id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
ce1aac7f-a3e5-4931-9066-9201fc0ac0b1 | trentmkelly/LessWrong-43k | LessWrong | OHGOOD: A coordination body for compute governance
Core to many compute governance proposals is having some kind of register that records who owns AI chips.
This article explores how this register could be implemented in practice, outlining an organisation that maintains such a register and its necessary processes. I... |
0c8fa67e-3e7b-44cb-98ac-172a9c4fc54b | trentmkelly/LessWrong-43k | LessWrong | AMA on Truthful AI: Owen Cotton-Barratt, Owain Evans & co-authors
We just published a long paper on Truthful AI (overview post). We’ll be running an Ask Me Anything on truthful AI from Tuesday (October 26) to Wednesday (October 27) at this post.
You may wish to ask about:
1. Anything discussed in the Truthful AI pa... |
d0c36270-b0e4-4e9a-b7a8-485a3ca11798 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post2457
I’m looking for alignment techniques that are indefinitely scalable and that work in any situation we can dream up . That means I spend time thinking about “exotic” problems — like AI systems reasoning about their own training process or about humanity’s far future. Yet I’m very optimistic about finding p... |
f12958e9-8f73-4ede-96db-1c24156987dc | trentmkelly/LessWrong-43k | LessWrong | The First World Takeover
Before Robin and I move on to talking about the Future, it seems to me wise to check if we have disagreements in our view of the Past. Which might be much easier to discuss - and maybe even resolve... So...
In the beginning was the Bang. For nine billion years afterward, nothing much happen... |
d547d63a-9804-47e4-972e-77c8efe5bd08 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3190
The user cousin_it has pointed out a problem with the counterfactual Oracle idea: the Oracle AIs may form a "bucket chain" bringing back a dangerous message from a future UFAI (unfriendly AI). This is certainly a problem, and though there are ways of reducing the risk, there doesn't seem to be any clean s... |
2342b2ca-e0c9-46de-97a9-64c5d3fedfbd | trentmkelly/LessWrong-43k | LessWrong | Reminders without times
Many times in life, a person wants to do a thing at a different time. For this to happen, the person has to remember about this, at the different time. We have very good systems this, as long as the time can be specified in terms of time. That is, if you can say ‘I want to do this in three day... |
6661b358-b9bd-487f-94e3-1031629cb4c8 | trentmkelly/LessWrong-43k | LessWrong | What is the difference between robustness and inner alignment?
Robustness, as used in ML, means that your model continues to perform well even for inputs that are off-distribution relative to the training set.
Inner alignment refers to the following problem: How can we ensure that the policy an AI agents ends up with... |
eec8c6f4-4f82-4068-bfcc-4d29913f4515 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Rationalist Housewarming at Isengard (Melbourne)
Discussion article for the meetup : Rationalist Housewarming at Isengard (Melbourne)
WHEN: 15 June 2013 08:00:00PM (+1000)
WHERE: 2105 / 5 Caraval Lane Docklands
The first and only (as far as we know) Australian Rationalist share house is finally here, our l... |
d5070350-8e59-4724-8ee4-fbe07b3b512b | trentmkelly/LessWrong-43k | LessWrong | A brief review of The Scout Mindset
I've been reading blogs like Less Wrong for almost a decade. So then, a lot of what was said in this book wasn't new to me. However, I still really liked it.
I'm of the opinion that in order to deeply understand a topic, it's not just enough to understand it conceptually, you have ... |
9c41b62a-bbfe-47b4-9b86-d6c7a484bbcf | StampyAI/alignment-research-dataset/arxiv | Arxiv | Constructing Unrestricted Adversarial Examples with Generative Models
1 Introduction
---------------
Machine learning algorithms are known to be susceptible to adversarial examples. Samples from the dataset can be perturbed to mislead cutting edge classifiers [szegedy2013intriguing](#bib.bib1) ; [goodfellow2014expl... |
4faa75f9-c2db-40fc-8884-b4e283a60366 | trentmkelly/LessWrong-43k | LessWrong | AIXI-style IQ tests
"Measuring universal intelligence: Towards an anytime intelligence test"; abstract:
> In this paper, we develop the idea of a universal anytime intelligence test. The meaning of the terms “universal” and “anytime” is manifold here: the test should be able to measure the intelligence of any biologi... |
33ad2d69-08af-4e31-bf74-6870ad45b082 | trentmkelly/LessWrong-43k | LessWrong | Anki Library
I have a few anki decks that I like, and am making more based on books I'm reading (currently doing one on Good Calories, Bad Calories). They don't replace reading the book, could probably supplement a summary, but definitely helps recall on specific details (at least for me).
Other people have some anki... |
10639669-074e-440e-92f5-b10d1a183651 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | DeepMind and Google Brain are merging [Linkpost]
The new organization, called Google DeepMind, will be led by Demis Hassabis (current CEO of DeepMind). Jeff Dean (co-founder of Google Brain) will be the chief scientist of the new organization.
 showed a potential technique for removing superposition from neurons using sparse coding. The original post shows ... |
d423788c-f906-46ef-8a2b-926998d9ca9a | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Deceptive Misaligned Mesa-Optimisers? It's More Likely Than You Think...
hi so the previous video explained what
mesa optimizers are why they're likely
to happen in advanced machine learning
systems
and why the optimal strategy for a mesa
optimizer
is to pretend to be aligned during the
training process and then t... |
1f7d23d2-cef8-4b90-b0c4-f0f182a0fdcf | trentmkelly/LessWrong-43k | LessWrong | Logical Foundations of Government Policy
Crosspost from the EA Forum.
Why think about goalsets?
Societies need many distinct systems: a transport system, a school system, etc. These systems cannot be justified if they are amoral, so they must serve morality. Each system cannot, however, achieve the best moral outcom... |
4f92eef9-cec2-4262-8113-a751469b0f18 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Christiano (ARC) and GA (Conjecture) Discuss Alignment Cruxes
The following are the summary and transcript of a discussion between Paul Christiano (ARC) and Gabriel Alfour, hereafter GA (Conjecture), which took place on December 11, 2022 on Slack. It was held as part of a series of discussions between [Conjecture](htt... |
7dec4d36-671d-4af6-b42c-75509e68dafe | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | How to Diversify Conceptual AI Alignment: the Model Behind Refine
*This work has been done while at* [*Conjecture*](https://www.conjecture.dev/)
Tl;dr: We need far more conceptual AI alignment research approaches than we have now if we want to increase our chances to solve the alignment problem. However, the conceptu... |
b3ee86f8-2157-4f23-a5d4-8474d1d54345 | trentmkelly/LessWrong-43k | LessWrong | Beware unfinished bridges
> This guy don't wanna battle, he's shook
> 'Cause ain't no such things as halfway crooks
>
> - 8 Mile
There is a commonly cited typology of cyclists where cyclists are divided into four groups:
1. Strong & Fearless (will ride in car lanes)
2. Enthused & Confident (will ride in unprotect... |
a475d33b-7bca-4f65-90db-dfb66de0b536 | trentmkelly/LessWrong-43k | LessWrong | Things I Learned Making The SB-1047 Documentary
Last week I published a 30-minute documentary about SB-1047 I had been working on since September.
Here's what I learned.
Story of the bill
1. The bill had to go through many committees before passing the California Senate and Assembly, eventually reaching the Govern... |
e41b8285-8033-45e8-8b96-a4628540ba85 | trentmkelly/LessWrong-43k | LessWrong | Moral strategies at different capability levels
Let’s consider three ways you can be altruistic towards another agent:
* You care about their welfare: some metric of how good their life is (as defined by you). I’ll call this care-morality - it endorses things like promoting their happiness, reducing their suffering,... |
cf893a47-7c0e-4b7b-bc54-9849a404b912 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Forecasting progress in language models
*Note:* this post was [cross-posted to Metaculus](https://www.metaculus.com/notebooks/8329/human-level-language-models/) over a week ago as part of their new [*Metaculus journal*](https://www.metaculus.com/project/journal/).
Here, I describe a way of measuring the performance ... |
a3650fc8-d694-46e7-a295-8be8f503968b | trentmkelly/LessWrong-43k | LessWrong | Intelligence explosion in plain, vanilla, mixed berry, and coffee flavors
John Scalzi took a throwaway line in a short rant he wrote about Atlas Shrugged as the basis for a short story about yogurt ruling the world. Enjoy! |
933a00a7-97c2-4165-a6ed-59afdf1f874e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Markus Anderljung and Ben Garfinkel: Fireside chat on AI governance
*Markus Anderljung and Ben Garfinkel discuss their paths into the field of AI governance, and how the field has developed in recent years. They explore the question “How sure are we about this AI stuff?”, and finish with an update on the latest resear... |
a6f62d76-78aa-425d-be24-de1c7fccdb3a | trentmkelly/LessWrong-43k | LessWrong | Momentum, Reflectiveness, Peace | Otium
None |
faf732d1-5e2c-4f4b-8aea-eb925d11951e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | A Simple Model of AGI Deployment Risk
What risks should we be willing to take when deploying superintelligent systems? Such systems could prove quite dangerous, and so we may wish to delay their use in order to improve their safety. But superintelligent systems can also be used to defend ourselves against other existe... |
bd35327f-8cc0-44ee-b95a-54966cdc0821 | trentmkelly/LessWrong-43k | LessWrong | Sam Altman's ouster at OpenAI was precipitated by letter to board about AI breakthrough - Reuters
|
42f3c4cf-34c7-4519-8f49-c631cedb492f | trentmkelly/LessWrong-43k | LessWrong | "Learning to Summarize with Human Feedback" - OpenAI
https://openai.com/blog/learning-to-summarize-with-human-feedback/
Eliezer's Summary/Thoughts: https://mobile.twitter.com/ESYudkowsky/status/1301954347933208578 |
4bf6476e-efe4-4d47-92c9-e5b72df91e82 | trentmkelly/LessWrong-43k | LessWrong | Things I Learned From Working With A Marketing Advisor
Epistemic Status: Opinions stated without justification
I’ve been getting a bunch of advice and help at LRI from a marketing/strategy expert, and it’s been an education. She’s been great to work with — she kicks my ass, in a good way. Basically, she takes my wri... |
335f65ba-46f2-451d-92ef-149f73129883 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Dunbar's Function
Today's post, Dunbar's Function was originally published on 31 December 2008. A summary (taken from the LW wiki):
> Robin Dunbar's original calculation showed that the maximum human group size was around 150. But a typical size for a hunter-gatherer band would be 30-50, cohesive onlin... |
c7db8c09-586c-4891-a5dd-e9645aad83d0 | trentmkelly/LessWrong-43k | LessWrong | Category Qualifications (w/ exercises)
This is the second post in the Arguing Well sequence. This post is influenced by A Human's Guide to Words and Against Lie Inflation.
Intro
In the last post, we discussed a common problem in arguments that Prove Too Much. In this post, we’ll generalize that problem to help deter... |
10610871-81fc-4f38-86c0-2e0950150a92 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Should we push for requiring AI training data to be licensed?
We frequently speak about AI capability gain being bad because it shortens the timeframe for AI safety research. In that logic, taking steps to decrease AI capability would be worthwhile.
At the moment the large language models are trained with a lot of da... |
0dac102f-92c2-4067-ab5d-886ee385568d | trentmkelly/LessWrong-43k | LessWrong | What Rate of Return Should You Expect?
If you invest money, perhaps because you want to give money in the future, you want a good estimate of the real rate of return so you can compare it to just giving now. Small differences in this estimate have very large effects over decades, and just today I've seen people giving... |
b99f04b3-b68f-4f41-b8b9-795bd24f0ade | trentmkelly/LessWrong-43k | LessWrong | A discussion of heroic responsibility
[Originally posted to my personal blog, reposted here with edits.]
Introduction
> You could call it heroic responsibility, maybe,” Harry Potter said. “Not like the usual sort. It means that whatever happens, no matter what, it’s always your fault. Even if you tell Professor McGo... |
80a98a31-a84b-410c-a8e9-faa80b62ae6c | StampyAI/alignment-research-dataset/arxiv | Arxiv | Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
1 Introduction
---------------
Deep learning has dramatically advanced the state of the art in vision, speech,
and many other areas. Stochastic gradient descent (SGD) has proved to be an effective way of training
deep netwo... |
e8de9e7d-84f0-465a-940f-1c265155d8b7 | trentmkelly/LessWrong-43k | LessWrong | The Summoned Heroine's Prediction Markets Keep Providing Financial Services To The Demon King!
The Summoned Heroine and the Demon King
The Summoned Heroine announced to the Circular Citadel that she would be creating a prediction market, to help them anticipate and counter the Demon King’s next moves. There was cheeri... |
acde6c9a-455f-457b-b951-4c214c298575 | trentmkelly/LessWrong-43k | LessWrong | Minor interpretability exploration #2: Extending superposition to different activation functions
Epistemic status: small exploration without previous predictions, results low-stakes and likely correct.
Introduction
As a personal exercise for building research taste and experience in the domain of AI safety and speci... |
d41a7d21-e60e-424d-b8a8-3d9f528baa13 | trentmkelly/LessWrong-43k | LessWrong | How to (un)become a crank
Ahhh, a human interest post. Well, sort of. At least it has something besides math-talk.
In the extreme programming community they have a saying, "three strikes and you refactor". The rationalist counterpart would be this: once you've noticed the same trap twice, you'd be stupid to fall prey... |
234d9089-b1e3-498a-a1f5-b473fbd4f739 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AGI safety from first principles: Control
It’s important to note that my previous arguments by themselves do not imply that AGIs will end up in control of the world instead of us. As an analogy, scientific knowledge allows us to be much more capable than stone-age humans. Yet if dropped back in that time with just our... |
b22fc904-e6f2-41f9-889c-f158b56a2c63 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Inner alignment requires making assumptions about human values
Many approaches to AI alignment require making assumptions about what humans want. On a first pass, it might appear that inner alignment is a sub-component of AI alignment that doesn't require making these assumptions. This is because if we define the prob... |
a6302529-0874-415f-bdbf-f12f9ef2b3b7 | trentmkelly/LessWrong-43k | LessWrong | Are conferences an inefficient/terrible discussion forum (in addition to academic papers)?
I'm asking this as a follow-up to http://lesswrong.com/lw/d5y/why_academic_papers_are_a_terrible_discussion/, which was written a few years ago, and which I find very interesting.
Many of the arguments advanced in http://lesswr... |
32f611e8-55d4-4dab-aafe-b817583ad0f0 | trentmkelly/LessWrong-43k | LessWrong | What are your thoughts on rational wiki
I have seen a a post on LW that suggest RationalWiki is a poor source of information. Their post on Pick-Up Artists also avoids some key questions and focuses on the controversial points, making the article misleading. That is just two observations. What do you think about the s... |
86528a2f-b1e8-4cf3-83f1-41792bd1d39f | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "AlphaGo Zero uses 4 TPUs, is built entirely out of neural nets with no handcrafted features, doesn't pretrain against expert games or anything else human, reaches a superhuman level after 3 days of self-play, and is the strongest version of AlphaGo yet.The architecture has been simplified. Previous AlphaGo had a polic... |
df58f8db-67d8-4fb3-a219-aff842f64e85 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Emotional attachment to AIs opens doors to problems
We are going to see more and more emotional bonds between humans and AIs
------------------------------------------------------------------------
Recently I red an interesting post ["How it feels to have your mind hacked by an AI"](https://www.lesswrong.com/posts/9k... |
06e217dc-a34c-486e-a07a-2a15d2991240 | trentmkelly/LessWrong-43k | LessWrong | Pantry Staples for DIY
It's often helpful to be able to build or fix things without needing to go to the store: perhaps it's urgent, perhaps you don't have a car, perhaps there's a pandemic on, perhaps you have an idea that you want to try out before your interest fades. In cooking, the foods that you keep on hand are... |
1cc458a9-224b-42aa-aceb-8bd6fb26dc81 | trentmkelly/LessWrong-43k | LessWrong | Space
Space begets space
----------------------------------------
Part III
Like an anaemic finding iron, a week into this process I had a profound sense that I'd gained something in which I'd been chronically deficient. I started calling whatever it was that I'd gained, "Space".
Space is a condition of abundant un... |
e8287d62-afa8-4a0a-afbe-58a079fa1e07 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Failures of Contingent Thinking
1 Motivation
-------------
When facing complicated and uncertain environments, economic agents often fail to act optimally, choosing actions which do not maximize their expected payoff. An intuitive explanation, and one often appealed to in the experimental and behavioral literature,... |
8056eac9-0c36-4284-98fa-2fc0d70e4944 | trentmkelly/LessWrong-43k | LessWrong | Plan to Be Lucky
Like other rationalists, I’ve written about the dangers inherent in our instincts for tribalism. Unlike other rationalists, I think that instead of being suppressed tribalism should be indulged in a productive manner. That’s why I take a couple of hours each week, alone or with friends, to scream at a... |
12b731cb-23cf-472f-b745-e57b252b22fd | StampyAI/alignment-research-dataset/arxiv | Arxiv | Understanding Hindsight Goal Relabeling from a Divergence Minimization Perspective
1 Introduction
---------------
Goal reaching is an essential aspect of intelligence in sequential decision making. Unlike the conventional formulation of reinforcement learning (RL), which aims to encode all desired behaviors into a ... |
29df8ea0-a1aa-42fc-8ba3-3f8227221af3 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post125
Subhash and Josh are co-first authors on this work done in Neel Nanda’s MATS stream. We recently released a new paper investigating sparse probing that follows up on a post we put up a few months ago. Our goal with the paper was to provide a single rigorous data point when evaluating the utility of SAEs. T... |
bc66ed82-16d2-4003-8de0-860cc0347abd | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Two More Decision Theory Problems for Humans
(This post has been sitting in my drafts folder for 6 years. Not sure why I didn't make it public, but here it is now after some editing.)
There are two problems closely related to the [Ontological Crisis in Humans](/posts/KLaJjNdENsHhKhG5m/ontological-crisis-in-humans). ... |
0cd53fe0-97b4-4ff0-873c-1ffca907ab3c | trentmkelly/LessWrong-43k | LessWrong | PICT: A Zero-Shot Prompt Template to Automate Evaluation
Produced as part of the SERI ML Alignment Theory Scholars Program - Winter 2022 Cohort.
I’d like to thank Rebecca Gorman and Stuart Armstrong for their mentorship and advice on the post.
This post describes a zero-shot template prompt that works surprisingly w... |
c52ff8ac-0e36-4c2a-b704-cd8558c9aabf | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Ingredients of Timeless Decision Theory
**Followup to**: [Newcomb's Problem and Regret of Rationality](/lw/nc/newcombs_problem_and_regret_of_rationality/), [Towards a New Decision Theory](/lw/15m/towards_a_new_decision_theory/)
Wei Dai [asked](/lw/15m/towards_a_new_decision_theory/11ji):
>
> "Why didn't you men... |
306504e3-9b48-4e28-8f44-d5061287f158 | trentmkelly/LessWrong-43k | LessWrong | How urgent is it to intuitively understand Bayesianism?
The current state of my understanding (briefly):
1. I very much understand reductionism and the distinction between the map and the territory. And I very much understand that probability is in the mind.
2. From what I understand, prior probability is just the ... |
194d5ff6-b219-46e6-9839-48225085ee96 | trentmkelly/LessWrong-43k | LessWrong | Weirdness at the wiki
Richard Kennaway has posted about an edit war on the wiki. Richard, thank you.
Unfortunately, I've only used the wiki a little, and don't have a feeling for why the edit history for an article is inaccessible. Is the wiki broken or has someone found a way to hack it? Let it be known that hacking... |
b5eddc01-cc96-4c41-a686-ca5f18b12eb5 | trentmkelly/LessWrong-43k | LessWrong | Observer-Based Values
In my current novel, Melt-Out, an AGI arise without participation by humans, but is nevertheless an aligned AI. How does this happen? My character, Aylina Tayu, is a skeptic and pessimist whose major contribution to AI is her protoc of an Ideal Bayesian Adapter, a synonym (or a placeholder) for a... |
171787e0-5e6b-4a5b-9d09-7292c95dea89 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Atlanta - Practical Rationality Meetup Session
Discussion article for the meetup : Atlanta - Practical Rationality Meetup Session
WHEN: 02 December 2012 05:00:00PM (-0500)
WHERE: Atlanta, ga
Here are the event details! Please let me know if you have a different preferred location or other suggestions and I... |
3b50a0ae-0a2b-4e42-9ce6-b4cd0906982a | trentmkelly/LessWrong-43k | LessWrong | The great quote of rationality a la Socrates (or Plato, or Aristotle)
Help a brother out?
There's a great quote by one of The Big 3 Greek Philosophers (EDIT: Reference to Cicero removed) which I can paraphrase by memory as:
"I consider it rather better for myself to be proven wrong than to prove someone else w... |
a4f422f0-3278-4740-8b42-2f359915f58e | trentmkelly/LessWrong-43k | LessWrong | Instrumental Rationality 3: Interlude I
[Instrumental Rationality Sequence 3/7].
[Here, we’ll cover two concepts: Acting into Uncertainty and Fading Novelty. They’re both sort of about two (generalizable, I hope) mental feelings (i.e. internal experiences) that can occur when you start trying to act on any of the in... |
8c261e6b-4629-4860-8d93-092be23f7698 | trentmkelly/LessWrong-43k | LessWrong | What's the difference between Wisdom and Rationality?
Are Wisdom and Rationality the same? If not, how are they different? Is one a subset of the other? Or does each have parts that aren't contained in the other?
Rationality is about cognitive algorithms, that either improve your map-territory correspondence or help ... |
79d17ba9-b311-4b9f-a799-a41b0d4e4ff1 | trentmkelly/LessWrong-43k | LessWrong | Reframing a Crush: Distilling the "like" out of "like like"
When I was a kid, I was moved by the idea that love was a powerful force. In the media, people did ridiculous things in the name of love; worse still when love was unrequited. I was terrified of feeling completely and helplessly vulnerable for someone who cou... |
53ffdd2f-007d-42f4-a3cb-d7c86832e5b3 | trentmkelly/LessWrong-43k | LessWrong | AI play for the next 3 years: Lemonade Insurance
After Tesla's stock surged tenfold around 2020, a friend asked me what I knew that others didn't. It’s a bit complicated to explain, but a similar opportunity might arise with Lemonade Insurance in the next 3-4 years. The current situation is a combination of a very pro... |
0221332e-885f-4521-b313-65f2aec2f31f | trentmkelly/LessWrong-43k | LessWrong | Worrying about the Vase: Whitelisting
> Suppose a designer wants an RL agent to achieve some goal, like moving a box from one side of a room to the other. Sometimes the most effective way to achieve the goal involves doing something unrelated and destructive to the rest of the environment, like knocking over a vase of... |
1ab85b2c-e44b-4155-9310-9b5c2c8c3390 | trentmkelly/LessWrong-43k | LessWrong | HOWTO: Screw Up The LessWrong Survey and Bring Great Shame To Your Family
Let's talk about the LessWrong Survey.
First and foremost, if you took the survey and hit 'submit', your information was saved and you don't have to take it again.
Your data is safe, nobody took it or anything it's not like that. If you took t... |
d8095969-5272-4ab4-8976-fd7f40b22f84 | trentmkelly/LessWrong-43k | LessWrong | Project Idea: Challenge Groups for Alignment Researchers
@Henrik Karlsson writes a lot about what makes people successful. One recurring theme I've found involves the people around them. Successful people almost always have a group of friends who they're constantly chatting with, bouncing ideas around with, learning f... |
fbf17330-0907-4437-80df-0ec16371934c | trentmkelly/LessWrong-43k | LessWrong | Making a Cheerful Bid
Asking someone for their cheerful price is asking them for the price at which they'd certainly do what you ask.
This opens the door to asking for whatever you want. Prostitution, murder, watching 'The Bachelor' on loop for all your waking hours for a year: what is your cheerful price? The answer... |
e88eaf7e-a74e-4b3a-a1f2-e352b16e1880 | trentmkelly/LessWrong-43k | LessWrong | Trying to hide bad signaling? To the Dark Side, lead you it will.
Edit: This is old material. It may be out of date.
Or is that just a point of view?
I'm going to assume familiary with the common use of the following two terms on this site:
* Dark Arts
* Signaling
Otherwise don't worry, I've hanged out here fo... |
40ea9795-591f-460d-adda-f219cfed4d33 | trentmkelly/LessWrong-43k | LessWrong | Book Review of 5 Applied Bayesian Statistics Books
There are 5 strong contender for best Bayesian Statistics book
TLDR
* Statistical Rethinking, henceforth SR[1]
* Up to speed fast, no integrals, very intuitive approach.
* Doing Bayesian Data Analysis, henceforth The Dog Book[2]
* This is the easiest book. I... |
33c7402a-dec7-4130-9581-963d79496c89 | trentmkelly/LessWrong-43k | LessWrong | [Humor] A Fearsome Rationality Technique
|
402b9c5d-dbaf-4a9d-a0f5-e5e4657a9223 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | An Agent is a Worldline in Tegmark V
If asked to define what an *agent* is, my usual answer -- one of them, anyway -- is “a [worldline](https://en.wikipedia.org/wiki/World_line) in Tegmark V”.
The Tegmark Level V Multiverse (the “V” here is a Roman numeral) is not defined by [Max Tegmark](https://arxiv.org/pdf/0905.... |
12075179-1ded-421d-b8cb-43ca0b509da3 | trentmkelly/LessWrong-43k | LessWrong | A Step-by-step Guide to Finding a (Good!) Therapist
The process of finding a good therapist is unlike the process of finding most professionals. If your dentist nags you to floss more every time you see them, it's annoying, but unlikely to really affect the quality of their work. In therapeutic work, however, most of ... |
4b8144b7-e680-4e99-a49f-1316146903f4 | trentmkelly/LessWrong-43k | LessWrong | Are we prepared for Solar Storms?
The COVID19- Pandemic has brought one awful fact to global attention:
The Powers that Be are woefully unprepared for once-in-century global risks.
Are we prepared for Solar Storms (solar flares, coronal mass ejection events, etc)?
A basic summary:
A more advanced explanation... |
515b236d-7251-458d-8a3e-b54bb4180970 | trentmkelly/LessWrong-43k | LessWrong | The Anti-Carter Basilisk
In 1983, astrophysicist Brandon Carter proposed the "Doomsday argument" that supposes we are "likely" living near the end of humankind just before the species dies out. He reasoned that since most species tend to die out suddenly after a period of exponential increase, therefore if humans will... |
2bdf8b4d-8273-44b1-92b1-6985a615e1b2 | trentmkelly/LessWrong-43k | LessWrong | Problems with Bayesianism: A Socratic Dialogue
Crossposted from my blog
In this fictional dialogue between a Bayesian (B) and a Non-Bayesian (N) I will propose solutions to some pre-existing problems with Bayesian epistemology, as well as introduce a new problem for which I offer a solution at the end. (Computer sc... |
e2525bdc-f795-4fe6-898c-ba55b0a3beb0 | trentmkelly/LessWrong-43k | LessWrong | Categorizing failures as “outer” or “inner” misalignment is often confused
Pop quiz: Are the following failures examples of outer or inner misalignment? Or is it ambiguous?
(I’ve spoilered the scenarios so you look at each one separately and answer before moving on to the next. Even if you don’t want to actually answ... |
5460ef5b-a4bd-40e9-92f8-693f5c8d448d | trentmkelly/LessWrong-43k | LessWrong | [LINK] 'Blue Brain' Project Accurately Predicts Connections Between Neurons
From http://www.sciencedaily.com/releases/2012/09/120917152043.htm
Could this be a tiny step towards an AGI?
> 'Blue Brain' Project Accurately Predicts Connections Between Neurons
> One of the greatest challenges in neuroscience is to ident... |
b756973b-e0b8-42e5-a5c2-20cef2e78bab | trentmkelly/LessWrong-43k | LessWrong | I am the Golden Gate Bridge
EASILY INTERPRETABLE SUMMARY OF NEW INTERPRETABILITY PAPER
Anthropic has identified (full paper here) how millions of concepts are represented inside Claude Sonnet, their current middleweight model. The features activate across modalities and languages as tokens approach the associated con... |
40c40c3b-85d6-4932-a30c-f00bb5378bf9 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Active Reinforcement Learning: Observing Rewards at a Cost
1 Introduction
---------------
Translating human objectives into rewards for a reinforcement learning (RL) agent is often challenging and time-consuming.
Tasks such as driving in a busy city or engaging in a conversation have complex reward structures which... |
39957539-3f63-4f49-b146-2ea6ccc8bc34 | trentmkelly/LessWrong-43k | LessWrong | Bounded Rationality: Two Cultures
Have not read this in any detail, but looked like interesting food for thought. Here's the first section of the paper.
> Introduction and Outline
> Bounded rationality does not speak with one voice. This is not only because bounded rationality is researched in various fields such as... |
a9fd8459-5a5b-476d-bf5b-bc56fcc66e0d | trentmkelly/LessWrong-43k | LessWrong | Effectiveness of Moderna booster shot and mix-and-match?
I read that the FDA has approved the Moderna booster shot as well as mix-and-match shots. I have two doses of Moderna and inclined to take a booster as Pfizer's booster showed 95.6% effectiveness in their Phase 3 trial.
Was wondering if any has substantial evi... |
c399e764-fe3c-44f7-90e3-33192a772eb1 | trentmkelly/LessWrong-43k | LessWrong | "incomparable" outcomes--multiple utility functions?
I know that this idea might sound a little weird at first, so just hear me out please?
A couple weeks ago I was pondering decision problems where a human decision maker has to choose between two acts that lead to two "incomparable" outcomes. I thought, if outcome A... |
7c8dad23-925c-497f-af6f-58ba5e286229 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Frontier Model Security
> As the capabilities of frontier artificial intelligence models continue to increase rapidly, ensuring the security of these systems has become a critical priority. In our previous posts, we’ve focused on Anthropic’s [approach to safety](https://www.anthropic.com/index/core-views-on-ai-safety)... |
17f59c2d-e6a7-440a-a8ee-14ebb390de02 | trentmkelly/LessWrong-43k | LessWrong | Conditional Prediction with Zero-Sum Training Solves Self-Fulfilling Prophecies
Thanks to Charlotte Siegmann, Caspar Oesterheld, Spencer Becker-Kahn, and Evan Hubinger for providing feedback on this post.
The issue of self-fulfilling prophecies, also known as performative prediction, arises when the act of making a p... |
c3f6606e-7192-4d21-9ece-b5880bbf9ad3 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Atlanta LessWrong: August Meetup
Discussion article for the meetup : Atlanta LessWrong: August Meetup
WHEN: 04 August 2013 06:00:00PM (-0400)
WHERE: 2388 Lawrenceville Hwy. Apt L. Decatur, GA 30033
For the first of our two planned August meetups, we will be continuing our work on group self-improvement an... |
bd6d8bb0-1eec-4dee-8c3a-ef4df1c37e2c | trentmkelly/LessWrong-43k | LessWrong | Fairness vs. Goodness
It seems that back when the Prisoner's Dilemma was still being worked out, Merrill Flood and Melvin Drescher tried a 100-fold iterative PD on two smart but unprepared subjects, Armen Alchian of UCLA and John D. Williams of RAND.
The kicker being that the payoff matrix was asymmetrical, with dual... |
a4c060d2-1f52-464b-b739-ba37dcb20e38 | StampyAI/alignment-research-dataset/arxiv | Arxiv | A Theory of Universal Learning
1 Introduction
---------------
In supervised machine learning, a learning algorithm
is presented with labeled examples of a concept,
and the objective is to output a classifier which
correctly classifies most future examples from the
same source. Supervised learning has been successfu... |
762b5a76-9e3c-4fd1-bdb7-f198e9154709 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Gradations of Agency
[Epistemic status: This post written hastily, for Blog Post Day.]
*I’ve de-prioritised* [*this sequence*](https://www.lesswrong.com/s/aek5ksSs2FHTeofsf) *due to other* [*recent*](https://www.lesswrong.com/posts/kpPnReyBC54KESiSn/optimality-is-the-tiger-and-agents-are-its-teeth)[*posts*](https://w... |
548440b0-387b-47b5-8605-0e618c1f3560 | trentmkelly/LessWrong-43k | LessWrong | Name of Problem?
If we expand out an arbitrary program, we get a (usually infinite) expression tree. For instance, we can expand
fact(n) := (n == 0) ? 1 : n*fact(n-1)
into
fact(n) = (n == 0) ? 1 : n*(((n-1) == 0) ? 1 : (n-1) * ( (((n-1)-1) == 0) ? ... ))))
Let's call two programs "expression-equivalent" if they ex... |
23588fac-e869-44d4-8b0c-b03acaa87e06 | trentmkelly/LessWrong-43k | LessWrong | AI safety without goal-directed behavior
When I first entered the field of AI safety, I thought of the problem as figuring out how to get the AI to have the “right” utility function. This led me to work on the problem of inferring values from demonstrators with unknown biases, despite the impossibility results in the ... |
81461f86-5b11-4b08-bbc7-1b92c6bb2ec8 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Final Words
Today's post, Final Words was originally published on 27 April 2009. A summary (taken from the LW wiki):
> The conclusion of the Beisutsukai series.
Discuss the post here (rather than in the comments to the original post).
This post is part of the Rerunning the Sequences series, where we... |
e6428036-3f34-413e-92ca-ae4798a73365 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Recall and Regurgitation in GPT2
*The first half of this post uses causal tracing to explore differences in how GPT2-XL handles completing cached phrases vs completing factual statements. The second half details my attempt to build intuitions about the high-level structure of GPT2-XL and is speculation heavy.*
*Some ... |
8f49faec-2739-4fdd-81a5-306adf5cffd9 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Optimal and Causal Counterfactual Worlds
.mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowrap; flo... |
d8fc8ec9-1d35-42a9-8479-5096611e8fa5 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Automatic Correction of Human Translations.
1 Introduction
---------------
| Error Type | Example Text |
| --- | --- |
| Monolingual: typos | s𝑠sitalic\_s: Do your feet roll inwards when running? |
| t𝑡titalic\_t: K ppen deine Füße beim Laufen nach innen? |
| t′superscript𝑡′t^{\prime}italic\_t start\_POSTSUPER... |
ee7ee536-4025-4129-b4b5-fedfaa042686 | trentmkelly/LessWrong-43k | LessWrong | Do bamboos set themselves on fire?
Cross-posted from Telescopic Turnip.
As we all know, the best place to have a kung-fu fight is a bamboo forest. There are just so many opportunities to grab pieces of bamboos and manufacture improvised weapons, use them to catapult yourself in the air and other basic techniques any ... |
8ff4996a-07ea-4b5a-8007-2e3832b5587e | trentmkelly/LessWrong-43k | LessWrong | In Defense of Lawyers Playing Their Part
This is a linkpost for In Defense of Lawyers Playing Their Part.
Michael Huemer writes about why he believes it’s wrong for lawyers to pursue unjust legal outcomes.
It's a good article, and one of the best defenses of this position I've seen. Still, I think this argument is m... |
9385272f-dd25-46d2-8986-f81defa2dc5f | trentmkelly/LessWrong-43k | LessWrong | Since figuring out human values is hard, what about, say, monkey values?
So, human values are fragile, vague and possibly not even a well defined concept, yet figuring it out seems essential for an aligned AI. It seems reasonable that, faced with a hard problem, one would start instead with a simpler one that has some... |
4b83529e-bd98-4c2a-969a-58e422f05629 | trentmkelly/LessWrong-43k | LessWrong | Meetup : LW Buffalo Meetup at Buffalo Labs
Discussion article for the meetup : LW Buffalo Meetup at Buffalo Labs
WHEN: 07 February 2013 07:00:00PM (-0500)
WHERE: 2727 Broadway Ave. Suite 210, Cheektowaga, NY
Hi everyone!
For those of you who couldn't make the last meetup because it was on a Thursday, vote in the d... |
47eb8fec-4424-4785-9df9-1edcb9b198f9 | trentmkelly/LessWrong-43k | LessWrong | Melbourne Meetup: Friday 1st July 7pm
When: Friday 1st July, 19:00 (recurring on the first Friday of each following month)
Where: TrikeApps office, lvl 2, 55 Walsh St, West Melbourne 3003 (http://trikeapps.com/contact)
Directions:
Enter the somewhat unfriendly building, climb the stairs to the top (2 floors), and tu... |
e2d122ae-1dca-416f-8f3e-82d951ad5351 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | A Concrete Proposal for Adversarial IDA
*Note: This post came out of a conversation with Geoffrey Irving and Buck Shlegeris.*
*Epistemic Status: I suspect Paul has already thought of most or all of the ideas presented here, though I nevertheless found the exercise of carefully specifying an IDA implementation helpful... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.