This is a special post for quick takes by RobertHarling. Only they can create top-level comments. Comments here also appear on the Quick Takes page and All Posts page.
Comment Permalink
EAG Bay Area Application Deadline extended to Feb 9th – apply now!
We've decided to postpone the application deadline by one week from the old deadline of Feb 2nd. We are receiving more applications than in the past two years, and we have a goal of increasing attendance at EAGs which we think this will help. If you've already applied, tell your friends! If you haven't — apply now! Don't leave it till the deadline!
You can find more information on our website.
I think there's a nice hidden theme in the EAG Bay Area content, which is about how EA is still important in the age of AI (disclaimer: I lead the EAG team, so I'm biased). It's not just a technical AI safety conference, but it's also not ignoring the importance of AI. Instead, it's showing how the EA framework can help prioritise AI issues, and bring attention to neglected topics.
For example, our sessions on digital minds with Jeff Sebo and the Rethink team, and our fireside chat with Forethought on post-AGI futures, demonstrate how there's important AI related work that EA is key in making happen, and that others will neglect. And I think sessions like the AI journalism lightning talks and the screening of the animated series 'Ada' also demonstrate how a wide variety of careers and skillsets are important in addressing risks from AI, and why it's valuable for EA to be a broad and diverse movement.
We of course still have some great technical content, such as Ryan Greenblatt discussing the Alignment Faking paper. (And actually perhaps my favourite sessions are the non-AI sessions... I'm really excited to hear more about GiveWell's re-evaluation of GiveDirectly!). But I think the content helps remind me and demonstrate to me why I think the EA community is so valuable, even in the age of AI, and why I think it's still worthwhile for me to work on EA community building!
Applications close this Sunday (Feb 9th) if you want to come join us in the Bay!
TL;DR: A 'risky' career “failing” to have an impact doesn’t mean your career has “failed” in the conventional sense, and probably isn’t as bad it intuitively feels.
- You can fail to have an impact with your career in many ways. One way to break it down might be:
- The problem you were trying to address turns out to not be that important
- Your method for addressing the problem turns out to not work
- You don’t succeed in executing your plan
- E.g. you could be aiming to have an impact by reducing the risk of future pandemics, and you do this by aiming to become a leading academic to bring lots of resources and attention to improving vaccine development pipelines. There are several ways you could end up not having much of an impact: pandemic risk could turn out to not be that high; advances in testing and PPE mean we can identify and contain pandemics very quickly, and vaccines aren’t as important; industry labs advance vaccine development very quickly and your lab doesn’t end up affecting things; you don’t succeed at becoming a leading academic, and become a mid-tier researcher instead.
- People often feel risk averse with their careers- we’re worried about taking “riskier” options that might not work out, even if they have higher expected impact. However there are some reasons to think most of the expect impact could come from the tail scenarios where you're really successful.
- I think we neglect is that there are different ways your career plan can not work out. In particular, many of the scenarios where you don’t succeed to have a large positive impact, you still succeed in the other values you have for your career- e.g. you’re still a conventionally successful researcher, you just didn’t happen to save the world.
- And even if your plan “fails” because you don’t reach the level in the field you were aiming for, you likely still end up in a good position e.g. not a senior academic, just a mid-tier academic or a researcher in industry, or not a senior civil servant but mid-tier civil servant. This isn’t true in every area- in some massively oversubscribed areas like professional sports failing can mean not having any job. Or when doing a start-up. But I’d guess this isn’t the majority of impactful careers that people consider.
- I also can imagine myself finding the situation of having tried and failed somewhat comforting in that I can think to myself “I did my bit, I tried, it didn’t work out, but it was a shot worth taking, and now I just have this normally good life to live”. Of course I ‘should’ keep striving for impact, but if that relaxing after I fail makes me more likely to take the risk initially, maybe it’s worth it.
Curated and popular this week
·
Notes
The following text explores, in a speculative manner, the evolutionary question: Did high-intensity affective states, specifically Pain, emerge early in evolutionary history, or did they develop gradually over time?
Note: We are not neuroscientists; our work draws on our evolutionary biology background and our efforts to develop welfare metrics that accurately reflect reality and effectively reduce suffering. We hope these ideas may interest researchers in neuroscience, comparative cognition, and animal welfare science.
This discussion is part of a broader manuscript in progress, focusing on interspecific comparisons of affective capacities—a critical question for advancing animal welfare science and estimating the Welfare Footprint of animal-sourced products.
Key points
Ultimate question: Do primitive sentient organisms experience extreme pain intensities, or fine-grained pain intensity discrimination, or both?
Scientific framing: Pain functions as a biological signalling system that guides behavior by encoding motivational importance. The evolution of Pain signalling —its intensity range and resolution (i.e., the granularity with which differences in Pain intensity can be perceived)— can be viewed as an optimization problem, where neural architectures must balance computational efficiency, survival-driven signal prioritization, and adaptive flexibility.
Mathematical clarification: Resolution is a fundamental requirement for encoding and processing information. Pain varies not only in overall intensity but also in granularity—how finely intensity levels can be distinguished.
Hypothetical Evolutionary Pathways: by analysing affective intensity (low, high) and resolution (low, high) as independent dimensions, we describe four illustrative evolutionary scenarios that provide a structured framework to examine whether primitive sentient organisms can experience Pain of high intensity, nuanced affective intensities, both, or neither.
Introdu
·
We’ve redesigned effectivealtruism.org to improve understanding and perception of effective altruism, and make it easier to take action.
View the new site →
I led the redesign and will be writing in the first person here, but many others contributed research, feedback, writing, editing, and development. I’d love to hear what you think, here is a feedback form.
Redesign goals
This redesign is part of CEA’s broader efforts to improve how effective altruism is understood and perceived. I focused on goals aligned with CEA’s branding and growth strategy:
1. Improve understanding of what effective altruism is
Make the core ideas easier to grasp by simplifying language, addressing common misconceptions, and showcasing more real-world examples of people and projects.
2. Improve the perception of effective altruism
I worked from a set of brand associations defined by the group working on the EA brand project[1]. These are words we want people to associate with effective altruism more strongly—like compassionate, competent, and action-oriented.
3. Increase impactful actions
Make it easier for visitors to take meaningful next steps, like signing up for the newsletter or intro course, exploring career opportunities, or donating.
We focused especially on three key audiences:
* To-be direct workers: young people and professionals who might explore impactful career paths
* Opinion shapers and people in power: journalists, policymakers, and senior professionals in relevant fields
* Donors: from large funders to smaller individual givers and peer foundations
Before and after
The changes across the site are aimed at making it clearer, more skimmable, and easier to navigate. Here are some side-by-side comparisons:
Landing page
Some of the changes:
* Replaced the economic growth graph with a short video highlighting different cause areas and effective altruism in action
* Updated tagline to "Find the best ways to help others" based on testing by Rethink
·
The company released a model it classified as risky — without meeting requirements it previously promised
This is the full text of a post first published on Obsolete, a Substack that I write about the intersection of capitalism, geopolitics, and artificial intelligence. I’m a freelance journalist and the author of a forthcoming book called Obsolete: Power, Profit, and the Race to Build Machine Superintelligence. Consider subscribing to stay up to date with my work.
After publication, this article was updated to include an additional response from Anthropic and to clarify that while the company's version history webpage doesn't explicitly highlight changes to the original ASL-4 commitment, discussion of these changes can be found in a redline PDF linked on that page.
Anthropic just released Claude 4 Opus, its most capable AI model to date. But in doing so, the company may have abandoned one of its earliest promises.
In September 2023, Anthropic published its Responsible Scaling Policy (RSP), a first-of-its-kind safety framework that promises to gate increasingly capable AI systems behind increasingly robust safeguards. Other leading AI companies followed suit, releasing their own versions of RSPs. The US lacks binding regulations on frontier AI systems, and these plans remain voluntary.
The core idea behind the RSP and similar frameworks is to assess AI models for dangerous capabilities, like being able to self-replicate in the wild or help novices make bioweapons. The results of these evaluations determine the risk level of the model. If the model is found to be too risky, the company commits to not releasing it until sufficient mitigation measures are in place.
Earlier today, TIME published then temporarily removed an article revealing that the yet-to-be announced Claude 4 Opus is the first Anthropic model to trigger the company's AI Safety Level 3 (ASL-3) protections, after safety evaluators found it may be able to assist novices in building bioweapons. (The
I don't seem to see the fireside chat with forethought on the agenda, will it be added later? I'd love to attend!