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 survey: 46% of Americans are concerned about extinction from AI; 69% support a six-month pause in AI development, published by Akash on April 5, 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.

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 Chasing AI Safety Press Coverage, published by RedStateBlueState on April 4, 2023 on The Effective Altruism Forum.

On March 31, I made a post about how I think the AI Safety community should try hard to keep the momentum going by seeking out as much press coverage as it can, since keeping media attention is really hard but the reward can be really large. The following day, my post proceeded to get hidden under a bunch of April Fools day posts. Great irony.

I think this point is extremely important and I'm scared that the AI Safety Community will not take full advantage of the present moment. So I've decided to write a longer post, both to bump the discussion back up and to elaborate on my thoughts.

Why AI Safety Media Coverage Is So Important

Media coverage of AI Safety is, in my mind, critical in the AI Safety mission. I have two reasons for thinking this.

The first is that we just need more people aware of AI Safety. Right now it's a fairly niche issue, both because AI as a whole hasn't gotten as much coverage as it deserves and because most people who have seen ChatGPT don't know anything about AI risk. You can't tackle an issue if nobody knows that it exists.

The second reason relies on a simple fact of human psychology: the more people hear about AI Safety, the more seriously people will take the issue. This seems to be true even if the coverage is purporting to debunk the issue (which as I will discuss later I think will be fairly rare) - a phenomenon called the illusory truth effect. I also think this effect will be especially strong for AI Safety. Right now, in EA-adjacent circles, the argument over AI Safety is mostly a war of vibes. There is very little object-level discussion - it's all just "these people are relying way too much on their obsession with tech/rationality" or "oh my god these really smart people think the world could end within my lifetime". The way we (AI Safety) win this war of vibes, which will hopefully bleed out beyond the EA-adjacent sphere, is just by giving people more exposure to our side.

(Personally I have been through this exact process, being on the skeptical side at first before gradually getting convinced simply by hearing respectable people concerned about it for rational reasons. It's really powerful!)

Who is our target audience for media coverage? In the previous post, I identified three groups:

Tech investors/philanthropists and potential future AI Safety researchers. The more these people take AI Risk seriously, the more funding there will be for new / expanded research groups and the more researchers will choose to go into AI Safety.

AI Capabilities people. Right now, people deploying AI capabilities - and even some of the people building them - have no idea of the risks involved. This has lead to dangerous actions like people giving ChatGPT access to Python's exec function and Microsoft researchers writing "Equipping LLMs with agency and intrinsic motivation is a fascinating and important direction for future work" in their paper. AI capabilities people taking AI Safety seriously will lead to fewer of these dangerous actions.

Political actors. Right now AI regulation is virtually non-existent and we need this to change. Even if you think regulation does nothing good but slow progress down, that would actually be remarkable progress in this case. Political types are also the most likely to read press coverage.

Note that press coverage is worth it even if few people from these three groups directly see it. Information and attitudes naturally flow throughout a society, which means that these three groups will get more exposure to the issue even without reading the relevant articles themselves. We just have to get the word out.

Why Maintaining Media Coverage Will Take A Lot Of Effort

The media cycle is brutal.

You work...

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 80,000 Hours Podcast on high-impact climate philanthropy, published by jackva on April 4, 2023 on The Effective Altruism Forum.

This is a linkpost for a new 80,000 hours episode focused on how to engage in climate from an effective altruist perspective.

The podcast lives here, including a selection of highlights as well as a full transcript and lots of additional links. Thanks to 80,000hours’ new feature rolled out on April 1st you can even listen to it!

My Twitter thread is here.

Rob and I are having a pretty wide-ranging conversation, here are the things we cover which I find most interesting for different audiences:

For EAs not usually engaged in climate:

(1) How ideas like mission hedging apply in climate given the expected curvature of climate damage (and expected climate damage, though we do not discuss this)

(2) How engaging in a crowded space like climate suggests that one should primarily think about improving overall societal response, rather than incrementally adding to it (vis-a-vis causes like AI safety where, at least until recently, EAs were the main funders / interested parties)

(3) How technological change is fundamentally the result of societal decisions and sustained public support and, as such, can be affected through philanthropy and advocacy.

For people thinking about climate more:

(1) The importance of thinking about a portfolio that is robust and hedgy rather than reliant on best-case assumptions.

(2) The problem with evaluating climate solutions based on their local-short term effects given that the most effective climate actions are often (usually?) those that have no impacts locally in the short-term.

(3) The way in which many prominent responses – such as focusing on short-term targets, on lifestyle changes, only on popular solutions, and on threshold targets (“1.5C or everything failed”) – have unintended negative consequences.

(4) How one might think about the importance of engaging in different regions.

(5) Interaction of climate with other causes, both near-termist (air pollution, energy poverty) and longtermist (climate is more important when disruptive ability is more dispersed, e.g. in the case of bio-risk concerns).

For people engaging with donors / being potential donors themselves:

(1) The way in which philanthropically funded advocacy can make a large difference, as this is something many (tech) donors do not intuitively understand. We go through this in quite some detail with the example of geothermal.

(2) The relative magnitudes of philanthropy, public funding etc. and how this should shape what to use climate philanthropy for, primarily.

(3) A description of several FP Climate Fund grants as well as the ongoing research that underpins this work.

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: LW Team is adjusting moderation policy, published by Raemon on April 4, 2023 on LessWrong.

Lots of new users have been joining LessWrong recently, who seem more filtered for "interest in discussing AI" than for being bought into any particular standards for rationalist discourse. I think there's been a shift in this direction over the past few years, but it's gotten much more extreme in the past few months.

So the LessWrong team is thinking through "what standards make sense for 'how people are expected to contribute on LessWrong'?" We'll likely be tightening up moderation standards, and laying out a clearer set of principles so those tightened standards make sense and feel fair.

In coming weeks we'll be thinking about those principles as we look over existing users, comments and posts and asking "are these contributions making LessWrong better?".

Hopefully within a week or two, we'll have a post that outlines our current thinking in more detail.

Generally, expect heavier moderation, especially for newer users.

Two particular changes that should be going live within the next day or so:

Users will need at least N karma in order to vote, where N is probably somewhere between 1 and 10.

Comments from new users won't display by default until they've been approved by a moderator.

Broader Context

LessWrong has always had a goal of being a well-kept garden. We have higher and more opinionated standards than most of the rest of the internet. In many cases we treat some issues as more "settled" than the rest of the internet, so that instead of endlessly rehashing the same questions we can move on to solving more difficult and interesting questions.

What this translates to in terms of moderation policy is a bit murky. We've been stepping up moderation over the past couple months and frequently run into issues like "it seems like this comment is missing some kind of 'LessWrong basics', but 'the basics' aren't well indexed and easy to reference." It's also not quite clear how to handle that from a moderation perspective.

I'm hoping to improve on "'the basics' are better indexed", but meanwhile it's just generally the case that if you participate on LessWrong, you are expected to have absorbed the set of principles in The Sequences (AKA Rationality A-Z).

In some cases you can get away without doing that while participating in local object level conversations, and pick up norms along the way. But if you're getting downvoted and you haven't read them, it's likely you're missing a lot of concepts or norms that are considered basic background reading on LessWrong. I recommend starting with the Sequences Highlights, and I'd also note that you don't need to read the Sequences in order, you can pick some random posts that seem fun and jump around based on your interest.

(Note: it's of course pretty important to be able to question all your basic assumptions. But I think doing that in a productive way requires actually understand why the current set of background assumptions are the way they are, and engaging with the object level reasoning)

There's also a straightforward question of quality. LessWrong deals with complicated questions. It's a place for making serious progress on those questions. One model I have of LessWrong is something like a university – there's a role for undergrads who are learning lots of stuff but aren't yet expected to be contributing to the cutting edge. There are grad students and professors who conduct novel research. But all of this is predicated on there being some barrier-to-entry. Not everyone gets accepted to any given university. You need some combination of intelligence, conscientiousness, etc to get accepted in the first place.

See this post by habryka for some more models of moderation.

Ideas we're considering, and questions we're trying to answer:

Wha...

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: Giant (In)scrutable Matrices: (Maybe) the Best of All Possible Worlds, published by 1a3orn on April 4, 2023 on LessWrong.

It has become common on LW to refer to "giant inscrutable matrices" as a problem with modern deep-learning systems.

To clarify: deep learning models are trained by creating giant blocks of random numbers -- blocks with dimensions like 4096 x 512 x 1024 -- and incrementally adjusting the values of these numbers with stochastic gradient descent (or some variant thereof). In raw form, these giant blocks of numbers are of course completely unintelligible. Many hold that the use of such giant SGD-trained blocks is why it is hard to understand or to control deep learning models, and therefore we should seek to make ML systems from other components.

There are several places where Yudkowsky or others state or strongly imply that because SGD-trained models with huge matrices are unintelligible, we should seek some more easily-interpretable paradigm.

I'm going to argue against that. I think that a better alternative is probably not possible; that the apparent inscrutability of these models actually has little-to-nothing to do with deep learning; and finally that this language -- particularly to the non-rationalist -- suggests unwarranted mystery.

0: It Is Probable That Generally Intelligent Systems Must be Connectionist

Imagine a universe in which it is impossible to build a generally intelligent system that is not massively connectionist. That is, imagine a world where the only way to get intelligence from atoms is to have a massive number of simple, uniform units connected to each other -- or something that is a functional equivalent of the same.

In such a world, all smart animals would have become smart by scaling up the number of such units that they have. The dominant evolutionary species might become intelligent by scaling up its head size, despite paying the evolutionary cost of making childbirth dangerous and painful by doing so. Flying species that could not afford this extra weight of scaling up skull volume might take another approach, shrinking their neurons to pack more of them into a given volume. Even animals far distant from the dominant species along the phylogenetic tree and in which the evolution of high levels of intelligence occurred entirely separately, would do so by scaling up their brains.

The dominant species, once it could make information-processing equipment, might try for many years to make some generally intelligent system without massive connectionism. They might scorn connectionism as brute force, or as lacking insight; thousands of PhDs and software engineers would spend time devising specialized systems of image classification, voice transcription, language translation, video analysis, natural language processing, and so on. But once they coded up connectionist software -- then in a handful of years, the prior systems built through hundreds of thousands of hours of effort would fall to simple systems that an undergrad could put together in his spare time. And connectionist systems would quickly vault out from the realm of such prior systems, to build things completely unprecedented to non-connectionist systems.

Of course, such a world would be indistinguishable from our world.

Is this proof that intelligence must be connectionist? Of course not. We still await a Newton who might build a detailed causal model of intelligence, which confirms or refutes the above.

But if the universal failure of nature and man to find non-connectionist forms of general intelligence does not move you, despite searching for millions of years and millions of man-hours -- well, you could be right, but I'd really like to see any predictions an alternate hypothesis makes.

1.a: Among Connectionist Systems That We Know To Be Possible, Synchronous Mat...

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: Excessive AI growth-rate yields little socio-economic benefit., published by Cleo Nardo on April 4, 2023 on The AI Alignment Forum.

Basic argument

In short

There exists an adaptive limit representing the rate at which new technology can disseminate from research labs to the economy.

Exceeding this limit yields no socio-economic benefit, in the general case.

Exceeding this limit imposes significant risks and costs.

AI growth-rate currently exceeds the adaptive limit.

Therefore, we should slow rate of AI growth-rate.

Is this argument sound?

(1) There exists an adaptive limit.

There exists an adaptive limit because the economy is a slow-moving complex adaptive system — prices are sticky, employees have contracts, and the NASDAQ closes at 8:00 PM. The economy reaches equilibrium at a timescale of years, not weeks.

Imagine if OpenAI had released ChatGPT-3.5 but hadn't released ChatGPT-4. The release of ChatGPT-3.5 would've had a profound impact on the economy. Here's the standard neo–Keynesian story

Firstly, people start using ChatGPT-3.5 to automate part of their jobs, increasing their productivity. Some employees use ChatGPT-3.5 to complete more tasks within the same working hours, whereas other employees use ChatGPT-3.5 to complete the same tasks within fewer working hours. In both cases, value is created.

Because employees are completing more tasks, the companies start shipping more products and services. Supply goes up, and quality goes up too. So investors and shareholders receive more money. Because employees take more leisure time, their jobs looks more attractive to job-seekers. Hence the supply for labour increases, and the wages decrease, and the demand for such labour, at the new price, increases. Because the labour costs have dropped, competing companies will lower prices to gain more market share. And new industries, which would've been unprofitable in the previous economy, are now profitable and viable. This leads to a real-terms wage increase.

After a few years, the economy reaches (or at least approaches) a new general equilibrium in which...

People have more money to spend.

The products and services are more abundant, cheaper, and of a higher quality.

People have more leisure to enjoy themselves.

This is what we call an "AI Summer Harvest".

Notice that the benefit of ChatGPT-3.5 dissipates slowly throughout the economy. Eventually, the economy approaches a new general equilibrium, but this is a non-instantaneous process.

(2) Exceeding this limit yields no socio-economic benefit.

Our story starts with ChatGPT-3.5 and ends with higher social welfare.

This is a common feature of technological progress — a new general-purpose technology will normally yield these three benefits:

People have more money to spend.

The products and services are more abundant, cheaper, and of a higher quality.

And people have more leisure to enjoy themselves.

This follows from standard neoclassical economics.

However, the neoclassical analysis only applies when the economy is near-equilibrium. When the economy is far-from-equilibrium the neoclassical analysis is agnostic, and we must turn to other macroeconomic methods.

According to mainstream macroeconomic theory, until the sticky prices are adjusted, fired employees are rehired, and investments are reallocated, the economy will not be in a better socio-economic situation.

(e.g. the sticky prices haven't yet been adjusted, fired employees haven't been rehired, investments haven't been reallocated, etc), then the economy will not (in general) be in a better socio-economic situation.

(3) Exceeding this limit imposes significant risks and costs.

This has been covered extensively elsewhere. I'm alluding to

Economic shocks (unemployment, inflation, inequality)

Misinformation, fraud, blackmail, other crimes.

Extinction-level risks.

(4)...

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 CEA’s Interim Managing Director, published by Ben West on April 4, 2023 on The Effective Altruism Forum.

A few weeks ago, Max Dalton stepped down as Executive Director of CEA. Recently, the EV US and UK boards approved my appointment as Interim Managing Director in his stead.

CEA accomplished a lot in 2022, and I’m honored to lead the team in this interim period.

My communication style is more lighthearted than Max’s was; my model is:

People don’t read announcements, and will engage longer if it’s funny.

It seems good to bring humor to the Forum, and generally make EA a more fun place to be.

Also, I would rather just be myself on the Forum. Maybe I will regret this if a journalist quotes me out of context, but I’d rather do the thing that seems right to me than the thing which seems best for (a seeming definition of) PR.

But obviously some people have different opinions, and if you just want to skip to the serious bit you can go here.

For everyone else: the remainder of this post goes over some of our highlights from the past year, as well as some suggestions I have for the future.

Some CEA Highlights by Team

Events Team

The Events Team’s core metric is the number of connections made. More information about this can be found here.

This metric has shown substantial growth, approximately doubling from 2021-2022. The team is to be congratulated for their hard work and success:

However, recent victories won by Shrimp Welfare Project and others have made clear that there is an important audience which is underserved by current EAGs:

I am therefore directing the events team to put EAG Bay Area 2024 actually in the San Francisco bay:

EAG Bay Area 2024 (artist’s rendition)

Online Team: EA Forum

Engagement with object-level posts (those not about the EA community) has approximately quintupled over the past two years:

Although weirdly engagement on posts about the EA community spiked in November and are only now going back down to normal levels:

This is baffling because we did not roll out any large features in mid-November. If you have any guesses about what might have occurred here, please let us know.

Groups Team

Our University Group Accelerator Program (UGAP) program has been growing rapidly:

And thousands of people have been through our virtual programs:

However, there is something important about meeting in person. Like many of you, I was disturbed to learn that CEA does not actually own a castle, despite the obvious community building benefits. I am therefore instructing Community Building Grants recipients that at least one third of grant expenditures must be castle-related. This relates to my AI safety strategy, which I hope to publish more about soon.

Executive Office

It’s a great mark for our transparency that almost all of the data in this post can be found publicly. However, many stakeholders do not have access to that webpage.

I am therefore instructing the executive office team to create a dashboard which can be accessed throughout the multiverse via correlated decision-making.

Communications Team

Perhaps surprisingly, recent polling data from Rethink Priorities indicates that most people still don’t know what EA is, those that do are positive towards it as a brand, overall affect scores haven't noticeably changed post FTX collapse, and only a few percent of respondents mentioned FTX when asked about EA open-ended. It seems like these results hold both in the general US population and amongst students at “elite universities”.

Since EA is about blindly following quantitative results, and the quantitative results indicate that the financial fraud scandals don’t affect our public image,... Nope, not even going to joke about that one. (Fraud is, famously, bad.)

Serious Part

My role

In all seriousness: this is a cruxy time for EA and I am honored to lead C...

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: Penalize Model Complexity Via Self-Distillation, published by research prime space on April 4, 2023 on The AI Alignment Forum.

When you self-distill a model (e.g. train a new model using predictions from your old model), the resulting model represents a less complex function. After many rounds of self-distillation, you essentially end up with a constant function. This paper makes the above more precise.

Anyway, if you apply multiple rounds of self-distillation to a model, it becomes less complex. So if the original model learned complex, power-seeking behaviors that doesn't help it do well on the training data, this behavior would likely go away after several rounds of self-distillation. Self-distillation allows you to essentially get the minimum complexity model that still does well on the test set. Thus, I think it's promising from an AI safety standpoint.

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: Giant (In)scrutable Matrices: (Maybe) the Best of All Possible Worlds, published by 1a3orn on April 4, 2023 on The AI Alignment Forum.

It has become common on LW to refer to "giant inscrutable matrices" as a problem with modern deep-learning systems.

To clarify: deep learning models are trained by creating giant blocks of random numbers -- blocks with dimensions like 4096 x 512 x 1024 -- and incrementally adjusting the values of these numbers with stochastic gradient descent (or some variant thereof). In raw form, these giant blocks of numbers are of course completely unintelligible. Many hold that the use of such giant SGD-trained blocks is why it is hard to understand or to control deep learning models, and therefore we should seek to make ML systems from other components.

There are several places where Yudkowsky or others state or strongly imply that because SGD-trained models with huge matrices are unintelligible, we should seek some more easily-interpretable paradigm.

I'm going to argue against that. I think that a better alternative is probably not possible; that the apparent inscrutability of these models actually has little-to-nothing to do with deep learning; and finally that this language -- particularly to the non-rationalist -- suggests unwarranted mystery.

0: It Is Probable That Generally Intelligent Systems Must be Connectionist

Imagine a universe in which it is impossible to build a generally intelligent system that is not massively connectionist. That is, imagine a world where the only way to get intelligence from atoms is to have a massive number of simple, uniform units connected to each other -- or something that is a functional equivalent of the same.

In such a world, all smart animals would have become smart by scaling up the number of such units that they have. The dominant evolutionary species might become intelligent by scaling up its head size, despite paying the evolutionary cost of making childbirth dangerous and painful by doing so. Flying species that could not afford this extra weight of scaling up skull volume might take another approach, shrinking their neurons to pack more of them into a given volume. Even animals far distant from the dominant species along the phylogenetic tree and in which the evolution of high levels of intelligence occurred entirely separately, would do so by scaling up their brains.

The dominant species, once it could make information-processing equipment, might try for many years to make some generally intelligent system without massive connectionism. They might scorn connectionism as brute force, or as lacking insight; thousands of PhDs and software engineers would spend time devising specialized systems of image classification, voice transcription, language translation, video analysis, natural language processing, and so on. But once they coded up connectionist software -- then in a handful of years, the prior systems built through hundreds of thousands of hours of effort would fall to simple systems that an undergrad could put together in his spare time. And connectionist systems would quickly vault out from the realm of such prior systems, to build things completely unprecedented to non-connectionist systems.

Of course, such a world would be indistinguishable from our world.

Is this proof that intelligence must be connectionist? Of course not. We still await a Newton who might build a detailed causal model of intelligence, which confirms or refutes the above.

But if the universal failure of nature and man to find non-connectionist forms of general intelligence does not move you, despite searching for millions of years and millions of man-hours -- well, you could be right, but I'd really like to see any predictions an alternate hypothesis makes.

1.a: Among Connectionist Systems That We Know To Be Possible, Sy...

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: Given the Restrict Act, Don’t Ban TikTok, published by Zvi on April 4, 2023 on LessWrong.

While others have already written similar posts, given my previous position it seems necessary for me to do so as well – the Restrict Act is a no good very bad bill, and having seen the bill I realize that I was wrong to support banning TikTok.

For a while I have been in favor of banning TikTok, if it does not divest its Chinese ownership and modify the software to stop collecting outside user data. TikTok is Chinese spyware. Letting it be installed on most phones, given the data it is collecting, is not something we can or should abide.

Then someone actually proposed a bill, S 686 or the Restrict Act. I was reminded why it is almost never a good idea to ban things. Rather than a narrow bill to allow the banning of TikTok, we got a bill that vastly expands government powers, a Patriot Act for the Internet.

This is why we cannot have nice things. These people clearly cannot be trusted to regulate such matters.

So I’m admitting I was wrong. TikTok is still Chinese spyware, it is still not a good use of your time, you personally should not use it.

If this is how things are, however, then that’s where this ends. Don’t ban TikTok.

TikTok Is Chinese Spyware

TikTok is Chinese Spyware. It can’t read your texts and emails directly, but it can do a lot of other things short of that. If we have the ability to ban TikTok without vastly destroying our civil liberties in general, we should ban TikTok.

Noah Smith agrees.

Spying is the most commonly cited reason for banning TikTok, because it’s the easiest to prove. Tiktok has admitted tracking journalists’ physical movements and sending the data to its Chinese parent company. But physical location is probably only the tip of the iceberg of the data TikTok can collect, which includes faceprints, voiceprints, browsing history, text messages, and pretty much anything you do on your phone. And as Ben Thompson wrote back in 2020, that information basically becomes the property of the Chinese Communist Party.

[From Ben]:

All Chinese Internet companies are compelled by the country’s National Intelligence Law to turn over any and all data that the government demands, and that power is not limited by China’s borders. Moreover, this requisition of data is not subject to warrants or courts, as is the case with U.S. government requests for data from Facebook or any other entity.If anything it would be a something of a surprise were it not[.]

What I do not understand is why Apple and Google haven’t taken care of this for us. If I was running an app store and I found out a popular app was collecting a lot of the data on everyone’s phone and sending it back, I’d have said “either you fix this in 72 hours or you don’t have an app in our store and we’re issuing a warning to everyone to delete it.” Somehow, that did not happen. It still is not happening. I don’t know why.

There are plenty of other places to post and get similar content. TikTok itself would be easy to spin off or sell to preserve its value. I don’t see any serious threat to free speech here.

It turns out that our government is coming around to the position that our kids all putting Chinese spyware on their phones, and watching whatever videos a Chinese company decides to show them on the basis of an opaque algorithm, might actually be a bad idea?

In response, TikTok is exploring how to make this work (Bloomberg link).

If ByteDance unloads TikTok, then this would be a pure win for all involved.

If TikTok is actually banned, in many ways that is far better. Americans are spending really quite a lot of time on the app, in ways I doubt are producing much value in any sense, so I’d love to shake that loose. Would there be fallout?

A plurality actually say they favor the ban, and young people mostly d...