I certainly feel like AI maximalism is the only way it all maths up. There is a bit that Matt Levine writes about AI companies, and you touch on basically the same points, that roughly goes "you are an AI company and you are making the god-computer. Why would you focus on anything else because if you're successful you could basically generate infinite money? Turns out to make the god-computer you need a lot of money upfront."[1]
So perhaps they would only respect money and property up to the point that it's necessary.
The more realist answer is that they will keep respecting money and property as capital works to reinforce the status quo and enrich shareholders. You think just because you can do your job faster--or completely--with a machine that you have to work less? Hardly! Now this is where interesting policy questions spring up but I digress.
This sounded familiar and then I realized I had read the same article a few weeks ago. This is some strange syndication I would propose changing the link to this one: https://biographic.com/death-by-a-thousand-dams/ it’s also way better layout and fewer ads.
In “Careless People”, in the chapter on the Rohingya Genocide, a case is made for Facebook’s complicity in it. Essentially threatening and violence-inciting content got past moderation because the font that was used didn’t support Myanmar text. So there would be a report on a post but all the mod team would see was missing Unicode characters and mark the post as okay. By the time the serious violence started it was too late.
Now, I am not claiming a lack of text support caused a genocide, and I don’t think Wynn-Williams is either. Facebook was certainly culpable though, the military had direct controls over what was allowed to be posted and suppressed dissent. The issue could have been caught much earlier on and perhaps mitigated had off-shore moderation teams been able to read posts that were being flagged.
This is from 2016, it's a fairly detailed study of what happens when a rural, barely (or illiterate) population gets access to a powerful social media tool:
If you read contemporary reporting about anti-Muslim riots in Myanmar https://www.reuters.com/article/world/special-report-buddhis... you may learn that there were Buddhist monks traveling the country, giving sermons to crowds of thousands of people, telling them to treat Muslims as enemies, and stickers being used to visibly signal alignment with the movement.
Careless People is a story about Facebook from inside Facebook, so of course it adopts the viewpoint that Facebook's moderation decisions are very important, but ultimately they were a sideshow to what was playing out on the ground. This goes double for military attacks on Rohingya villages, which can hardly be explained as spontaneous reactions to some Facebook post.
In any case, Facebook moderators literate in Burmese would've known to use the right font, whereas for Facebook moderators trying to deal with a language they can't even read, the font choice would be the least of their problems.
The internet and the real world don't exist in separate vacuums. Internet communication and amplification of hate speech helped create an environment where hateful sermons were planned and popularized, and vice-versa.
Facebook pages and posts espousing these viewpoints (using sensationalized or completely fictional justifications) were then amplified by the algorithm. The amplification reached people that were not otherwise inclined to participate, but were convinced through information they received online.
You can't have one without the other. Influence is driven through both in-person physical events/marketing and online platforms. Categorizing the online influence as a "sideshow" completely misinterprets how important online communities have become in radicalization. "Reach" is a key social media metric for a reason.
Reach is indeed important. But the internet is not the only way to generate reach, nor necessarily the most effective one. During the riots in 2012 and 2013, 2% of the population were using the internet https://data.worldbank.org/indicator/IT.NET.USER.ZS?location... and there were 14 mobile phone subscriptions per 100 population https://data.worldbank.org/indicator/IT.CEL.SETS.P2?location... Meanwhile, a traveling monk wanting to hold a sermon about the importance of shunning Muslims could rely on the local religious community to deliver him an audience, and on word of mouth, flyers and posters to inform everyone who didn't attend.
That's not the full story. A few quotes from Amnesty International's report:
> Even before internet access became widely available in Myanmar, Facebook was already the dominant platform in the country.
> In 2011, despite widespread restrictions on internet access, it was estimated that approximately 80% of Myanmar's few internet users had Facebook accounts.
> By 2014, Facebook was estimated to have had hundreds of thousands of users in Myanmar, many of whom were accessing the platform through internet cafes or devices belonging to their friends or family members.
> Prior to 2015, Facebook was only available in Myanmar via an English-language interface, but a specific Myanmar version was launched in 2015.
Hundreds of thousands of Facebook users in a country of 50 million works out to less than 2%. It's the same story. The Amnesty report does not strike me as starting from specific instances of mob violence and working backwards to determine the source of radicalization in each case. (And of course the later military campaign against Rohingya villages that they mostly focus on was not mob violence at all.) Instead, they work forward from the information leaked out of Facebook, which of course comes with a Facebook-centric worldview.
They would just run everything through machine translation and throw African labor at moderating it. There's no way they hire moderators locally. Or probably they just let LLM/Jev/Laya handle it nowadays.
And so, the standards of moderation will be increasingly more inconsistent and chaotic as cultural distances between the company HQ in US to the poster of a particular content increases - I see people complaining about food pics getting flagged as pornography in Japanese all the time.
I can't imagine how random the moderation infra will respond to: anti-Muslim propaganda briefs created by violent fundamentalist Buddhism terrorism groups circulated in the minority inhabited contested forested border areas of Myanmar. That description alone sounds Neuromancer enough to that the system might just fail open, before it digests the content and likewise fail open.
Good points. Also, thanks for the Reuters link. Buddhists as a group have for centuries been (fairly) associated with peace and nonviolence. The situation in Myanmar is a tragic anomaly. The article you shared sheds at least a bit of light on it:
> Relishing his extremist reputation, Wirathu describes himself as the "Burmese bin Laden."
Wirathu is clearly a sociopath -- and the furthest thing from a real Buddhist. If only people weren't so easily led by those who use religion as a means of control....
Facebook's direct culpability in the ethnic cleansing in Myanmar had already been well-known and thoroughly reported on many years before Careless People.
I heard it was because .. in a country where they did tie-ups with mobile operators to 0 rate face book traffic, to make it to the point where the internet and facebook were the same thing to many people, they also only employed a handful of moderators. Meanwhile Politicians were posting incitements to violence on the platform.
I think OP was more pointing out a supreme irony with the farmer demographic. They rely heavily on seasonal immigrant labor yet they voted for immigrants to be rounded up and forbidden to enter the US. Their sales rely on exports (e.g. soybeans) and imports (tractor parts, fertilizer) yet they voted for a complete upending of the global trade system that benefitted them. So that they are complaining about the current economic climate while simultaneously politically being fully behind it is schadenfreude for many.
Both political parties are made up of constituent cohorts. All Republicans are not anti-immigrant, anti-trade junkies in the same way all Democrats are not anti-work socialists. I think you'll find farmers in particular are generally much more pro-trade and pro-immigrant than the rest of the party.
The story is these farmers voted for Trump despite him tilting against their best interest. But the reality is they probably put up with his views on immigration and trade hoping they would be more than offset by his views on land use and tax policy (and have been sorely mistaken).
I know that people really want to make schadenfreude out of this But when the shoe is on the other foot and Republicans make these same kind of ad hominem attacks they equally fall flat. No, just because someone voted for Joe Biden doesn't make them a police-hating communist.
> these farmers voted for Trump despite him tilting against their best interest
There is a good amount of research that suggests people do not vote using reason, or necessarily consider their own best interests. Rather, it is suggested that most people who vote affiliate themselves with a political party that appeals to their interpretation of morality; and then they vote on tribal lines.
It's a little hard to feel bad for these people when they already saw how Trump didn't actually push their interests during his first term, and yet they voted for him again.
Meanwhile, the Project 2025 handbook was there for everyone to read, and Trump's rhetoric was based on cracking down on immigration (bye-bye, seasonal labor), beating up our trade partners (need to import farm tools? pay more!), and slashing federal government spending (haphazard and largely a failure to actually save money or tackle deficits or debt).
The 2017 Tax Cuts & Jobs Act was a fine preview for more of the same in 2025, so I'm not sure what these voters were looking for when it comes to tax policy.
Land use is generally a state-level concern, so I'm not sure what they think Trump would be doing for them there.
I could buy it if we heard about reluctant Trump voters who don’t really like him and disagree with a bunch of his actions but reluctantly voted for him because they thought he was better than the alternative. This seems to be extremely rare, though. Maybe those voters just don’t get air time.
Okay. So all that opposition turnout disappeared in 2024? Where did they go? Were they so disappointed with the alternative that they just decided to just stay home (effectively a vote for Trump)?
> All Republicans are not anti-immigrant, anti-trade junkies in the same way all Democrats are not anti-work socialists
Democrats are not anti-work socialists and never were. They never had rhetoric that would imply so. However, republicans do have anti-immigration, anti-trade rhetoric and politics.
> No, just because someone voted for Joe Biden doesn't make them a police-hating communist.
Of course, because Biden was not hating police and was not a communist. Republican party however is exactly what it is accused of being here. And Trump is its ruler and champion.
I think this is a great explanation. To add to it, I believe there is something durability too. My basic Jabra wireless earbuds from 2020 are just now getting faulty (just battery stuff). In a similar time period from 2010 and 2015 I went through at least 6 pairs of wired headphones or earbuds, the jacks would fail or the wires would just stop working.
I'm a little confused about the value of scanning rare books since this story came out. I have a small collection of "rare" books and they're not really bounties of information, at least not modern information. I know novel training corpus is important but the information in rare non-fiction books is commonplace or out-dated. And the information in old rare fiction-books are originals for which reprints exist or just uninteresting stories that aren't really worth anyone's time except collectors'.
I think from the manufacturer's perspective there is some misplaced idea that automation via an app somehow makes work easier. For household chores there are so many manual steps that an app doesn't actually make the process any easier, it complicates it. You have this dependency on a phone that you're relying on whereas before it was just some simple button presses. It's like having to call your mechanic every time you wanted to start the car.
> As a thought experiment, imagine only one person had a superintelligent lawyer. They would have an unfair advantage in court -- even if they were wrong on the merits. That would lead to a worse society. But now imagine everyone has a superintelligent lawyer. In this case, justice would be carried out much more fairly and efficiently than it is today when there is often an imbalance in skills and resources in litigation.
This would only be true in a vacuum. Ignoring several leaps of progress where it would suddenly be allowed for an AI to represent one in court (imagine the size of the context window) and you could also call it from prison. That is not what is imbalanced about the US court system. Whoever has more money can drown someone in court filings and court fees and then drop their case financially ruining someone.
It's also funny how the interest of "AI for everyone" only extends to the national interest but I guess he's got to make sure he stays in good graces with the current administration.
I'm not a fan of Zuckerberg in the least, but one area a super intelligent lawyer would be fine at is being drowned in court filings and paperwork
The bigger problem with his argument IMO is that even with open models, it's still the person or organisation with the most money / access to GPU compute winning. They run the larger model (or collection of models), they can process more tokens through them in the same amount of time etc
I think there's a point of diminishing returns here. A more competent lawyer can only do so much if you're wrong on the merits, assuming your opponent also has a decently competent lawyer.
> it's still the person or organisation with the most money / access to GPU compute winning.
Is it? Legal is rarely a genuine war of attrition and delays are the norm. AI is years/decades away from not needing humans in the loop verifying output, legal more than most. Correctness is more important than speed. Main skill is doc search but most cases easily fit in modern contexts.
So I don't think legal in particular needs the fastest/bestest compute.
Naively, you'd think the judicial systems would invoke "judgement" on the underlying purpose of dumping content in the form of filings, paperwork, etc...
I think that's what's ringing to me about the whole piece. There is a large assumption of equity. There is this section header:
> Everyone will have free or affordable access to these tools.
"access" is doing a lot of heavy lifting, yes everyone has access to the US healthcare system. Is it affordable for everyone? Not even close, but politicians will always phrase it that way as a non-agreement to proposals of universal healthcare.
This is the only valid future for tech like this IMO.
Should any kind of "superintelligence" exist, it must be a public good, available to all. Not controlled by a small handful of oligarchs and profit seeking entities.
Only in a world were AI companies can subsidize AI usage indefinitely. In reality AI is not cheap and larger & more potent model it's usage costs is more expensive
I think his hope is that a either a fully automated superintelligent AI lawyer or something close that won't be as costly to guide would be infinitely more affordable than the current human lawyers.
Throwback to a brief moment when everyone wanted to support IFTTT. If I remember, it ran out of steam because it takes away eyeballs from the actual application (social networks in-particular). The use cases are also rather niche.
So perhaps they would only respect money and property up to the point that it's necessary.
The more realist answer is that they will keep respecting money and property as capital works to reinforce the status quo and enrich shareholders. You think just because you can do your job faster--or completely--with a machine that you have to work less? Hardly! Now this is where interesting policy questions spring up but I digress.
https://archive.is/20251015175242/https://www.bloomberg.com/...
reply