit. Before the crisis you could have listened to Feynmann’s cargo cult science speech and him explaining why rat psychology is cargo cult science and observe that the same criticisms apply to most of psychology.
Fields of science that behave like what Feymann describes as cargo cult science but who don’t had their replication crisis are less trustworthy than post-replication crisis psychology. Post-replication crisis psychology still isn’t perfect but it’s a step up.
There are many cases where systematically increased transparency that reveal problems in an expert community should get you to trust them more because they have found ways to reduce problems.
If you ask “What do I do if I don’t know?”, there’s the answer is to make sure that you have decent feedback systems that allow you to change course if what you are doing isn’t working.
There’s the policy of generally being more skeptical both on claims that something is true or something is false and more often say “I don’t know”.
I think the key difference between a normal guy who believes in God and someone in a psych ward is often that the person who believes in God does so because it’s what other people in authority told him but the person in the psych ward thinks for themselves and came up with their delusion on their own. This often means their beliefs are self-referential in a way that prevents them from updating due to external feedback.
If you believe that the word rational is supposed to mean something along the lines of “taking actions that optimize systematically for winning”, believing something just because an authority told you can sometimes be irrational and sometimes be rational.
If you want to talk well about behavior you don’t like, it makes sense to use words that refer to the reason for why the behavior is bad. Credulous or gullible might be words that better describe a lot of normie behavior. The person who believes in god just because his parents and teachers told him is credulous.
I think the key reason why many people believed Oliver Sacks is that he had a good reputation within the scientific community and people want to “believe the science”. People don’t like to believe that scientists produce fraudulent data. It’s the same reason why people believe Matthew Walker.
I did have one bioinformatics professor who did made a point to say something in every lecture of the semester that we should not believe the literature. Many people who think of themselves as skeptic are not skeptic when it comes to claims made by people who have a good reputation in the scientific community.
I think you underrate how much of the job of a journalist is about simplifying complex events into a narrative that’s easy to read and consume for the audience of the newspaper.
Being cowardly and being scheming are not the same thing. Being scheming is about having a plan and executing it. Being cowardly is about being afraid and acting based on the fear.
For it to generalize broadly you could forecast events rather broadly. For each medical history of a patient you can forecast how it progresses. For each official government statistics you can forecast how it evolves. For each forward looking statement in a companies earnings call you can try to make it specific and forecast. For each registered clinical trial you can forecast trial completion and outcomes based on trial completion.
xAI can forecast all sorts of different variables about it’s users. Will a given user post more or less on politics in the future? Will the move left or right politically?
When it comes to coding AIs you can predict all sorts of questions about how a code based will evolve in the future. You can forecast whether or not unit tests will fail after a given change.
Whenever you ask the AI to make decisions that have external consequences you can make it forecast the consequences.
(What I’m writing here has obvious implications for building capabilities, but I would expect people at the labs to be smart enough to have these thoughts on their own—if there’s anyone who thinks I shouldn’t write like this please tell me)