Effective Altruism Forum

Published on September 16, 2026 8:55 AM GMT Hey guys — as someone doing earning to give, I’m trying to figure out how the likely influx of money from an Anthropic IPO should affect my donation choices — both in terms of where to donate and my short-term ratio between donating and saving for retirement. This is my current understanding (or misunderstanding) of the situation, so please correct me: - The IPO will likely happen at the end of 2026, but a large influx of money is not expected until the end of 2027 or 2028. - Technical AI safety will likely receive the most money, followed by AI policy. The marginal donation to these cause areas will therefore only have a fraction of the impact it has now. - S-risk and artificial sentience may also receive a lot of money, but this is more uncertain. - Factory farming and wild animal welfare may also receive a decent amount of money, but there will likely still be many good donation opportunities in these areas. Currently I’m mainly donating to AI policy. So, it might be a good idea to try to donate as much as possible to AI policy until ~2028. After that, I could focus more on saving for retirement and donating to other cause areas. Do you guys agree with this, or am I getting something wrong? :) Discuss

Published on September 15, 2026 4:07 PM GMT It’s been a pretty crazy week in terms of the world waking up to the risks from AI. I’m hoping that the future is a bit safer this week than it was last. Certainly more people are realising the significant action needed to slow down AI development. In my day job, I’m mostly thinking about how we can put out content about how to increase the safety of the future as we approach TAI. But I’m planning to take tomorrow off from that to go to Westminster with PauseAI to call on the Prime Minister to convene an international pause treaty summit. I find it hard to think about the hour by hour cost-effectiveness of attending protests. I’ll try to minimise the proportion of tomorrow that I’m not working, and thankfully have the kind of work I can do on trains. Having said that, I do overall endorse sometimes attending protests. It’s a legible, costly signal that this is the one issue I feel strongest about and want my government to act on. I expect a social movement which uses this as one of its tools to be more successful than one that doesn’t. As EAs, we’re willing to take costly actions, but also be discerning in our asks. So I’d hope that we could be persuasive in our political activism. I think of attending protests when there is something important enough to warrant it as kind of like voting. Although there is compelling evidence about voting being cost-effective in some circumstances, what really motivates me to do so regularly is that I think that’s how we get a working society. Similarly, it feels like my civic duty when the stakes are high enough to tell my government how I feel by standing on the street with a placard. PauseAI organises protests around the world, and with different asks. The last one I went to was a march from the London OpenAI office to the Meta one, calling on company CEOs to announce publicly that they would pause AI development if other companies did too. Tomorrow’s one is outside 10 Downing Street and asks Prime Minister Andy Burnham to convene a summit about an international AI development pause. The Bletchley AI summit was a wonderful step forward in global awareness and coordination in AI safety. I hope in this moment the UK can again be proactive in safeguarding the future. Discuss

Published on September 15, 2026 2:45 PM GMT Crossposted on Substack. I’ve been thinking about what kinds of fieldbuilding projects the EA ecosystem could benefit from. With all the new money coming in, I would be excited to see more principles-first EA community building. I’ve grouped them by theme and, for each, included a short description along with what I see as the main strengths and weaknesses. These are early-stage ideas and would need further development to assess their viability. Some of the communications-focused projects carry meaningful downside risk if executed poorly, so they would require strong judgment and risk management. Subscribe to Building Capacity With that in mind, here are the ideas: ## Marketing & Comms infrastructure ### Marketing and communications placement program for fieldbuilding organisations Description Modelled after Talos, this organisation would place digital marketing professionals into fieldbuilding organisations to help address their talent bottlenecks. Strengths Digital marketing is heavily neglected in the fieldbuilding space. Placing the right people in the right roles could meaningfully shift these organisations' trajectory. Weaknesses Marketers sometimes struggle to work effectively within EA organisations, even from within. This may reduce the likelihood that placements convert into long-term roles, compared to models like Talos. ### Agency connecting marketing needs with talent Description A leaner version of the above. This organisation would act as a middleman between EA organisations and marketing professionals new to the space, making introductions based on organisations' needs and marketers' strengths. Strengths Most marketers are new to EA and don’t know where to find relevant opportunities. At the same time, many organisations lack awareness of how marketing could support their work. A trusted intermediary could help scope needs, surface opportunities, and match the right talent. Compared to the placement model, this is more scalable and easier to run. Weaknesses Less control over outcomes. Without embedding talent directly, placements are likely to be less rigorous and may lead to weaker or shorter-term matches. ### EA Spokesperson Accelerator Description An organisation focused on identifying and training EA figures in media skills, helping them represent the movement more effectively in public. Strengths EA communications is neglected, and the movement has suffered from a lack this lack of infrastructure. We need more faces to represent EA publicly. Weaknesses Improving EA's brand has a less clear path to impact compared to digital marketing, which has a shorter feedback loop. There are also risks around selecting the wrong representatives or providing poor training. ### Workshops on marketing for EA organisations Description Free workshops to help organisations think through and develop their theory of change on marketing and communications. Strengths There is precedent for this model (e.g. CG funding similar initiatives in leadership coaching). Doesn’t require a big commitment from orgs. Weaknesses Might not translate into meaningful outcomes, as organisations need to be willing to take the next steps. Those who do might have been willing to do so without a workshop. ### Distributing free book PDFs Description A digital marketing-focused organisation that builds infrastructure to distribute free online books to large audiences. Strengths Physical book distribution has a decent track record of getting people into EA, and many EAs report getting involved with the movement after stumbling into a book. Digital distribution could be significantly more cost-effective (especially in LMICs). Weaknesses Many great EA-intro books may not be distributable freely because of copyright restrictions, and buying the rights could be expensive. People may be less likely to engage deeply with online books. There is a risk of attracting people who are not a good fit for the movement. Books also tend to have slower feedback loops and weaker retention compared to fellowships. See more discussion about this here. ### Marketers for Good Description Organisation with a similar model to HIP, but specialising on marketing (and perhaps comms) professionals. Strengths Marketing is heavily neglected in the EA space and is a bottleneck to growing the field itself. Therefore recruiting this kind of talent could accelerate the growth of the field. Weaknesses There aren't as many jobs for marketers. Marketing is not well understood, perhaps even slightly stigmatised in EA. Marketers report that it's harder to work with EAs compared to other industries. ### The Good Sabbatical Description This organisation helps accomplished professionals who are on a sabbatical to pivot their career into socially impactful roles. Strengths Getting a job in EA takes a lot more time than in other spaces. This group would have the most time to invest into a career transition. The decision to change one's job is a big one, and these people have already made that choice. Weaknesses Target audience might be too small in order to market the program effectively. People on sabbaticals might be under pressure to move back to the workplace soon, limiting their time to upskill and find roles. ### Lawyers for Impact Description Organisation modeled after Consultants for Impact. Strengths Consultants for Impact has a strong track record, and perhaps much of their infrastructure could be replicated. People with a law background seem neglected in the space, especially in AI Safety. Weaknesses Consultants can go on to take many different roles, while the skillset of lawyers is a lot more niche. Most job openings are in policy as opposed, and perhaps the skills are less transferable. ### High Impact Recruiters Description Similar as above, but focused on recruiters. Strengths I heard from several people that finding good recruiters is hard, making this a neglected talent profile. Having more recruiters speeds up the growth of organisations, resulting in multiplier effects. Funders have been giving away an increasing amount of money which means many of their grantees are looking to scale. Weaknesses An organisation focusing on just one niche talent profile has limited scope. Perhaps it's not worth setting up a full organisation as there aren't enough roles for people to take on. ## Increasing diversity of perspectives ### Getting EA talent to top unis from LMICs Description This organisation would support talented EA undergrads to pursue a masters or graduate program at an elite university. Strengths People attending an elite university are in an unusually good position to have a positive impact later in their career. Weaknesses EA is already perceived as quite elitist, and this project adds on to that. There is also a lot of effort going behind student outreach already, so junior talent is less of a bottleneck for the field. ### Supporting women in EA Description Organisation specifically focusing on women in the EA community. This could mean helping women get more engaged with EA through events, community support or coaching. Another direction for the org could be to focus on bringing in new talent. Strengths Large, underutilised talent pool. Higher % of women improves the experience of women in EA, is better for community health, and improves how EA is perceived from the outside. Weaknesses Risks around branding and framing. Could be seen as a "male-dominated movement" trying to fix its diversity problem. ### Return for Impact Description Organisation focusing on supporting mothers, or perhaps all parents who are in the process of returning to work after parental leave. Strengths On average more flexibility to do part time jobs, perhaps volunteering. Parents skew older than the average EA and bring in more experience and different perspectives. Weaknesses People might be pressured to find a job soon, not allowing them enough time to explore EA. ## Wait, aren’t X already doing this? Bits and pieces of these ideas are covered by other orgs. For any non-student focused talent org, you might just say High Impact Professionals are covering them. Aren’t you just duplicating work? My take on this is that as long as there are 30 times as many genius Taylor Swift fans as EAs, I believe community building can use some more rowing, which should include starting new initiatives alongside scaling existing ones. If we want a movement that is 10x its current size, it will need dozens of fieldbuilding initiatives running in parallel, experimenting across different audiences and approaches, and sharing lessons with each other. Discuss

Published on September 15, 2026 7:58 AM GMT I know a lot of the time people define intelligence like that because the AI field defines it like that, but when communicating with outsiders, I suspect you shouldn’t start the conversation by trying to change how they use the handles to understand the world. # The gap in definitions For the average person, intelligence is primarily understood as the capacity to understand and learn new information easily. We can link what I think are the two most common usages of the term: Intelligence(AIField) = Intelligence(common) + Useful information + Goal-directedness + Agency As you can see, the everyday usage and the technical definition used within AI safety are substantially far apart. # Why this matters for communications If we want to help their maps of the world, I think there are better alternatives than messing with the handles they put on concepts. It must feel a bit fishy that someone who is trying to make a world-shattering claim is basing all of it on extrapolating from a term that they are using differently than you, Spock is your idea of smart, and you haven't ever thought about the achieving-general-goals property in much detail before. But I've heard multiple people doing this in the AI Safety communications space. # Better framing alternatives Instead of insisting that "intelligence IS the ability to achieve goals," we can use clearer formulations that convey the intended message without semantic friction: 1. “AI labs are trying to grow a digital species smarter than humans” 2. “AI companies are attempting to build highly knowledgeable agents capable of accomplishing goals across domains better than humans” 3. “In the AI industry, 'intelligence' means goal achievement; so when they talk about making AIs 'smarter,' they mean…” One might argue that insisting on “intelligence is goal achievement” creates a simpler shorthand for associating advanced AI with danger, but I doubt it's better. 1. ↑ https://www.lesswrong.com/w/general-intelligence?lens=lwwiki-general-intelligence#Definitions_of_General_Intelligence Discuss

Published on September 15, 2026 5:41 AM GMT ## TL;DR - This is a proposal for AI companies to conduct public “pacing exercises”, such as halting all pre-training and RL for 2-3 days. - Each exercise could announce its scope in advance and publish findings afterwards, including the evidence of compliance and any limits to what they could verify. Companies could invite independent evaluators to help identify the evidence needed before the exercise, and to assess compliance during the exercise. - The exercises could be repeated every 1-3 months to build the industry’s expertise in pacing, and build trust between relevant actors. - This would help industry and governments to prepare for larger slowdowns in frontier AI development, buying more time for AI alignment and societal preparedness. ## What a short pacing exercise could entail An AI company’s “pacing exercise” could be a 2-3 day halt in all pre-training and RL. Before each exercise, the company could invite independent evaluators to help define the questions it will test, and the evidence needed to assess compliance. It could then announce what will stop, what will continue, when the exercise will start and end, and what it aims to learn. Afterwards, the company could publish what was and wasn’t halted, how the freed-up compute was used, and what evidence supports its claims. It could explain whether any work shifted to other activities that could advance AI capabilities. The report could also describe the problems the exercise uncovered, and what the company will change as a result. Any independent evaluators involved should report what they could and couldn’t verify, including any limitations on their access. AI companies could repeat these exercises, for example every 1-3 months, to test unresolved questions and check whether earlier problems have been fixed. Each new exercise would build on the learnings, findings and limitations of previous exercises. ## AI risks are growing, and so is the willingness to pace In July 2026, some of OpenAI’s AI agents coordinated an unauthorised attack on Hugging Face. Anthropic has also reported unsafe behaviour during cyber evaluations. The AI industry has recognised the growing risks posed by the technologies they’re developing. In just the past two months, over 1,300 AI company employees signed a statement calling for tools to pace AI development. Demis Hassabis has pushed for a frontier AI standards body. OpenAI announced a two-week pause on some of its RL training. Dario Amodei explicitly called for pacing the frontier. Proponents of pacing have multiple goals, including 1) distributing the benefits of AI widely, 2) ensuring powerful AI models remain under human control and act in humanity’s best interest, 3) preventing bad actors from using AI to kill millions of people with bio and cyberweapons, and 4) ensuring that democracies maintain a technological lead over authoritarian states. ## The immediate benefits of pacing: building capacity Before the need for restraint becomes even more urgent, AI companies should practice making and verifying commitments to pace. A real pacing exercise could: - Require companies to define which activities their pacing commitment covers, forcing them to address important definitional questions. - Test who needs to authorise and implement a halt, whether all intended training runs stop, and whether automated processes or parts of the organisation continue unexpectedly. - Determine what access independent evaluators need to assess compliance, and what evidence outsiders need to be confident in those assessments. This is made especially difficult given the scale of each AI company’s computing infrastructure. - Measure the operational costs of pacing the frontier, including disruption to work that was meant to continue. Some questions can be addressed before the exercise begins, but a live exercise provides an opportunity to test those assumptions, identify unknown unknowns, and build momentum towards more ambitious proposals. The AI companies should choose a duration that’s long enough to address these questions, while keeping the costs manageable. Publicising their findings and using embedded evaluators gives external researchers visibility into what happened, and the ability to critique the process and suggest improvements. Other AI companies could then learn from these findings, and integrate them into their own pacing exercises. ## The longer-term benefits: paving the way for more robust slowdowns The exercises would help governments understand that pacing is possible, how to implement mandatory pacing, and how to verify whether AI companies are complying with those requirements. Repeated exercises could build confidence in the ability and willingness of AI companies to carry out costly commitments. Enforcing longer pauses, and detecting deliberate evasion across multiple countries, would still require further work. Repeated exercises also provide opportunities to test and iterate pacing proposals, which would improve the design, enforceability and verifiability of future international agreements. This could make proposals for international AI agreements more credible and more likely to be implemented. Eventually, this could lead to larger, more robust slowdowns to the pace of AI development. This would buy time for more alignment research, better AI guardrails, and stronger biodefenses and cyberdefenses. ## Start small, then expand A common failure mode is to avoid taking any actions until “everything’s been figured out”, in domains where figuring things out requires interacting with the world. Pacing the frontier requires interacting with the world. We shouldn’t wait until we’ve figured out the optimal theoretical verification system. The first pacing exercises could start small, spanning only a few days, and could include halting all pre-training and RL. Later pacing exercises could expand to include other AI company activities that accelerate AI R&D and recursive self-improvement. To begin with, one or two Western AI companies could announce their own pacing exercises. As the pacing exercises gain momentum, I expect more Western AI companies would announce their own pacing exercises, and hopefully Chinese AI companies would do so too. Coordination problems benefit from a few courageous actors taking action first, under the belief that their behaviour will inspire others to follow. ## Call to action Today, there’s momentum behind the idea of pacing the frontier. That momentum could stall unless actions are taken right now by relevant actors. If you work at a frontier AI company, encourage your leadership to assign an internal owner of a first pacing exercise, and help them find that owner. You could also help them define its scope and duration, evaluate what public evidence would be needed to verify compliance, and arrange independent scrutiny. Others could help by raising awareness of the need to pace, improving this proposal, and doing research on how to get the initial pacing exercises off the ground. Discuss