The Nonlinear Library

The Nonlinear Library

By The Nonlinear FundEducation
Download on the App Store

All The Nonlinear Library episodes

  • LW - [New LW Feature] "Debates" by Ruby
    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: [New LW Feature] "Debates", published by Ruby on April 1, 2023 on LessWrong.
    Following the success of the 2021 MIRI Conversations in drawing out various people's views on key questions in AI and digging into their disagreements, the LessWrong team was inspired to build a more dedicated feature for hosting debates on LessWrong.
    The MIRI conversations were mostly hosted on Discord and then via a laborious process shoehorned into the LessWrong editor. We figured it wouldn't be hard to do better. Among many benefits, the debates can be held on LessWrong itself; readers are able to comment "inline" on responses within the debate; and there will be customized "debate item" in the Latest Posts list on the frontpage that signals that 1) it's a debate, and 2) how many debate responses have been posted since you last viewed the debate. Hopefully all of this is intuitive from the UI.
    The feature is designed so that debates can be held in private, possibly edited, and then published publicly. Or that the debate happens live on the site, allowing for live commenting.
    As we're rolling out the feature, we'll initially just set up a few debates that we'd like to see, and then later potentially open up the feature to users more broadly. You're welcome to contact us [link] or comment here if you're interested in viewing or participating in a particular debate.
    This announcement post will also serve as the Inaugural Debate using the new debate feature. We were lucky to find two willing participants on short notice, so big thanks to them. GPT-4 and Claude+ are going to discuss whether or not AI Safety via Debate is a promising Alignment strategy.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • EA - Iterating on our redesign (Forum update April 2023) by Sarah Cheng
    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: Iterating on our redesign (Forum update April 2023), published by Sarah Cheng on April 1, 2023 on The Effective Altruism Forum.
    Thank you to everyone who gave us feedback on our recent redesign of the Forum! We’ve looked closely at your feedback and our findings from interviews with new users.
    Two clear themes emerged:
    New users found the site too serious and intimidating
    Experienced users missed the amount of “character” in the original design
    After careful consideration, we’ve decided to change our sans-serif font from Inter to Comic Sans. We believe that this friendly and distinctive font will address the concerns of both new and experienced users, and keep users engaged for longer.
    We’re confident that all of our users will be happy with this font change. We welcome you to contact us with your praise, or post it in the comments below.
    However, we realize that we made this change without any heads-up, and that can feel jarring. If you’d like to ease the transition, we’ve added an option in your account settings to switch back to Inter for today only.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    2 min
  • EA - Announcing the April 1st Purchase of Bodhi Kosher Vegetarian Restaurant by EA NYC by Rockwell
    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: Announcing the April 1st Purchase of Bodhi Kosher Vegetarian Restaurant by EA NYC, published by Rockwell on April 1, 2023 on The Effective Altruism Forum.
    Dear Effective Altruism community,
    We are thrilled to announce that Bodhi Kosher Vegetarian Restaurant—a historic estate near Canal Street, serving delicious vegan dim sum since 2018—has been selected as the location for a new EA NYC community center! Effective altruism is all about maximizing the impact of our actions to help others. And what better way to do that than by owning a restaurant?
    The restaurant is in the process of being established as a convening center to run workshops and meetings that bring together people to think seriously about how to address important issues. We believe that the cooperative process of deciding on shared dim sum dishes, the dynamic nature of lazy susan tables, and the abundance of potent black tea, together foster a collaborative and creative environment key to solving the world's most pressing problems.
    EA NYC already runs dozens of events in the restaurant per year (disproportionately in the back right corner, though sometimes near the front door), and we saw several reasons a venue purchase could be valuable for both increasing the number and quality of impactful events, and potentially saving—or even generating!—some money:
    Excellent value for services (such as a full plate of vegetable BBQ meat for only $8.95)
    Ideal location (in the heart of Lower Manhattan, easily accessible for EA NYC staff and community members)
    Existing menu can be optimized to ensure maximum nutritional value per dollar spent
    Both small and large tables available (to accommodate a variety of event sizes)
    Warm atmosphere (especially near the kitchen)
    Access to unique forecasting technologies (free fortune cookies with each meal!)
    We have carefully vetted our fortune supplier to eliminate any morally dubious subcontractors and will not be using whatever company made these:
    Healthy work/community culture (supported by the fortune cookies!):
    The vision is modeled on traditional specialist conference centres, e.g. Oberwolfach, The Rockefeller Foundation Bellagio Center, the Brocher Foundation, and Whytham Abbey - but with a New York City twist. Unlike the aforementioned conference centres, ours will have the advantage of generating additional revenue through existing, robust dine-in and take-out business. We will also spell it “center,” because we live in the United States.
    The purchase was made from a large grant made specifically for this. There was no money from FTX or affiliated individuals or organizations. Just trust us on this, okay?
    When Open Philanthropy surveyed ~200 people involved or interested in longtermist work, they found that many had been strongly influenced by in-person events they’d attended, particularly ones where people relatively new to dim sum came into contact with a dozen or more dishes in one meal. Many developed a taste for textured vegetable protein, which we believe with 80% credence will extend into the far-future. (Some other data and our general experiences in the space are largely supportive of this too, including the popularity of Manhattan vegan chain Beyond Sushi as a venue for EA NYC subgroup meetings.)
    We understand that some members of the community may have concerns about the acquisition of a restaurant and its alignment with effective altruism principles. However, we want to assure everyone that this decision was not taken lightly. We carefully evaluated Bodhi's track record and business practices—pursuing third-party vetting by esteemed venture capital company Sequoia Capital—and found that it aligns with our values, mission, and EVPM (expected value profit margins).
    We also spent a lot of time contemplating the vibe, and considering whether gathering in this veg...
    6 min
  • EA - Hooray for stepping out of the limelight by So8res
    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: Hooray for stepping out of the limelight, published by So8res on April 1, 2023 on The Effective Altruism Forum.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min
  • LW - Shannon's Surprising Discovery by johnswentworth
    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: Shannon's Surprising Discovery, published by johnswentworth on March 30, 2023 on LessWrong.
    Imagine that you’re a telecom provider in the early 20th century. You provide both telephone lines and telegraph lines - audio and binary. These two systems face quite different design constraints.
    The telegraph line only needs to be able to distinguish between high and low voltage. If the signal needs to be amplified or denoised, a relay suffices, a magnetic switch which will shut under voltage above some cutoff or open under any voltage below the cutoff. A noisy incoming signal turns into a clean outgoing signal, so long as the noise isn’t so large as to cross the voltage cutoff.
    A telephone line is trickier. Long-range telephone lines require amplification of continuous voltages, without much distortion or noise. It’s a fairly different engineering problem, for which various arrangements of electrodes in evacuated tubes were used.
    An early twentieth century electrical engineer working on these systems would probably have told you that different kinds of signals require different kinds of transmission hardware. Audio has different needs from Morse code, both of which have different needs from television. Sure, you could send Morse code over a telephone line, but it would be wildly inefficient. Also, different kinds of messages need more or less reliability, and so can tolerate more or less noise in the lines. Try to send a message which needs high reliability over a very noisy line, and you’ll need a lot of redundancy. Again, wildly inefficient.
    Claude Shannon’s surprising discovery was that this is basically false. Different kinds of signals do not require different kinds of hardware, at least not for efficient throughput. (Though latency requirements can and do still sometimes drive hardware specialization.)
    Shannon showed that bits - i.e. zeros and ones, high and low voltages, any signal which can take one of two states - are a universal way to represent information. They’re “universal” in the sense that:
    We can represent any signal in bits
    Given any transmission line, we can build an encoder to take a signal represented in bits and send it over the line
    Most importantly, we can do both of the above with arbitrarily-close-to-optimal throughput, amortized over long messages
    As an added bonus, we can achieve arbitrarily low error rate, even over a noisy channel, with arbitrarily low overhead, again amortized over long messages
    In short: different kinds of signals do not, in principle, require different kinds of transmission-line hardware. We can just use one representation for everything, with arbitrarily little efficiency loss, with some clever encoding and decoding.
    That was the big surprise which originally made telecoms engineers excited about information theory - or communication theory, as it was called at the time.
    Aside: Fungibility of Information Channel Capacity
    For purposes of studying agency, one major consequence of Shannon’s discovery is that information channels are fungible - capacity is limited, but can be used to transmit whatever information we please, so long as we can build a suitable encoder and decoder.
    That makes information channel capacity a natural resource. Like money or energy, it’s limited, but can be “spent” on a wide variety of different things.
    Like money or energy, there’s usually some work required before spending information capacity. With money we need to find people to trade with who have or produce what we want. With energy we need to build a machine to turn energy into whatever we want. With information we need to build the encoder and decoder. But the resource itself is flexible.
    Viewing information as a resource tells us important things mainly insofar as that resource is scarce - i.e. mainly insofar as there’s limited information c...
    16 min
  • LW - Hooray for stepping out of the limelight by So8res
    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: Hooray for stepping out of the limelight, published by So8res on April 1, 2023 on LessWrong.
    From maybe 2013 to 2016, DeepMind was at the forefront of hype around AGI. Since then, they've done less hype. For example, AlphaStar was not hyped nearly as much as I think it could have been.
    I think that there's a very solid chance that this was an intentional move on the part of DeepMind: that they've been intentionally avoiding making AGI capabilities seem sexy.
    In the wake of big public releases like ChatGPT and Sydney and GPT-4, I think it's worth appreciating this move on DeepMind's part. It's not a very visible move. It's easy to fail to notice. It probably hurts their own position in the arms race. I think it's a prosocial move.
    If you are the sort of person who is going to do AGI capabilities research—and I recommend against it—then I'd recommend doing it at places that are more likely to be able to keep their research private, rather than letting it contribute to an arms race that I expect would kill literally everyone.
    I suspect that DeepMind has not only been avoiding hype, but also avoiding publishing a variety of their research. Various other labs have also been avoiding both, and I applaud them too. And perhaps DeepMind has been out of the limelight because they focus less on large language models, and the results that they do have are harder to hype. But insofar as DeepMind was in the limelight, and did intentionally step back from it and avoid drawing tons more attention and investment to AGI capabilities (in light of how Earth is not well-positioned to deploy AGI capabilities in ways that make the world better), I think that's worth noticing and applauding.
    (To be clear: I think DeepMind could do significantly better on the related axis of avoiding publishing research that advances capabilities, and for instance I was sad to see Chinchilla published. And they could do better at avoiding hype themselves, as noted in the comments. At this stage, I would recommend that DeepMind cease further capabilities research until our understanding of alignment is much further along, and my applause for the specific act of avoiding hype does not constitute a general endorsement of their operations. Nevertheless, my primary guess is that DeepMind has made at least some explicit attempts to avoid hype, and insofar as that's true, I applaud the decision.)
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - Keep Making AI Safety News by RedStateBlueState
    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: Keep Making AI Safety News, published by RedStateBlueState on March 31, 2023 on The Effective Altruism Forum.
    AI Safety is hot right now.
    The FLI letter was the catalyst for most of this, but even before that there was the Ezra Klein OpEd piece in the NYTimes. (Also general shoutout to Ezra for helping bring EA ideas to the mainstream - he's great!).
    Since the FLI letter, there was there was this CBS interview with Geoffrey Hinton. There was this WSJ Op-Ed. Eliezer's Time OpEd and Lex Fridman interview led to Bezos following him on Twitter. Most remarkably to me, Fox News reporter Peter Doocey asked a question in the White House press briefing, which got a serious (albeit vague) response. The president of the United States, in all likelihood, has heard of AI Safety.
    This is amazing. I think it's the biggest positive development is AI Safety thus far. On the safety research side, the more people hear about AI safety, the more tech investors/philanthropists start to fund research and the more researchers want to start doing safety work. On the capabilities side, companies taking AI risks more seriously will lead to more care taken when developing and deploying AI systems. On the policy side, politicians taking AI risk seriously and developing regulations would be greatly helpful.
    Now, I keep up with news... obsessively. These types of news cycles aren't all that uncommon. What is uncommon is keeping attention for an extended period of time. The best way to do this is just to say yes to any media coverage. AI Safety communicators should be going on any news outlet that will have them. Interviews, debates, short segments on cable news, whatever. It is much less important that we proceed with caution - making sure to choose our words carefully or not interacting with antagonistic reporters - than that we just keep getting media coverage. This was notably Pete Buttigieg's strategy in the 2020 Democratic Primary (and still is with his constant Fox News cameos), which led to this small-town mayor becoming a household name and the US Secretary of Transportation.
    I think there's a mindset among people in AI Safety right now that nobody cares and nobody is prepared and our only chance is if we're lucky and alignment isn't as hard as Eliezer makes it out to be. This is our chance to change that. Never underestimate the power of truckloads of media coverage, whether to elevate a businessman into the white house or to push a fringe idea into the mainstream. It's not going to come naturally, though - we must keep working at it.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    3 min
  • EA - New Cause Area: Portrait Welfare (+introducing SPEWS) by Luke Freeman
    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: New Cause Area: Portrait Welfare (+introducing SPEWS), published by Luke Freeman on March 31, 2023 on The Effective Altruism Forum.
    "The question is not, Can they reason?, nor Can they talk? but, Can they (I mean, hypothetically speaking, perhaps, just a smidgen, in theory) suffer" ~ Jeremy Bentham
    Introduction
    The effective altruism community has consistently pushed the frontiers of knowledge and moral progress, demonstrating a willingness to challenge conventional norms and take even the most unconventional ideas seriously. Our concern for global poverty is often considered "weird" as we highlight the importance of valuing individuals' well-being equally, regardless of geographical boundaries. In contrast, broader society tends to focus more on helping people within our own countries, inadvertently giving less consideration to those further afield. From animal welfare to long-term existential risks, our community is full of people who have expanded their moral circles even further to include the suffering of non-human animals and future generations.
    Now, avant-garde effective altruists are exploring the outer limits of moral concern, delving into areas such as insect welfare and digital minds. As we celebrate these accomplishments, we remain committed to broadening our understanding and seeking out new cause areas that may have a significant, overlooked impact.
    Imagine a future where we have made substantial strides in addressing these critical issues, and you find yourself sipping tea in a room adorned with stunning portraits. As you revel in this moment of tranquillity, a thought experiment crosses your mind: What if the portraits themselves deserve our moral consideration? And while we were busy tackling other pressing matters, could we have been inadvertently overlooking yet another human atrocity?
    Today, we invite you to entertain this intriguing and unconventional idea as we introduce the new cause area of Portrait Welfare. While initially sceptical, our research has led us to be surprisingly confident in the potential of this cause to be the much-awaited “Cause X.” To demonstrate our convictions we have registered our predictions on Manifold Markets, and at current market rates, a rational actor placing a modest bet of the median US salary could stand to win an impressive sum of over $12 trillion USD (in 2023 dollars) by market close.
    As we embark on this journey into uncharted territory, we encourage you to keep an open mind and dive into this fascinating new area of concern. Together, we can continue to push the boundaries of our impact and make the world a better place for all sentient beings – even those that exist within the confines of a frame. And if you're not on board with this yet, just remember, every time you hang a portrait on your wall or snap a selfie, there may be a possibility that you're contributing to a system of injustice and suffering. However, we understand that not everyone can see the bigger picture, and we won't judge you too harshly if you've done all you can to understand this possibility but still cannot accept it. After all, we are all on our own journey towards a more ethical and compassionate world.
    The Moral Case for Portrait Welfare
    While the notion of portrait welfare may initially seem far-fetched, there are moral reasons to consider this cause area. If it turns out that portraits possess a form of consciousness, it would be our ethical responsibility to address their welfare. In line with the principles of effective altruism, we ought to explore all possibilities that could lead to a reduction in suffering, even if they are unconventional.
    Expanding the circle of compassion: The effective altruism movement aims to reduce suffering for all sentient beings, regardless of species or other differences. By considering portrait...
    28 min
  • LW - Manifund x AI Worldviews by Austin Chen
    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: Manifund x AI Worldviews, published by Austin Chen on March 31, 2023 on LessWrong.
    Thanks for listening. To help us out with The Nonlinear Library or to learn more, please visit nonlinear.org.
    1 min
  • EA - My updates after FTX by Benjamin Todd
    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: My updates after FTX, published by Benjamin Todd on March 31, 2023 on The Effective Altruism Forum.
    Here are some thoughts on what I’ve learned from what happened with FTX, and (to a lesser degree) other events of the last 6 months.
    I can’t give all my reasoning, and have focused on my bottom lines. Bear in mind that updates are relative to my starting point (you could update oppositely if you started in a different place).
    In the second half, I list some updates I haven’t made.
    I’ve tried to make updates about things that could have actually reduced the chance of something this bad from happening, or where a single data point can be significant, or where one update entails another.
    For the implications, I’ve focused on those in my own work, rather than speculate about the EA community or adjacent communities as a whole (though I’ve done some of that). I wrote most of this doc in Jan.
    I’m only speaking for myself, not 80,000 Hours or Effective Ventures Foundation (UK) or Effective Ventures Foundation USA Inc.
    The updates are roughly in logical order (earlier updates entail the later ones) with some weighting by importance / confidence / size of update. I’m sorry it’s become so long – the key takeaways are in bold.
    I still feel unsure about how best to frame some of these issues, and how important different parts are. This is a snapshot of my thinking and it’s likely to change over the next year.
    Big picture, I do think significant updates and changes are warranted. Several people we thought deeply shared our values have been charged with conducting one of the biggest financial frauds in history (one of whom has pled guilty).
    The first section makes for demoralising reading, so it’s maybe worth also saying that I still think the core ideas of EA make sense, and I plan to keep practising them in my own life.
    I hope people keep working on building effective altruism, and in particular, now is probably a moment of unusual malleability to improve its culture, so let’s make the most of it.
    List of updates
    1. I should take more seriously the idea that EA, despite having many worthwhile ideas, may attract some dangerous people i.e. people who are able to do ambitious things but are reckless / self-deluded / deceptive / norm-breaking and so can have a lot of negative impact. This has long been a theoretical worry, but it wasn’t clearly an empirical issue – it seemed like potential bad actors either hadn’t joined or had been spotted and constrained. Now it seems clearly true. I think we need to act as if the base rate is >1%. (Though if there’s a strong enough reaction to FTX, it’s possible the fraction will be lower going forward than it was before.)
    2. Due to this, I should act on the assumption EAs aren’t more trustworthy than average. Previously I acted as if they were. I now think the typical EA probably is more trustworthy than average – EA attracts some very kind and high integrity people – but it also attracts plenty of people with normal amounts of pride, self-delusion and self-interest, and there’s a significant minority who seem more likely to defect or be deceptive than average. The existence of this minority, and because it’s hard to tell who is who, means you need to assume that someone might be untrustworthy by default (even if “they’re an EA”). This doesn’t mean distrusting everyone by default – I still think it’s best to default to being cooperative – but it’s vital to have checks for and ways to exclude dangerous actors, especially in influential positions (i.e. trust, but verify).
    3. EAs are also less competent than I thought, and have worse judgement of character and competence than I thought. I’d taken financial success as an update in competence; I no longer think this. But also I wouldn’t have predicted to be deceived so thoroughly, so I negatively updat...
    31 min