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: [MLSN #3]: NeurIPS Safety Paper Roundup, published by Dan Hendrycks on March 8, 2022 on The AI Alignment Forum.

As part of a larger community building effort, I am writing a safety newsletter which is designed to cover empirical safety research and be palatable to the broader machine learning research community. You can subscribe here or follow the newsletter on twitter here.

Welcome to the 3rd issue of the ML Safety Newsletter. In this edition, we cover:

NeurIPS ML safety papers

experiments showing that Transformers have no edge for adversarial robustness and anomaly detection

a new method leveraging fractals to improve various reliability metrics

a preference learning benchmark

... and much more.

Robustness

Are Transformers More Robust Than CNNs?

This paper evaluates the distribution shift robustness and adversarial robustness of ConvNets and Vision Transformers (ViTs). Compared with previous papers, its evaluations are more fair and careful.

After controlling for data augmentation, they find that Transformers exhibit greater distribution shift robustness. For adversarial robustness, findings are more nuanced. First, ViTs are far more difficult to adversarially train. When successfully adversarially trained, ViTs are more robust than off-the-shelf ConvNets. However, ViTs’ higher adversarial robustness is explained by their smooth activation function, the GELU. If ConvNets use GELUs, they obtain similar adversarial robustness. Consequently, Vision Transformers are more robust than ConvNets to distribution shift, but they are not intrinsically more adversarially robust.

Fractals Improve Robustness (+ Other Reliability Metrics)

PixMix improves both robustness (corruptions, adversaries, prediction consistency) and uncertainty estimation (calibration, anomaly detection).

PixMix is a data augmentation strategy that mixes training examples with fractals or feature visualizations; models then learn to classify these augmented examples. Whereas previous methods sacrifice performance on some reliability axes for improvements on others, this is the first to have no major reliability tradeoffs and is near Pareto-optimal.

Other Recent Robustness Papers

A new adversarial robustness state-of-the-art by finding a better way to leverage data augmentations.

A highly effective gradient-based adversarial attack for text-based models.

A new benchmark for detecting adversarial text attacks.

Adversarially attacking language models with bidirectional and large-scale unidirectional language models.

First works on certified robustness under distribution shift: [1], [2], [3].

A dataset where in-distribution accuracy is negatively correlated with out-of-distribution robustness.

Improving performance in tail events by augmenting prediction pipelines with retrieval.

A set of new, more realistic 3D common corruptions.

Multimodality can dramatically improve robustness.

Monitoring

Synthesizing Outlier for Out-of-Distribution Detection

The authors model the hidden feature representations of in-distribution examples as class-conditional Gaussians, and they sample virtual outliers from the low-likelihood region. The model is trained to separate in-distribution examples from virtual outliers.

A path towards better out-of-distribution (OOD) detection is through generating diverse and unusual examples. As a step in that direction, this paper proposes to generate hidden representations or “virtual” examples that are outliers, rather than generate raw inputs that are outliers. The method is evaluated on many object detection and classification tasks, and it works well. It is not evaluated on the more difficult setting where anomalies are held-out classes from similar data generating processes. If the authors evaluated their CIFAR-10 model’s ability to detect CIFAR-100 anomalies, then we would have more of...

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: Hello from the new Content Specialist at CEA, published by Lizka on March 8, 2022 on The Effective Altruism Forum.

Hello!

The tl;dr:

I’m the new Content Specialist at the Centre for Effective Altruism, taking over from Aaron Gertler.

I’m taking on a lot of his old responsibilities, and you can reach out to me if you have any questions about or issues with the Forum.

More below.

My name is Lizka. You may have seen some of my old posts. You’ll likely be seeing me post more frequently now. :)

Some things I’ll be doing in my new role

Running the EA Newsletter and the Forum Digest

Encouraging people to post on the Forum

Or to sign up and add a bio

Or to comment, or upvote

Etc.

Offering Forum workshops (e.g. at EAGx Oxford)

Offering feedback on drafts

Some moderation

Setting up AMAs and contests

General Forum support — if you’re experiencing an issue with the Forum, please reach out!

Generally trying to make the Forum as great and welcoming as possible

Some things I won’t be doing

Awesome software engineering that makes the Forum actually exist

Reading and responding to every post, sadly (because, happily, there are too many)

Feel free to get in touch with me!

DM me on the Forum or email me at [email protected]. (And you can always just comment on this post.)

Somewhat random image

I have yet to meet another “Lizka” in the EA community. Please let me know if your name is also Lizka! (And as a minor point of clarification, my name is pronounced: "lease-kah.")

At least for now. We’ll see exactly how my role evolves and we’re leaving it somewhat flexible to adapt to how things go.

I think more posts should have pictures. This is a visualization I made of certain points’ orbits under the action of a kleinian group.

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

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: Grantmaking is more like a skill than a path, published by OllieBase on March 8, 2022 on The Effective Altruism Forum.

Epistemic status: Moderately strongly held based on anecdotal data, but I could easily imagine learning about broader patterns and withdrawing this claim.

TL;DR: Grantmaking, in my view, is best viewed as a set of skills or a stage in a career path than a career path itself. To be excellent at grantmaking, I think you need to be an excellent generalist in a specific field first.This post is mostly a critique of the 80,000 Hours career profile 'Grantmaker focused on pressing problems’. I agree with most of the article but object to the framing (grantmaking as a career path). 80,000 Hours staff reviewed this post and might make some changes to their article, but I’ll let them comment on that if they want to.

My perspective

Without intending to, I have spent a good chunk of my career trying to be a good grantmaker. I spent ~1/3rd of ~1.5 years at what is now Longview Philanthropy making grants to support projects focused on protecting future generations and I now make grants as part of my role as Community Events Manager at CEA.

I don’t claim to be a good or an experienced grantmaker but I have been asked by a few people who are interested in grantmaking what they should do now to become a grantmaker focused on pressing problems.

I find myself explaining to people who ask this question what I wish I knew before I tried grantmaking: to be excellent at grantmaking you need to be an excellent generalist in a specific field first. More specifically, I think you need at least the following:

A “bird’s eye view” of a field - Which projects already exist? What is the track record of work in the space? Who’s who?

A theory of change - What needs to happen in your field to solve an important problem? Why hasn’t that happened already? What’s the best path to the solution?

A strong network - You often need to be able to talk to someone working in the field to understand whether a new project might work. Who do you know who could pick up the phone to explain something to you?

Technical know-how - Can you quickly understand and vet proposals for projects in your field? Can you spot errors?

I think the best way to attain this set of skills is to pursue a generalist path in a specific field, rather than pursue a “grantmaker path”. Examples of more generalist paths might be researcher, policy analyst or manager.

Much of this is covered in the 80,000 Hours article and I agree with most of that article. But I think that the framing of grantmaking as a “path,” rather than a “skill” or “stage,” is confusing for two reasons:

Framing grantmaking as a path encourages people to try to learn generalist skills outside the context of a specific problem.

Framing grantmaking as a path encourages people early in their career to apply for an extremely narrow set of jobs (e.g. research analyst at Open Phil).

Framing grantmaking as a path encourages people to try to learn generalist skills outside the context of a specific problem.

I agree with the contents of the section “How to assess your fit” but I worry that people early in their career will try to work on things like 'judging people accurately' or 'thinking of grant ideas' before they have built a birds-eye view of a field or theory of change for a problem.

For example, I spoke to a student at EA Global who was interested in becoming a grantmaker in AI safety. I asked them what they thought of various lines of attack on the alignment problem and they said something along the lines of “I’m not sure I want to do that kind of research directly, I’m more of a people person which is why I’m focused on becoming a grantmaker.” This sounds like it could be a big career misstep - it would be very hard to be a good grantmaker in the field of...

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: How to Best Use Twitter, published by Zvi on March 7, 2022 on LessWrong.

[Note: While I do intend to write more about Russia’s invasion of Ukraine, this post is intended to address this only indirectly rather than directly, by helping illustrate how to find other sources of information, and you are once again implored to rely on other news sources. I need to get inside the war’s OODA loop to write about it usefully, which requires more dedicated time than I’ve had this past week.]

It is high time for me to talk about the only practical way I know about to follow developments in the world in real time, whether they be a war, a pandemic or something else entirely, which is Twitter and in particular Twitter Lists.

I do not know of any practical alternative. One can of course watch or read the usual news reports, which are mostly effectively State Media of various quality, for various different States. When you’re reading about an actual war, the State part of State Media becomes more prominent and harder to miss.

The best TV sources for international events like the war that I have are Bloomberg (as far as I can tell, the closest thing to non-state media) for the economic side of things and the BBC World News for the more general side. Occasionally I’ll take a glance at CNN or Fox News or various other networks to get a sense of what they are focusing on. I have not attempted to directly watch any Russian broadcasts but am curious what is the best option for that.

For domestic American events, there are no non-obvious TV sources I have found, and TV is essentially useless other than to know what the Narratives are saying or to cover discrete events like debates, elections or the State of the Union. Any kind of #Analysis is strictly fuhgeddaboudit.

For written media, the usual suspects are what they are so choose your favorites. None of them seem able to keep up with the pace of play other than Bloomberg offering insight into markets, so they are mostly again useful for ‘how are things being presented and sold’ than insight into actual events.

For what is happening, Twitter is where it is at.

To use Twitter properly, there are four vital pieces of technology.

Tweetdeck or another similar alternative application.

Knowing who to follow and read.

Lists.

Unfollows, filters, mutes and blocks.

I’ll also give advice for how to start out, and some thoughts on regulating use.

Tweetdeck

Twitter by default will present posts using an algorithm. Things it thinks you will be more interested in, or that will generate more engagement, or that otherwise it decides on a whim to show you, will be shown to you. Things it disfavors will be dropped on the floor.

This is no good. The biggest issues are:

It makes it impossible to follow things in real time.

It makes it impossible to reliably see things you decide you want to see.

It makes it impossible to know when you’re ‘done’ with new content.

It makes it impossible to avoid seeing things more than once.

It inserts random stuff you never asked for.

Twitter’s website is fine for sending direct messages and chatting. It’s acceptable for checking individual profiles to see what they have to say and whether you want to include such a person in your information diet. It’s also fine for viewing an individual list. That’s another reason to emphasize lists over following people.

However, for the purpose of following the people you follow, Twitter’s website is useless. You want to use something else.

My choice has been Tweetdeck. Tweetdeck allows you to have lots of columns, one for people you follow, one for notifications and then any number of additional lists. Also it’s free.

It is definitely not perfect. When I want to explore a particular Tweet or screenshot it, I still open a default Twitter window for that.

So there is probably something out ...

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: You can tell a drawing from a painting, published by alkexr on March 7, 2022 on LessWrong.

Epistemic status: I've had this concept for a while, and applied it in practice many times. It's fairly easy to convey in an argument, and can clear up some confusion. But there are many loose ends, and I'm not sure it's well formulated. Also, I'm stuck, so I'll just throw it out there, hoping that it grows into something.

Drawings and paintings

You are presented with a .bmp file containing a digitalized version of two art pieces. You are told that one is a drawing and the other is a painting. Can you tell which one is which?

On one hand, of course you can't. The .bmp file format doesn't contain information about the history of some clumps of matter on some rock in space. All you have is the color of each pixel (plus some technical stuff). Clearly, the one that looks like a painting could be the result of the artist very carefully drawing every tiny square milimeter of the image so that the digitalized version is equivalent, pixel by pixel, to the .bmp file you are looking at right now...

On the other hand, that's not what happened. The only way doubt can even arise is if the images were adversarially designed for this very purpose. Even then, it'd take a skillful artist and some serious effort to succesfully confuse people, especially other artists.

Plotting

Some writers prefer to start working on their novel by planning the whole plot before actually putting words on paper, while others just start writing a story and see where it goes. If you haven't encountered this distinction before, you may want to pause for a few seconds to figure out how it will affect the sort of book that results.

Stories that are written without a pre-planned plot tend to be more locally consistent. The characters' decisions are not constrained by where the plot needs to go, so they will be better aligned with their personalities, motivations, and pasts. Consequences follow more clearly from causes, because the causes aren't backwards-engineered to fit the pre-determined consequences. But you'll get a meandering story that doesn't seem to move towards a conclusion. A good example is A Song of Ice and Fire. Stories that are written with a clear outline planned in advance feel more coherent as a whole, have a conclusion, and plot twists make sense. However, characters may at times act in ways that make little sense just to keep the plot going. An example is the Harry Potter series.

But you could just write the exact same words with the other method, right?

Bottom lines

You are considering an argument. Did the person making it draw the conclusion from the argument, or did they tailor the argument to fit their pet conclusion?

In theory, you can come up with the same argument either way. In practice though...

Go

So your friend has decided to decorate an empty wall in their room with two images of board states of Go after 50 moves. One, apparently, was produced by your friend downloading some Go software and throwing stones on the board against an AI. The other one is a game between two professionals. There is a clear difference between the two, even functionally: Go players visiting that room will draw aesthetic pleasure from one but not the other.

Of course there is nothing stopping your friend from playing moves indistinguishable from those of a top player. In practice, though, he can't. What's more, I think it's impossible for a professional to mimic the play of a beginner, even with access to a source of randomness. And yet, if you aren't a Go player, to you the two may look fundamentally no different. Telling the difference requires at least a little expertise.

Programming languages

You are presented two long lists of numbers. One is a list of prime numbers generated by an algorithm in Python, the ot...

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 CERI Summer Research Fellowship, published by Dewi Erwan on March 7, 2022 on The Effective Altruism Forum.

The Cambridge Existential Risks Initiative (CERI, pronounced /ˈkɛri/) has opened applications for an in-person, paid, 10-week Summer Research Fellowship focused on existential risk mitigation, taking place from July to September 2022 in Cambridge, and aimed at all aspiring researchers, including undergraduates.

To apply and find out more, please visit the CERI Fellowship website.

If you’re interested in mentoring fellows on this programme, please submit your name, email and research area here, and we will get in touch with you in due course.

The deadline to apply is April 3 2022 23:59 UTC.

The CERI Fellowship is a fantastic opportunity to:

Build your portfolio by researching a topic relevant to understanding and mitigating existential risks to human civilisation.

Receive guidance and develop your research skills, via weekly mentorship from a researcher in the field.

Form lasting connections with other fellows who care about mitigating existential risks, while also engaging with local events including discussions and Q&As with experts.

Why we are running this programme

Our mission as an organisation is to reduce the probability of an existential catastrophe. We hope to do this by creating more opportunities for aspiring researchers, thus increasing the number of people with the motivation, skills, relevant knowledge, relationships and career options to mitigate existential risks. Upcoming posts will give more concrete and specific theories of change for our main cause areas.

What we provide

A generous stipend of £5,600 (approx. $7,400) for 10 weeks.

Mentorship from a researcher working in a related field.

Free accommodation at a Cambridge college, plus food, all travel expenses and a desk to work with other fellows at our office in central Cambridge.

Opportunity to work on a group research project with other fellows.

Networking and learning opportunities through various events, including trips to Oxford and London.

What we are looking for

We are excited to support a wide range of research, from the purely technical to the philosophical, as long as there is direct relevance to mitigating existential risk. This could also include social science or policy projects focusing on implementing existential risk mitigation strategies.

Incredibly successful projects would slightly reduce the likelihood that human civilization will permanently collapse, that humans will go extinct, or that the future potential of humanity will be permanently reduced. A secondary goal of this project is for fellows to learn more about working on existential risk mitigation, develop relevant skills, and test their fit for further research or work in this field.

Who we are looking for

Anyone can apply to the fellowship, though we expect it to be most useful to students (from undergraduates to postgraduates) and early-career individuals looking to test their fit for existential risk research. We particularly encourage undergraduates to apply, to develop their research experience.

We are looking to support proactive individuals from a wide range of subject areas who have a high potential to do impactful work in the future. Candidates will be assessed both on ability to contribute to the existential risk research community and motivation to reduce existential risks.

The application process

The first stage consists of essay-style questions, and is the main portion of the application process. We will evaluate applications on a rolling basis and applicants who progress to the next stage will be invited for a short interview, which is the final stage. Successful applicants will be notified by mid/late April, and afterwards we will work with accepted fellows to develop their project idea...

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: We're announcing a $100,000 blog prize, published by Nick Whitaker on March 7, 2022 on The Effective Altruism Forum.

We want to encourage a broader, public conversation around effective altruism and longtermism. To that end, we’re offering up to 5 awards of $100,000 each for the best new and recent blogs. We’re also making grants to promising young writers in the community.

You can learn more about the project and get on our radar here.

Why does this matter?

Top-of-funnel community growth in EA is slower than it could and should be. At the same time, EA is relatively underrepresented in intellectual discourse compared to newer and smaller movements like Progress Studies. EA is producing a ton of thoughtful writing, but the majority takes place in internal discussions and private documents. For some discussions, this would be the only sensible way to have them. But having other discussions in public should help to raise the salience of EA in the broader discourse and bring more people in. It could also help spark new ideas.

Further, we think EA needs more strong writers who can share key ideas in prestigious and popular venues — to persuade people to work on the most pressing issues of our time and to advance our thinking about them. We want to incentivize EAs to develop those skills.

What’s the plan?

We want to jumpstart these ambitions with the Blog Prize. Over the course of 2022, we want to find the very best new blogs exploring themes related to effective altruism and longtermism. Up to 5 winning bloggers will receive a prize of $100,000 each. (We were inspired by Tyler Cowen’s “Liberalism 2.0” blog prize). You can read more about our rules and guidelines on our website.

The judging panel for the blog prize is me (Nick Whitaker), Leopold Aschenbrenner, Avital Balwit, and Fin Moorhouse. Most blogs will be considered via our self-nomination form, but please feel free to send us recommendations.

What next?

We hope this can be a first step towards a more ambitious effort to support an ecosystem of public-facing writing for EA and longtermism. We believe that EA blogs could soon make up a major part of the general blogosphere, finding audiences (and potential EAs) we wouldn’t have found otherwise. Hopefully, we will also inspire the writing of foundational blog posts and posts that evolve into great projects.

To help potential bloggers, we’ve compiled a “How to start a blog” guide. We’re also offering mentorship on writing and editorial strategy to bloggers amid our private Slack community of bloggers. Self-nominate your blog on our website if you’d like to join. We have fostered a lively community for discussion, cross-promotion, and peer-to-peer feedback. Eventually, we hope to offer bloggers seminars from established EA writers.

While this is our first big announcement, stay tuned for future plans and follow ongoing efforts from Effective Ideas to build and foster a written media ecosystem for EA. Watch this space.

This project is supported by FTX Future Fund and Longview Philanthropy.

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

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: How are you keeping it together?, published by rando2022 on March 6, 2022 on The Effective Altruism Forum.

Not sure this is the best place to post this, but I think it's a problem a lot of us are having now: Given the recent news, I've been struggling a lot with hopelessness about the possibility of a catastrophic nuclear war. It's become a major impediment to both my mental health and my ability to care about EA causes. I've been thinking of things my family could realistically do in the event that things appear to be moving southward (packing bags and having a plan to getting out of the area if we have enough warning, and planning on ways to end things quickly if we don't). But I'm still pretty horrified at it all and finding it hard to function.

Any advice? How do you keep enough hope to keep at it?

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

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: Projecting compute trends in Machine Learning, published by Tamay Besiroglu on March 7, 2022 on The AI Alignment Forum.

Summary

Using our dataset of milestone Machine Learning models, and our recent analysis of compute trends in ML, we project forward 70 years worth of trends in the amount of compute used to train Machine Learning models. Our simulations account for (a) uncertainty in estimates of the growth rates in compute usage during the Deep Learning (DL)-era and Pre-DL era, and (b) uncertainty over the ‘reversion date’, i.e. the date when the current DL-era compute trend (with a ~6 month doubling time) will end and revert to the historically more common trend associated with Moore’s law. Assuming a reversion date of between 8 to 18 years, and without accounting for algorithmic progress, our projections suggest that the median of Cotra 2020’s biological anchors may be surpassed around August 2046 [95% CI: Jun 2039, Jul 2060].

This suggests that historical rates of compute scaling, if sustained briefly (relative to how long these trends have been around so far), could result in the emergence of transformative models.

Our work can be replicated using this Colab notebook.

Note: we present projections, not predictions. Our post answers the question of: “What would historical trends over the past 70 years when naively extrapolated forward imply about the future of ML compute?” It does not answer the question: “What should our all-things-considered best guess be about how much compute we should expect will be used in future ML experiments?”

Introduction

Recently, we put together a dataset of over a hundred milestone Machine Learning models, spanning from 1952 to today, annotated with the compute required to train them. Using this data, we produce simple projections of the amount of compute that might be used to train future ML systems.

The question of how much compute we might have available to train ML systems has received some attention in the past, most notably in Cotra’s Biological Anchors report. Cotra’s report investigates TAI timelines by analyzing: (i) the training compute required for the final training run of a transformative model (using biological anchors), and (ii) the amount of effective compute available at year Y. This article replaces (ii) the compute estimate by projecting 70 years worth of trends in the amount of compute used to train Machine Learning models.

Cotra's amount of effective compute available at year Y is broken down into forecasts of (a) compute cost, (b) compute spending, and (c) algorithimic progress. By contrast, we do not decompose the estimate, and rather project it on our previous investigation of training compute of ML milestone systems. This trend includes the willingness to spend over time including the reduced compute costs over time; however, it does not address algorithmic progress. We explicitly do not forecast the cost of compute or compute spending.

In this post, we present projections based on previously observed trends and some basic insights about how long the current 6-month doubling time can be sustained. That is, our post answers the question of: what would current trends imply about the future if you naively extrapolate them forwards.

One key reason we don’t expect these projections to be particularly good predictions is that it seems likely that Moore’s law might break down in some important way over the next few decades. We therefore might expect that that the doubling-time in compute usage, when the dollar-budgets to scale compute grow at the economic growth-rate, will be substantially longer than the historically common ~20-month doubling period.

When will the current scaling trend revert back to Moore’s law?

In our recent analysis of compute trends in ML (Sevilla et al., 2022), we find that, since the advent o...

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: CE Research Report: Road Traffic Safety, published by vicky cox on March 7, 2022 on The Effective Altruism Forum.

This report is the result of Charity Entrepreneurship’s research to identify the most promising charities to start in the area of global health and development. Improving road traffic safety is one of our top ideas for 2022. A new organization operating in this space would focus heavily on policy and advocacy work in the countries where these issues are most neglected.

We are currently accepting applications to our 2022 Incubation Program, through which we hope to launch this charity along with others that will focus on aid quality advocacy, postpartum family planning, tobacco taxation, and EA Meta. The 2022 Incubation Program will take place from late June to late August. The deadline to apply is March 31, 2022. We provide a 2-month, full-time, cost-covered training program (online), cash stipends to cover living expenses, considerable funding (of up to $175,000), and a wide range of ongoing support. Visit our website to learn more.

The CE team is happy to answer any questions you may have about the following research report, and we welcome critical feedback and conversation.

Health and Development Policy – Road Traffic Safety

Author: Vicky CoxReview: Sam Hilton and Melanie BasnakDate of publication: January 2022Research period: 2021

Thanks to Zakariyau Yusuf for his assistance with the early research for this report, and to Antonia Personette and Urszula Zarosa for their contributions. We are also grateful to the experts who took the time to offer their thoughts on this research.

For questions about the content of this research, please contact Vicky Cox at [email protected]. For questions about the research process, please contact Sam Hilton at [email protected]

Executive summary

Legislation is a commonly used tool to reduce road traffic injuries and fatalities in high-income countries. However, legislation is often lacking in low- and middle-income countries and, as a result, these countries disproportionately bear the global burden of road-related injuries and deaths.

We evaluated legislation across 5 key risk areas – speeding, drink-driving, helmet use, seat belt use, and child restraint use – and found that speed limits and seat belt legislation look to be the most promising . Therefore, this report will focus on speeding, speed limits, and seat belt use (both improving existing seat belt legislation to ensure it extends to all occupants and introducing new legislation in countries where it is currently non-existent).

There is strong evidence that speed is a key risk factor in road traffic injuries and fatalities, and that speed limits can reduce speeding. A meta-analysis of 98 studies explored the relationship between average speed and the incidence of road-related injuries and fatalities and created a formula that can be used to calculate the expected reduction in injuries/fatalities as a result of a change in speed (as the expected effectiveness depends on the speed before and after). For example, when speed is reduced from 100 km/h to 90 km/h, the number of fatal accidents is reduced by 34.4%. A meta-analysis of 51 studies found that speed limits reduce average speeds by 25% of the change in speed limit .

There is also strong evidence that seat belt use reduces road-related injuries and deaths. A meta-analysis of 30 studies found that wearing a seat belt can reduce fatalities among front-seat occupants by 40-50% and among rear-seat occupants by up to 25%.

Moreover, there is good evidence that advocacy leads to policy change. Looking at 84 case studies of advocacy in this space across the globe, we found an average success rate of ~48%, with 40 of 84 campaigns resulting in policy change in an average of 2.6 years.

Ma...