So it appears that the European baby formula that has been de facto banned in the US is consistently regulated to higher standards than US formula. For example, US formula can use corn syrup instead of lactose and need not contain omega 3 fatty acids.
@SenBlumenthal
What part of “Congress shall make no law abridging the freedom of speech” seems ambiguous to you? How much clearer does it need to be? Do you think it doesn’t count unless it says “and we really mean it” at the end or something?
For those who are unaware: the FDIC is brutally efficient, vastly more than almost any other government agency I can name. On Monday morning, the former Silicon Valley Bank branches will re-open under the name of another bank.
From what I can tell, the real reason European formula can’t be sold in the US is that US manufacturers have succeeded in getting regulations put in place for the true purpose of preventing European competition.
The regulations seem on their surface to be for the protection of consumers, but in reality exist to mostly to keep out competitors at the behest of the lobbyists of the US manufacturers. This is so common a situation economists have a name for it: regulatory capture.
The situation could be fixed with the stroke of a pen, but instead of simply suspending the regulations preventing European baby formula from being sold here, the government has focused its attention on pretending that the problem is elsewhere.
I don't get the anti-vaxers. Sure, the architecture was pretty bloated and slow compared to modern RISC instruction sets, but it was a fine machine for its time, and I remember it fondly.
Perhaps this situation will be fixed quickly enough that babies will not go hungry or starve. Perhaps it won’t. Either way, though, this isn’t a natural event, it’s man-made, and it’s not even isolated. Regulatory capture is common.
@runasand
@nytimes
Here’s a dirty open secret: HR isn’t ever going to help you (or any employee). They exist purely to keep the company out of legal trouble. They should always be viewed with suspicion.
there are additional features here, of course, including US government policy in purchasing formula for the WIC program, the way that formula manufacturing is regulated in the US, etc. But generally, the situation isn’t an unfortunate accident.
They do this all the time. They have a lot of practice at it. And, whether one likes the government deposit insurance system or not, they are very efficient at avoiding disruption to customers.
The FDIC always closes a bank on a Friday, and always transitions it to the new ownership over a weekend, and it always reopens Monday. Payroll systems, electronic transfers, or anything else in-flight will simply function as before.
Apple has won enormous amounts of goodwill by declaring that privacy is a human right, and is about to destroy all of it at once by building a technology to have your phone scan your pictures and turn you over to law enforcement if they’re the wrong sort of pictures. 1/
Almost certainly all of the accounts will be exactly where they were at the moment the bank closed. It is extremely rare in the United States that a single depositor loses money, even if they are over the $250,000 limit.
1. The rumors for GPT-4 are that it's as big a leap over GPT-3 as GPT-3 was over GPT-2 and that it will be arriving within months.
2. Even though this is an incredible inflection point in the history of civilization, to most people (1) might as well be gibberish.
Today's news about the Marriott breach should finally drive home a lesson that has been missed for years now: "we've been doing what every other big company does" means you are insecure and have to change your ways, because the median large company has terrible security. 1/
Why do I spend time refuting Eliezer Yudkowsky’s claims? Because he is on a constant PR blitz demanding society do foolish and even monstrous things in response to a threat that doesn’t exist, and because some people have come to believe him.
It’s tempting to think of humans as sentient, but their brains are just a bunch of nerve cells mechanically firing according to the laws of physics. There’s no there there. To really have a soul you need the purity and transcendence of matrix multiplication.
Eliezer and his acolytes believe it’s inevitable AIs will go “foom” without warning, meaning, one day you build an AGI and hours or days later the thing has recursively self improved into godlike intelligence and then eats the world. Is this realistic?
🧵Many people love the phrase “yelling fire in a crowded theater” to explain that not all speech should be protected. Few know the phrase was coined in a court opinion in which the Supreme Court held that public opposition to the draft was not protected speech. Really.
In the 1950s-1960s, Hollywood discovered Westerns made lots of money, and so they kept making Westerns well past the point where the market was burned out on them and didn't want to see them again for a generation. These days, it's superhero movies.
I finally realized one of the things that bugs me about most security "certifications" out there. Computer security is warfare. No, really, it's war. There's an opponent who doesn't care about you, doesn't play by the rules, and wants to screw you as fully as possible. 1/
Sometimes I hear people ask why
@SpaceX
doesn’t just get everything right the first time on a program like Starship. When I see that, I know they’re either not an engineer, or they’re unused to doing pioneering work on a hard problem at high speed.
It’s possible, on a crazy
@TalBenM
The FDIC has a strong incentive to prevent bank runs. If there is an expectation that depositors need to withdraw money to avoid losses, the FDIC would have to pick up the tab more often.
“AI can’t even play chess.”
“It can play chess, but it can’t play Go.”
“It can play Go, but it can’t paint a painting.”
“It can paint a painting, but it can’t write a poem.”
“It can write a poem, but it can’t really write an essay I subjectively see as sufficiently original.”
One of Eliezer Yudkowsky's essays, which his acolytes have frequently pushed me towards, is called "The Hidden Complexity of Wishes". In it, he describes an unnecessarily complicated thought experiment that can be summarized in only a few words without real loss of content:
Vernor Vinge has died. In pace requiescat. I only met him once, many years ago, though I recall we had a long and interesting conversation. Vinge saw farther and earlier. His influence, though quiet, cannot be understated.
Ever wonder why it’s harder and harder to get an appointment with a doctor? It’s because some folks decided decided to create a shortage. (I don’t know why this started making the rounds again but it’s still very relevant.)
I cannot tell you on what date deeply superhuman AGI systems will appear, but when they do, I can guarantee that a considerable fraction of the chattering classes will dismiss them as trickery or say “we don’t even have a good definition of intelligence.”
Listen to this man speak. He gets it.
Every call to stop AI is a call to keep this man from having a beautiful future. Every call to stop AI is a call to hold off cures to diseases, to hold off cheap energy sources, to keep mankind poor.
It doesn’t matter what sort of pictures motivated this feature; eventually governments will force its use for all sorts of things, and many governments do not respect human rights. I’m completely aghast that this is being contemplated. 2/
The crux of Eliezer's claims is this: it is necessary to get AI right "the first time" because once we build an AI that is smart enough we're all doomed when (not if) it decides to kill all of us. My strong claim is that this makes no sense at any stage of the reasoning chain.
Many people are familiar with the notion of the hedonic treadmill. I recently realized certain groups are on an anti-hedonic treadmill, constantly seeking out new reasons to be miserable. They hunt for novel things to be angry about the way some other people seek out new foods.
@theRealAhirrao
@zachcoelius
That's the way it works. The FDIC closes banks many times a year. There are about 5000 banks in the US and another 5000 credit unions. Some pop on a regular basis. They have a very efficient system for managing this.
The real “mathematics of cognition” turned out to be the fact that if you have an arbitrarily complex function (like “is there a cat in this picture?”) that you can describe only with a ton of examples, that you can find a good close approximation to the function with gradient
It’s been known for years now from both direct and indirect evidence that Alzheimer’s may be caused by herpesviruses. They and their close relatives like EBV and CMV almost certainly also cause other horrible diseases like MS. It’s well past time to wipe them out.
Biggest thing to ever come out of my little group. Pls help spread this finding!
We found clean, CAUSAL evidence that the shingles vaccine prevents a good chunk of dementia cases. So, could a virus cause Alzheimer’s->YES!
Hear me out & see preprint:
🧵1/
Repeating something that needs repeating: not everyone who has been thinking hard about AI for a long time agrees with
@ESYudkowsky
's claim that we're all doomed. Some of us know and understand his arguments and yet do not agree with his conclusions.
Mira Murati reached out to me in 2022 for a one-hour zoom call. Sam Altman never essayed any such contact. Also, I don't think Murati has made any jokes about how funny it would be if the world ended. I'm tentatively 8.5% more cheerful about OpenAI going forward.
I cannot take people who claim to worry about carbon dioxide emissions but who simultaneously oppose nuclear energy seriously. The positions are incompatible, and holding both at once indicates that one has no real understanding of the underlying issues.
I'm fed up. Something has to be done about the public discourse on AI. The doomers are spending totally insane amounts of money spreading their message while pretending to be meek underdogs. Time to get organized, even if that requires I do some of the organizing.
I cannot imagine owning a phone that might decide on its own to send copies of all my photographs to the government for analysis if some piece of software I’m forced to run decides the photos are bad, but that’s precisely what is apparently being deployed. 3/
@sf_mills
This sounds wonderful. You have to be a pretty hateful person to dislike this. The idea that we should be against having children becoming smarter and more accomplished is literally saying it’s better for people to be stupider and to have fewer opportunities.
If the Google Gemini fiasco (and it is a fiasco) has taught us anything, it is that the risk we face is not a lack of understanding of how to “align” machines. We already seem to be developing tools for brainwashing AIs after all. It is that alignment itself is an inchoate and
It doesn’t even matter if some of the AI doom fantasies have a germ of truth.
Imagine it is 1800. A group of fanatics starts explaining that if engine technology continues to be developed, people will be able to build tanks, bombers, machine guns, metal military ships that no
@IntuitMachine
Because 99% of the time they have a franchise to print money that has a high regulatory moat preventing easy entry. If you own bank stocks as part of a diversified portfolio you probably win most of the time.
I've discovered some people aren't aware of why Rust is a really cool programming language. They think it's just a strongly typed compiled language of some sort, which is true, but it doesn't really explain why it is interesting. 1/
Eliezer is doing a bait and switch again by saying the underlying math is simple. The rocket equation is also simple and you can’t read it and build a Falcon 9. All he did was an example from a textbook. This doesn’t constitute actually understanding how the stuff works in depth.
Man, I'm probably not going to win this; the gatekeeping tactic is simple and effective exactly because the mundanes in the audience don't know and can't trust that there *isn't* some deep and macro-relevant arcane science that *you* don't know about, if you're not wearing the
Your bluetooth device (headset, 2FA fob, keyboard, whatever) turns out to be insecure. And not fixable. Third parties can force your key length down to one byte. What a clusterfuck.
“AI will replace search” reminds me of “better agriculture will mean we can feed horses more cheaply than before, making horses cheaper”, except now we all drive cars. Only the lack of foresight seems even more extreme.
Reminder: Effective Altruism associated organizations have spent hundreds of millions of dollars deliberately trying to stir up AI risk fears. It’s a giant astroturfing operation funded by wealthy cult members like convicted embezzler Sam Bankman-Fried.
Just a reminder: any magic algorithm you think will reliably detect AI generated content can be used as an oracle to train AI systems to avoid the magic detector.
We’re going to have to live with being skeptical of images and text we see, just as we always should have been.
I don’t think that current LLMs are conscious, but it feels pretty sociopathic to torture something for kicks if it pretty faithfully behaves like a distressed human.
Eliezer has written thousands of pages of such essays. "To really understand," it is claimed, "you have to follow the full argument in all its majesty." And when you read the things, the extraordinarily verbose and sloppy writing conceals the sloppy reasoning, to a point.
Reminder: the claim that AGI will kill everyone is based on a series of stacked premises that are all either completely incorrect or almost completely incorrect:
...that AGIs will necessarily have arbitrary / inexplicable goals unless perfectly "aligned";
...that they will be
Most are unaware of how much of space technology was theoretically anticipated and worked out by a handful of people decades before anyone actually tried to achieve orbit. Tsiolkovsky’s achievements, for example, are astounding and worth reading about.
@tszzl
It’s intellectually honest, but it is also a reflection of a profound fanaticism and lack of intellectual humility. It is the sort of belief system that leads to massive suffering because it holds that there is literally nothing more important than the central belief.
Every time you invoke “fire in a crowded theater”, what you’re doing is quoting an opinion that was renounced by its author, which said political speech could be censored, which isn’t current precedent, and which most modern people would be ashamed of supporting.
The EA / Doomers never struck me as particularly good thinkers on topics like AI, engineering, human society, etc., but the sheer depth of miscalculation and misunderstanding associated with the OpenAI / Sam Altman affair has been breathtaking.
I tweeted this paper on Alpha Zero earlier, but it deserves to be tweeted again. AI developments are happening very, very, very fast at this point. A general algorithm that solved Chess, Go, Shogi, etc. on its own with 24 hours of training is big news.
@RichardHanania
So the problem isn’t the “innocent until proven guilty” idea. It’s the “actual criminals are as worthy of our care as honest people” idea.
@AlexVeshev
@tszzl
As I’ve repeatedly noted, John von Neumann was convinced, based on what he thought was an airtight game theory argument, that a nuclear war was inevitable and a first strike against the Soviets was necessary.
I'm often asked by AI doomers to explain why AI will not kill all life on earth. I'd like to suggest that this reverses the burden of proof. I'd like to see a compact, well argued, and persuasive case made for why we should believe that in the first place.
So what happens when you actually ask a primitive AI, not even one that is as good as what we will have in a few years, what it would do in response to the request? Let's ask GPT-4! Looks like its first thought isn't "throw her out the window."
The worst part about the position of AI doomers: they make the odds of disaster larger, not smaller. If they get what they want, we could have a future where only totalitarian governments have access to AI and use it to create astonishing levels of repression worldwide.
The danger is not AI generated images. The danger is an academic publishing system in which we pretend to have peer review but don’t. Humans have been generating nonsense papers for a long time. Peer review is supposed to stop it; if it didn’t, that’s the problem.
@GeorgeWHerbert
@theRealAhirrao
@zachcoelius
SVB can't hold out for anything. As soon as the FDIC seizes them, the shareholders no longer have any say at all. They're wiped out. The FDIC hands the assets over to a bank of their choice that they negotiate a deal with.
Frankly, I do not want to live in a society where everyone divides up into two economies, two internets, two sets of app stores, all segregated by political tribe. That's the path we're going. That path leads to violence at an unprecedented scale.
He literally posted an essay today advocating that anyone who refuses to go along with a global moratorium on AI research should have their data centers subjected to aerial bombardment.
Please note: There seems to be a campaign to FAKE the story that AI alignment theorists advocate violence. Everyone remember: *WE* never say this, it is *THEM* who find it so useful to claim we do - who fill the air with talk of violence, for their own political benefit.
I think cryptographers working on blockchains have come up with loads of fascinating new ideas. However, the touts, conmen, religious fanatics, etc. involved in these things aren't really capable of having honest discussions about the flaws and strengths. 1/
@TalBenM
Furthermore, it appears that if the 10 years SVB owns are held to maturity there are no losses. A much larger bank would have no trouble with that, and gets to eat the shareholder equity too.
I'm sorry, but no matter what spin Eliezer Yudkowski's acolytes put on it, it's simply not normal to claim that it is reasonable and proportional to bomb data centers in nuclear armed foreign countries if they refuse to stop doing AI research.
Whether by accident or intent, the Apple Silicon unified memory architecture means high end Macs are now really amazing machines for running big AI models and doing AI research.
Gedankenexperiment: what will the software industry look like once you can sit down and say to an AI: "Write me a clone of Linux, except written in Rust. Provide formal proofs of correctness of all of the code." and a little while later you have exactly what you asked for.
One of Eliezer Yudkowsky's persistent claims is that it's nearly impossible to get an to AI understand your requests well enough that it's safe to ask for even a fairly mundane thing like two identical strawberries on a plate.
@matthew_d_green
When the paper was released on my mailing list, I said that Bitcoin couldn't work because no one would be silly enough to do this; eventually it would need to eat all the power in the world to keep running. Except it turns out that I was wrong, people are that silly.
@ID_AA_Carmack
@PalmerLuckey
I am pleased by these postings. They indicate that we are developing a better culture. All too often, lives have been crushed based on hearsay.
I get that it’s tempting to think that Noam Chomsky is thinking, but if you look closely, his responses often demonstrate a mechanical sameness. His programmers were pretty talented but they haven’t gotten the hang of fooling a skilled observer.
Remember that Eliezer has essentially no evidence for this position. His argument consists of a long series of stacked hypotheticals, each of which has far less than 100% probability, which he claims together somehow create 99.9% certainty we are all going to die.
Remember: The argument for AGI ruin is *never* that ruin happens down some weird special pathway that we can predict because we're amazing predictors.
The argument is *always* that ordinary normal roads converge on AGI ruin, and purported roads away are weird special hopium.
Some of you may be glad that I stopped most of my COVIDposting in the last couple of days. You may be wondering if I at last got sick of it. In fact, I simply got sick. Probably COVID-19, but there's no way to know because in spite of trying hard I couldn't get tested. 1/
Imagine pizza shop owners are having difficulty making money. The government hits on a great solution: set a minimum price for pizza of $200 a pie. The result isn’t that all the pizza store owners are saved, it’s that they go out of business.
I've said this before: the single easiest way you can improve the lives of the bulk of US citizens (and the bulk of Europeans too) is to eliminate restrictions on real estate development. Housing will become far cheaper, but that's just the most obvious benefit.
A link between various herpesviruses and dementia has been suspected for years. This evidence is consistent with the hypothesis. Ending Alzheimers has a net present value to society well north of $20T, probably in practice many times that.
Herpes is terrible, but did you know it might also be making people demented?
Studies of the effects of antivirals support this contention and show that we may be able to greatly reduce dementia rates with existing drugs.
Check out this Taiwanese result.
A friend points out to me that there's an extent to which modern AI doom discourse resembles the worries that the LHC at CERN was going to destroy the world by creating micro-black holes or triggering vacuum decay.
This is a bad one.
TL;DR you can send a packet of death to any Linux 4 kernel before recent fixes, and it will make the processor spin in an infinite loop.
There is a frequent mistaken belief, which shows up in all sorts of discussions (including the recent fracas about AI) that perfect scientific understanding precedes engineering. This is not true. It never has been.
It is not possible for a neural network that runs on (say) ten gigabytes of parameters to faithfully store many hundreds of terabytes of training set. Claims that this is possible are routinely made but risible on information theory grounds.
Moral panics about people asking LLMs for information about making drugs and then getting answers seem no better than the moral panics of thirty years ago about being able to find such information online on web pages, or the moral panics decades before that about publications
Fascinating story. A dispute between two groups of scientists ran for years because one had done simulations using buggy software that they didn't release with their paper. My conclusion: if the software isn't public, the paper shouldn't be published.