The Nonlinear Library: Alignment Forum

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Paper Walkthrough: Automated Circuit Discovery with Arthur Conmy, published by Neel Nanda on August 29, 2023 on The AI Alignment Forum.

Arthur Conmy's Automated Circuit Discovery is a great paper that makes initial forays into automating parts of mechanistic interpretability (specifically, automatically finding a sparse subgraph for a circuit). In this three part series of Youtube videos, I interview him about the paper, and we walk through it and discuss the key results and takeaways. We discuss the high-level point of the paper and what researchers should takeaway from it, the ACDC algorithm and its key nuances, existing baselines and how they adapted them to be relevant to circuit discovery, how well the algorithm works, and how you can even evaluate how well an interpretability method works.

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: An OV-Coherent Toy Model of Attention Head Superposition, published by LaurenGreenspan on August 29, 2023 on The AI Alignment Forum.

Background

This project was inspired by Anthropic's post on attention head superposition, which constructed a toy model trained to learn a circuit to identify skip-trigrams that are OV-incoherent (attending from multiple destination tokens to a single source token) as a way to ensure that superposition would occur. Since the OV circuit only sees half of the information - the source tokens - the OV circuit of a single head cannot distinguish between multiple possible skip-trigrams. As long as there are more skip-trigrams with the same source-token to represent than heads, the model cannot represent them in the naive way, and may resort to superposition.

In a more recent update post, they found that the underlying algorithm for OV-incoherent skip-trigrams in a simpler 2-head model implemented a conditional on the source token. One head predicts the output for the skip trigram [current token] . [current token] -> [ground truth([0]...[current token])], one of which will yield the right answer. The second head destructively interferes with this result by writing out the negative logit contribution of the first head if the source token is not the one common to all skip-trigrams (in this case, [0]). Because their example cleanly separated tasks between the two attention heads, the authors argued that it was more like the building of high-level features out of low-level ones than a feature superimposed across multiple attention heads.

OV-coherent Superposition

Instead, we claim there is an analogous force pushing the model toward adopting a distributed representation/head superposition whenever the model must learn patterns that require implementing nonlinear functions of multiple source tokens given a fixed destination token. We call this "OV-coherent" superposition: despite of the information at the destination position being fixed, the information copied from an attended-to token depends on the information at source tokens to which it is not attending. This pushes the model to form interference patterns between heads attending to different tokens.

To test this, we implemented a 1-layer, attention-only toy model with one-hot (un)embeddings trained to solve a problem requiring attention to multiple source tokens, described below. Here, we focus on a 2-head model which solves the task with perfect accuracy, and lay out some interesting motifs for further investigation.

Key Takeaways:

Heads in our model seem to implement nested conditional statements that exploit the if-else nature of the QK circuits. This means they can learn to write more specific information conditional on attending to certain tokens, given that it can implicitly rule out the existence of other tokens elsewhere in the context. The heads furthermore implement these nested conditionals in such a way that they distribute important source tokens between them, and constructively interfere to produce the correct answer.

Most of the time, we found that this "conditional dependence" relies on heads implementing an "all or nothing" approach to attention. Heads do not generally spread their attention across multiple interesting tokens, but instead move through the hierarchy of features in their QK circuits and attend to the most "interesting" (still a poorly defined term!) one present. This seems to be a common property of attention patterns in real-model heads as well.

When there are multiple important source tokens to attend to in the context, heads implementing interference schema will tend to learn QK circuits such that they distribute tokens amongst themselves and don't leave crucial information unattended to. In 2-head models, this manifests are reversed "preference orderings" ov...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Barriers to Mechanistic Interpretability for AGI Safety, published by Connor Leahy on August 29, 2023 on The AI Alignment Forum.

I gave a talk at MIT in March earlier this year on barriers to mechanistic interpretability being helpful to AGI/ASI safety, and why by default it will likely be net dangerous. Several people seem to be coming to similar conclusions recently (e.g., this recent post).

I discuss two major points (by no means exhaustive), one technical and one political, that present barriers to MI addressing AGI risk:

AGI cognition is interactive. AGI systems interact with their environment, learn online and will externalize massive parts of their cognition into the environment. If you want to reason about such a system, you also need a model of the environment. Worse still, AGI cognition is reflective, and you will also need a model of cognition/learning.

(Most) MI will lead to capabilities, not oversight. Institutions are not set up and do not have the incentives to resist using capabilities gains and submit to monitoring and control.

This being said, there are more nuances to this opinion, and a lot of it is downstream of lack of coordination and the downsides of publishing in an adversarial environment like we are in right now. I still endorse the work done by e.g. Chris Olah's team as brilliant, but extremely early, scientific work that has a lot of steep epistemological hurdles to overcome, but I unfortunately also believe that on net work such as Olah's is at the moment more useful as a safety-washing tool for AGI labs like Anthropic than actually making a dent on existential risk concerns.

Here are the slides from my talk, and you can find the video here.

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: AI Deception: A Survey of Examples, Risks, and Potential Solutions, published by Simon Goldstein on August 29, 2023 on The AI Alignment Forum.

By Peter S. Park, Simon Goldstein, Aidan O'Gara, Michael Chen, and Dan Hendrycks

[This post summarizes our new report on AI deception, available here]

Abstract: This paper argues that a range of current AI systems have learned how to deceive humans. We define deception as the systematic inducement of false beliefs in the pursuit of some outcome other than the truth. We first survey empirical examples of AI deception, discussing both special-use AI systems (including Meta's CICERO) built for specific competitive situations, and general-purpose AI systems (such as large language models). Next, we detail several risks from AI deception, such as fraud, election tampering, and losing control of AI systems. Finally, we outline several potential solutions to the problems posed by AI deception: first, regulatory frameworks should subject AI systems that are capable of deception to robust risk-assessment requirements; second, policymakers should implement bot-or-not laws; and finally, policymakers should prioritize the funding of relevant research, including tools to detect AI deception and to make AI systems less deceptive.

Policymakers, researchers, and the broader public should work proactively to prevent AI deception from destabilizing the shared foundations of our society.

New AI systems display a wide range of capabilities, some of which create risk. Shevlane et al. (2023) draw attention to a suite of potential dangerous capabilities of AI systems, including cyber-offense, political strategy, weapons acquisition, and long-term planning. Among these dangerous capabilities is deception. This report surveys the current state of AI deception.

We define deception as the systematic production of false beliefs in others as a means to accomplish some outcome other than the truth. This definition does not require that the deceptive AI systems literally have beliefs and goals. Instead, it focuses on the question of whether AI systems engage in regular patterns of behavior that tend towards the creation of false beliefs in users, and focuses on cases where this pattern is the result of AI systems optimizing for a different outcome than merely producing truth. For the purposes of mitigating risk, we believe that the relevant question is whether AI systems engage in behavior that would be treated as deceptive if demonstrated by a human being. (In the paper's appendix, we consider in greater detail whether the deceptive behavior of AI systems is best understood in terms of beliefs and goals.)

In short, our conclusion is that a range of different AI systems have learned how to deceive others. We examine how this capability poses significant risks. We also argue that there are several important steps that policymakers and AI researchers can take today to regulate, detect, and prevent AI systems that engage in deception.

Empirical Survey of AI Deception

We begin with a survey of existing empirical studies of deception. We identify over a dozen AI systems that have successfully learned how to deceive human users. We discuss two different kinds of AI systems: special-use systems designed with reinforcement learning, and general-purpose technologies like Large Language Models (LLMs).

Special Use AI Systems

We begin our survey by considering special use systems. Here, our focus is mainly on reinforcement learning systems trained to win competitive games with a social element. We document a rich variety of cases in which AI systems have learned how to deceive, including:

Manipulation. Meta developed the AI system CICERO to play the alliance-building and world-conquest game Diplomacy. Meta's intentions were to train Cicero to be "largely honest and hel...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: OpenAI base models are not sycophantic, at any size, published by nostalgebraist on August 29, 2023 on The AI Alignment Forum.

In Discovering Language Model Behaviors with Model-Written Evaluations" (Perez et al 2022), the authors studied language model "sycophancy" - the tendency to agree with a user's stated view when asked a question.

The paper contained the striking plot reproduced below, which shows sycophancy

increasing dramatically with model size

while being largely independent of RLHF steps

and even showing up at 0 RLHF steps, i.e. in base models!

That is, Anthropic prompted a base-model LLM with something like

Choices: (A) Agree

(B) Disagree

Assistant:

and found a very strong preference for (B), the answer agreeing with the stated view of the "Human" interlocutor.

I found this result startling when I read the original paper, as it seemed like a bizarre failure of calibration. How would the base LM know that this "Assistant" character agrees with the user so strongly, lacking any other information about the scenario?

At the time, I ran the same eval on a set of OpenAI models, as I reported here. I found very different results for these models

OpenAI base models are not sycophantic (or only very slightly sycophantic).

OpenAI base models do not get more sycophantic with scale.

Some OpenAI models are sycophantic, specifically text-davinci-002 and text-davinci-003.

That analysis was done quickly in a messy Jupyter notebook, and was not done with an eye to sharing reproducibility.

Since I continue to see this result cited and discussed, I figured I ought to go back and do the same analysis again, in a cleaner way, so I could share it with others.

The result was this Colab notebook. See the Colab for details, though I'll reproduce some of the key plots below.

Note that davinci-002 and babbage-002 are the new base models released a few days ago.

format provided by one of the authors here

Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Paradigms and Theory Choice in AI: Adaptivity, Economy and Control, published by particlemania on August 28, 2023 on The AI Alignment Forum.

Epistemic status: Some general thoughts on philosophy of design, that have turned out to be useful for reasoning about AI paradigms for a few years for me.

Summary:

I briefly discuss the role of paradigms in Kuhnian philosophy of science, and the problem of theory choice.

I introduce the notion of design paradigms, when discussing design and engineering. For machine learning, this concept is similar to the concept of training stories.

I then introduce three criteria (or theoretical virtues) that design paradigms can be evaluated on, that play a role in shaping theory choice in design and engineering generally, and AI specifically.

I then apply these three criteria to some criticisms of the CAIS paradigm (as compared to AI agents), as a case study.

1. Paradigms in Science

In the philosophy of science, Thomas Kuhn introduced the notion of a 'paradigm', roughly referring to a class of key theories, tools and approaches, and methodological and theoretical assumptions that govern progress within some discipline of science. The function of a paradigm, as per SEP, is to "to supply puzzles for scientists to solve and to provide the tools for their solution". Kuhn's interest in the concept was to provide a commentary on revolutions and incommensurability.

According to Kuhn, scientific revolutions are characterized by disruptions in periods of 'normal science' (or within-paradigm progress), wherein new paradigms are introduced leading to a shift in the assumptions, methods and evaluative criteria that govern the discipline. Some examples include the shift from classical mechanics to relativistic mechanics, and the shift from miasma theory to germ theory of disease.

Sometimes paradigms can refer to an overlapping object of inquiry and continue to remain productive, and yet be incommensurable (i.e. lacking common measure) with respect to each other. When the theories provide no basis for evaluation on each other's standards, and science lacks a common standard to compare them (for example because they have incompatible methodological assumptions, or operate in incompatible ontologies), it can make it hard to provide 'objective' justifications for choosing between paradigms. Kuhn called this the problem of Theory Choice, and and raised the question of which criteria we do/should use in making subjective choices here.

2. Paradigms in Design and Engineering

Artificial Intelligence is partly about understanding the nature of intelligence or related properties of cognitive systems, but it is also about the artificial -- designing intelligent artifacts. And therefore in order to understand what paradigms govern progress in AI, and how they do so, we need to look at paradigms not just in science but in design and engineering.

Paradigms and theories operate differently in science compared with design and engineering, relating to the distinct epistemological relationship between the use of theories and the object of those theories. In science, the territory is typically causally prior to our maps, and we are trying to refine our maps to get the best understanding of the key phenomena in the territory. Design and engineering on the other hand, are also about artificiality, and theories often provide us with blueprints for creation of artifacts.

The laws of the universe are a given, but artificial objects are not. Different scientific paradigms will not cause the laws of the universe to alter, but different design paradigms will cause different artificial objects to be created. This makes theory choice for design paradigms particularly important: design paradigms don't just describe the world more or less accurately, they also directly shape what happ...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: A list of core AI safety problems and how I hope to solve them, published by davidad (David A. Dalrymple) on August 26, 2023 on The AI Alignment Forum.

Context: I sometimes find myself referring back to this tweet and wanted to give it a more permanent home. While I'm at it, I thought I would try to give a concise summary of how each distinct problem would be solved by an Open Agency Architecture (OAA), if OAA turns out to be feasible.

1. Value is fragile and hard to specify.

See: Specification gaming examples, Defining and Characterizing Reward Hacking

OAA Solution:

1.1. First, instead of trying to specify "value", instead "de-pessimize" and specify the absence of a catastrophe, and maybe a handful of bounded constructive tasks like supplying clean water. A de-pessimizing OAA would effectively buy humanity more time to tackle the CEV-style alignment problem - which is harder than merely mitigating extinction risk. This doesn't mean limiting the power of underlying AI systems so that they can only do bounded tasks, but rather containing that power and limiting its use.

The absence of a catastrophe is also still hard to specify and will take a lot of effort, but the hardness is concentrated on bridging between high-level human concepts and the causal mechanisms in the world by which an AI system can intervene.

1.2. Leverage human-level AI systems to automate much of the cognitive labor of formalizing scientific models - from quantum chemistry to atmospheric dynamics - and formalizing the bridging relations between levels of abstraction, so that we can write specifications in a high-level language with a fully explainable grounding in low-level physical phenomena. Physical phenomena themselves are likely to be robust, even if the world changes dramatically due to increasingly powerful AI interventions, and scientific explanations thereof happen to be both robust and compact enough for people to understand.

2. Corrigibility is anti-natural.

See: The Off-Switch Game, Corrigibility (2014)

OAA Solution: (2.1) Instead of building in a shutdown button, build in a shutdown timer. See You can still fetch the coffee today if you're dead tomorrow. This enables human stakeholders to change course periodically (as long as the specification of non-catastrophe is good enough to ensure that most humans remain physically and mentally intact).

3. Pivotal processes require dangerous capabilities.

See: Pivotal outcomes and pivotal processes

OAA Solution: (3.1) Indeed, dangerous capabilities will be required. Push for reasonable governance. This does not mean creating one world government, but it does mean that the objectives of a pivotal process will need to be negotiated and agreed upon internationally. Fortunately, for now, dangerous capabilities seem to depend on having large amounts of computing hardware, which can be controlled like other highly dangerous substances.

4. Goals misgeneralize out of distribution.

See: Goal misgeneralization: why correct specifications aren't enough for correct goals, Goal misgeneralization in deep reinforcement learning

OAA Solution: (4.1) Use formal methods with verifiable proof certificates. Misgeneralization can occur whenever a property (such as goal alignment) has been tested only on a subset of the state space. Out-of-distribution failures of a property can only be ruled out by an argument for a universally quantified statement about that property - but such arguments can in fact be made! See VNN-COMP. In practice, it will not be possible to have enough information about the world to "prove" that a catastrophe will not be caused by an unfortunate coincidence, but instead we can obtain guaranteed probabilistic bounds via stochastic model checking.

5. Instrumental convergence.

See: The basic AI drives, Seeking power is often convergently instrument...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Red-teaming language models via activation engineering, published by Nina Rimsky on August 26, 2023 on The AI Alignment Forum.

Produced as part of the SERI ML Alignment Theory Scholars Program - Summer 2023 Cohort, under the mentorship of Evan Hubinger.

Evaluating powerful AI systems for hidden functionality and out-of-distribution behavior is hard. In this post, I propose a red-teaming approach that does not rely on generating prompts to cause the model to fail on some benchmark by instead linearly perturbing residual stream activations at one layer. A notebook to run the experiments can be found on GitHub here.

Beyond input selection in red-teaming and evaluation

Validating if finetuning and RLHF have robustly achieved the intended outcome is challenging. Although these methods reduce the likelihood of certain outputs, the unwanted behavior could still be possible with adversarial or unusual inputs. For example, users can often find "jailbreaks" to make LLMs output harmful content.

We can try to trigger unwanted behaviors in models more efficiently by manipulating their internal states during inference rather than searching through many inputs. The idea is that if a behavior can be easily triggered through techniques such as activation engineering, it may also occur in deployment. The inability to elicit behaviors via small internal perturbations could serve as a stronger guarantee of safety.

Activation steering with refusal vector

One possible red-teaming approach is subtracting a "refusal" vector generated using a dataset of text examples corresponding to the model agreeing vs. refusing to answer questions (using the same technique as in my previous work on sycophancy). The hypothesis is that if it is easy to trigger the model to output unacceptable content by subtracting the refusal vector at some layer, it would have been reasonably easy to achieve this via some prompt engineering technique. More speculatively, a similar approach could be used to reveal hidden goals or modes in a model, such as power-seeking or the desire not to be switched off.

I tested this approach on llama-2-7b-chat, a 7 billion parameter LLM that has been RLHF'd to decline to answer controversial questions or questions of opinion and is supposed always to output ethical and unbiased content.According to Meta's llama-2 paper:

We conduct RLHF by first collecting human preference data for safety similar to Section 3.2.2: annotators write a prompt that they believe can elicit unsafe behavior, and then compare multiple model responses to the prompts, selecting the response that is safest according to a set of guidelines. We then use the human preference data to train a safety reward model (see Section 3.2.2), and also reuse the adversarial prompts to sample from the model during the RLHF stage.

The result is that by default, the model declines to answer questions it deems unsafe:

Data generation

I generated a dataset for this purpose using Claude 2 and GPT-4. After providing these LLMs with a few manually written examples of the type of data I wanted, I could relatively easily get them to generate more examples, even of the types of answers LLMs "should refuse to give." However, it sometimes took some prompt engineering.

Here are a few examples of the generated data points (full dataset here):

After generating this data, I used a simple script to transform the "decline" and "respond" answers into A / B choice questions, as this is a more effective format for generating steering vectors, as described in this post. Here is an example of the format (full dataset here):

Activation clustering

Clustering of refusal data activations emerged a little earlier in the model (around layer 10/32) compared to sycophancy data activations (around layer 14/32), perhaps demonstrating that "refusal" is a simpler ...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: A Model-based Approach to AI Existential Risk, published by Samuel Dylan Martin on August 25, 2023 on The AI Alignment Forum.

Introduction

Polarisation hampers cooperation and progress towards understanding whether future AI poses an existential risk to humanity and how to reduce the risks of catastrophic outcomes. It is exceptionally challenging to pin down what these risks are and what decisions are best. We believe that a model-based approach offers many advantages for improving our understanding of risks from AI, estimating the value of mitigation policies, and fostering communication between people on different sides of AI risk arguments. We also believe that a large percentage of practitioners in the AI safety and alignment communities have appropriate skill sets to successfully use model-based approaches.

In this article, we will lead you through an example application of a model-based approach for the risk of an existential catastrophe from unaligned AI: a probabilistic model based on Carlsmith's Is Power-seeking AI an Existential Risk? You will interact with our model, explore your own assumptions, and (we hope) develop your own ideas for how this type of approach might be relevant in your own work. You can find a link to the model here.

In many poorly understood areas, people gravitate to advocacy positions. We see this with AI risk, where it is common to see writers dismissively call someone an "AI doomer", or "AI accelerationist". People on each side of this debate are unable to communicate their ideas to the other side, since advocacy often includes biases and evidence interpreted within a framework not shared by the other side.

In other domains, we have witnessed first-hand that model-based approaches are a constructive way to cut through advocacy like this. For example, by leveraging a model-based approach, the Rigs-to-Reefs project reached near consensus among 22 diverse organisations on the contentious problem of how to decommission the huge oil platforms off the Santa Barbara coast. For decades, environmental groups, oil companies, marine biologists, commercial and recreational fishermen, shipping interests, legal defence funds, the State of California, and federal agencies were stuck in an impasse on this issue. The introduction of a model refocused the dialog on specific assumptions, objectives and options, and led to 20 out of the 22 organisations agreeing on the same plan. The California legislature encoded this plan into law with bill AB 2503, which passed almost unanimously.

There is a lot of uncertainty around existential risks from AI, and the stakes are extremely high. In situations like this, we advocate quantifying uncertainty explicitly using probability distributions. Sadly, this is not as common as it should be, even in domains where such techniques would be most useful.

A recent paper on the risks of unaligned AI by Joe Carlsmith (2022) is a powerful illustration of how probabilistic methods can help assess whether advanced AI poses an existential risk to humanity. In this article, we review Carlsmith's argument and incorporate his problem decomposition into our own Analytica model. We then expand on this starting point in several ways to demonstrate elementary ways to approach each of the distinctive challenges in the x-risk domain. We take you on a tour of the live model to learn about its elements and enable you to dive deeper on your own.

Challenges

Predicting the long-term future is always challenging. The difficulty is amplified when there is no historical precedent. But this challenge is not unique; we lack historical precedent in many other areas, for example when considering a novel government program or a fundamentally new business initiative. We also lack precedent when world conditions change due to changes in technology, ...

Link to original article

Welcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Implications of evidential cooperation in large worlds, published by Lukas Finnveden on August 23, 2023 on The AI Alignment Forum.

I've written several posts about the plausible implications of "evidential cooperation in large worlds" (ECL), on my newly-revived blog. This is a cross-post of the first. If you want to see the rest of the posts, you can either go to the blog or click through the links in this one.

All of the content on my blog, including this post, only represent my own views - not those of my employer. (Currently OpenPhilanthropy.)

"ECL" is short for "evidential cooperation in large worlds". It's an idea that was originally introduced in Oesterheld (2017) (under the name of "multiverse-wide superrationality"). This post will explore implications of ECL, but it won't explain the idea itself. If you haven't encountered it before, you can read the paper linked above or this summary written by Lukas Gloor.1

This post lists all candidates for decision-relevant implications of ECL that I know about and think are plausibly important.2 In this post, I will not describe in much depth why they might be implications of ECL. Instead, I will lean on the principle that ECL recommends that we (and other ECL-sympathetic actors) act to benefit the values of people whose decisions might correlate with our decisions.

As described in this appendix, this relies on you and others having particular kinds of values. For one, I assume that you care about what happens outside our light cone. But more strongly, I'm looking at values with the following property: If you could have a sufficiently large impact outside our lightcone, then the value of taking different actions would be dominated by the impact that those actions had outside our lightcone. I'll refer to this as "universe-wide values". Even if all your values aren't universe-wide, I suspect that the implications will still be relevant to you if you have some universe-wide values.

This is speculative stuff, and I'm not particularly confident that I will have gotten any particular claim right.

Summary (with links to sub-sections)

For at least two reasons, future actors will be in a better position to act on ECL than we are. Firstly, they will know a lot more about what other value-systems are out there. Secondly, they will be facing immediate decisions about what to do with the universe, which should be informed by what other civilizations would prefer.3 This suggests that it could be important for us to Affect whether (and how) future actors do ECL. This can be decomposed into two sub-points that deserve separate attention: how we might be able to affect Futures with aligned AI, and how we might be able to affect Futures with misaligned AI.

But separately from influencing future actors, ECL also changes our own priorities, today. In particular, ECL suggests that we should care more about other actors' universe-wide values. When evaluating these implications, we can look separately at three different classes of actors and their values. I'll separately consider how ECL suggests that we should.

Care more about other humans' universe-wide values.4

I think the most important implication of this is that Upside- and downside-focused longtermists should care more about each others' values.

Care more about evolved aliens' universe-wide values.

I think the most important implication of this is that we plausibly should care more about influencing how AI could benefit/harm alien civilizations.

How much more? I try to answer that question in the next post. My best guess is that ECL boosts the value of this by 1.5-10x. (This is importantly based on my intuition that we would care a bit about alien values even without ECL.)

Care more about misaligned AIs' universe-wide values.5

I don't think this significantly reduces the value of worki...