Posts mostly crossposted from my substack.
Yair Halberstadt(Yair Halberstadt)
Karma: 2,027
Might this be an issue even for people without this gene? What is the risk that constantly producing low levels of ethanol can cause oral cancer for ordinary people?
I think this is a really interesting question since it seems like it should neatly split the “LLMs are just next token predictors” crows from the “LLMs actually display understanding” crowd.
If in order to make statements about chairs and tables an LLM builds a model of what a chair and a table actually are, and to answer questions about fgeyjajic and chandybsnx it builds a model of what they are, it should be able to notice that these models correspond. At the very least it should be surprising if it can’t do that.
If it can’t generalize beyond stuff in the training set, and doesn’t display any ‘true’ intelligence, then it would be surprising if it can translate between two languages where it’s never seen any examples of translation before.
Taking a thirder position:
The reason you don’t usually accept a bet which you can only accept at most once is that you have a 2/3s credence the coin is tails, but given that the coin is tails there’s a 50% chance the bet is fake (since you’ve already accepted it the first time you woke up). This halves your expected earnings, so the bet is net negative.
But this gives you a way to break the symmetry. You only take the bet if the room is blue, which means your expected takings are now = 1⁄3 * -$200 + 2⁄3 * $100 = $0. So you still don’t take the bet, but would if slightly better odds are available.
Taking a halfer position:
When you wake up and see the wall is blue, your credence that the coin was tails must remain %50 (by conservation of expected evidence—you knew the wall would be either red or blue before you opened your eyes). So why should you take the bet?
The correct answer is that there is a better strategy than always refusing the bet. Namely: choose either Red or Blue beforehand and bet Tails only when you see that the room is in this color. This way the Beauty bets 50% of time when the coin is Heads and every time when it’s Tails, which allows her to systematically win money at 2:3 odds.
You place $200 down, and receive $300 if the coin was indeed tails.
If the coin toss ends up heads, you have a 50% chance of losing $200 - expected utility is $-100.
If the coin toss is tails, you have a %100 chance of gaining $100 - expected utility is $100.
So you end up with expected 0 utility.
The point stands, but the odds have to be better than 2:3.
I don’t see why the colour makes a difference...
So I refuse the bet same as in regular sleeping beauty.
Could be, but South Korea and Taiwan are in similarly precarious situations and have abysmal fertility rates.
When you look at other countries which are constantly at, threatened by, or threatening war they also don’t necessarily look great: Ukraine, Russia, Iran, China, North Korea, Finland, etc.That’s not to say this isn’t a factor, but I think you’ll need to add in enough extra details to differentiate from other countries that it will be very difficult to prove one way or another.
Falling fertility explanations and Israel
The way the auditing works in the UK is as follows:
Students will be given an assignment, with a strict grading rubric. This grading rubric is open, and students are allowed to read it. The rubric will detail exactly what needs to be done to gain each mark. Interestingly, even students who read the rubric often fail to get these marks.
Teachers then grade the coursework against the rubric. Usually two from each school are randomly selected for review. If the external grader finds the marks more than 2 points off, all of the coursework will be remarked externally.
The biggest problem with this system is that experienced teachers will carefully go over the grading rubric with their students, and explain precisely what needs to be done to gain each mark. They will then read through drafts of the coursework, and point out which marks the student is failing to get it. When they mark the final coursework they will add exactly one point to the total.
Meanwhile less experienced teachers don’t actually understand what the marking rubric means. They will pattern match the students response to the examples in the rubric, and give their students a too high mark. It will then be regraded externally and the students will end up with a far lower grade than they had expected.
Thus much of the difference in grades between schools is explainable by the difference in teacher quality/experience. This is bad for courses which are mostly graded in coursework, but fortunately most academic subjects are 90% written exams.
I believe that the US is nearly unique in not having national assessments. Certainly in both the UK and Israel most exams with some impact on your future life are externally marked, and those few that are not are audited. From my perspective the US system seems batshit insane, I’d be interested in what a steelman of “have teachers arbitrarily grade the kids then use that to decide life outcomes” could be?
Another huge difference between the education system in the US and elsewhere is the undergraduate/postgraduate distinction. Pretty much everywhere else an undergraduate degree is focused in a specific field, and meant to teach you sufficiently well to immediately get a job in that field. When 3 years isn’t enough for that the length of the degree is increased by a year or 2 and you come out with a masters or a doctorate at the end. For example my wife took a 4 year course and now has a master’s in pharmacy, allowing her to work as a pharmacist. Friends took a 5 or 6 year course (depending on the university) and are not Doctors. Second degrees are pretty much only necessary if you want to go into academia or research.
Meanwhile in the US it seems that all an undergraduate degree means is you took enough courses in anything you want to get a certificate, and then have to go to a postgraduate course to actually learn stuff that’s relevant to your particular career. 8 years total seems like standard to become a doctor in the US, yet graduate doctors actually have a year or 2 less medical training than doctors in the UK. This seems like a total dead weight loss.
Newtonian mechanics is a bunch of maths statements. It doesn’t predict anything at all.
The students constructed a model of the world which used Newtonian mechanics for one part of the model. That models predictions fell flat on its head. They are right to reject the model.
But the model has many parts. If they’re going to reject the model, they should reject all parts of the model, not just pick on Newtonian mechanics. There’s no such thing as gravity, or pendulum, or geometry, or anything at all. They should start from scratch!
Except that’s obviously wrong. Clearly some parts of the model are correct and some parts of the model aren’t.
So we have here a large Bayesian update that the model as a whole is incorrect, and a small Bayesian update that each individual part of the model is incorrect. The next thing to do is to make successive changes to the different parts of the model, see what they predict, and make Bayesian updates accordingly.
They will soon fine that it they model the base of the Pendulum as unattached to the ground, they will predict what happened perfectly, and so will make a large Bayesian update in favour of that being the correct model. Fortunately it still has Newtonian mechanics as one of it’s constituent assumptions.
That’s clearly not true in a general sense. Here’s a pattern that points to a different sum:
1 + 2 + 3 + … = 1 + (1 + 1) + (1 + 1 + 1) + … = 1 + 1 + 1 + … = − 1⁄2
Now the problem is this pattern leads to a contradiction because it can equally prove any number you want. So we don’t choose to use it as a definition for an infinite sum.
So you need to do a bit more work here to define what you mean here.
In precisely the same sense that we can write 1 + 1⁄2 + 1⁄4 + … = 2, despite that no real-world process of “addition” involving infinitely many terms may be performed in a finite number of steps, we can write 1 + 2 + 3 + 4 + 5 + … = −1/12
I think this is overstating things (which is fair enough to make the point you’re making).
The first is simply a shorthand for “the limit of this sum is 2”, which is an extremely simple, general definition, which applies in almost all contexts, and matches up with what addition means in almost all contexts. It preserves far more of the properties of addition as well—it’s commutative, associative, etc. In most cases where you want to work with the sum of an infinite series, the correct value to use for this series is 2.
The second is a shorthand for something far more complex, which applies in a far more limited range of cases, and doesn’t preserve almost any of the properties we expect of addition. It’s not linear or stable. In most cases where you want to work with sums of infinite series, the correct sum for this series is infinity. Only very rarely would you want −1/12.
Submission: Turing
MMAcevado simulates a Turing Machine in his mind, itself running a lossily compressed simulation of base MMAcevado. The simulated mind runs at 1⁄10,000th speed, and MMAcevado routes all IO through to the simulated mind.
On the assumption we have self navigating drones that can detect the weakest point in a tank as soon as it gets live of site, and head straight towards it, we would presumably have developed the ability to detect such drones via cameras on the tank as soon as they have line of site.
Than all you need is a bunch of pretty weak guns on turrets mounted on the tank to shoot the drone as soon as they are detected.
Most of these pieces already exist—modern Merkavas have cameras with 360 degrees view, the software to detect a moving drone quickly from a camera is pretty trivial, hardest part is avoiding false positives, but that seems easier than navigating, software to control guns and track targets has existed for a long time.
I assume that mounting a m16 style gun on a turret with 360 by 180 degrees rotation, and sub second rotation to any position is fairly straightforward. Imagine a few of these mounted along the sides of a tank. Most of the time they’re lying flat for protection but can shoot a drone within a second of it becoming visible.
A drone moving at 70 km/h would take 5 seconds to cover the last hundred metres to a tank, plenty of time to shoot it down.
This is mostly proven technology—it’s basically what trophy does, just we can use cheaper bullets against unarmoured drones, and use the theorised AI advances to use cheap cameras instead of more complex solutions, and the ability to distinguish enemy targets that are less obviously projectiles.
To a mother drone located farther from the enemy at higher altitude, but not high enough to be engaged. Using laser or directional (phased array) RF.
If it has line of sight to the drones, then it has line of sight to the target, and can be engaged by them.
I feel like the goalposts keep changing. This is not what was described in the original post.
So a few questions:
How do these drones communicate? Low on the ground P2P communications will have awful range, as will most low energy communication systems. Are they so autonomous they don’t need to communicate at all?
What’s their range? Existing drones only fly for about 20 minutes, and at a speed of about 70 km/h. Their range is usually about 10 to 20 km. Flying low to the ground and having to navigate will imply much lower speeds, and less efficient flight, as will having to run a powerful GPU, and whatever communication system you end up using. They also have to carry a payload capable of destroying a tank. Unlike in Ukraine that requires getting past the ADS (e.g. trophy), so is going to be more sophisticated than a grenade.
Again, how are you actually destroying the tank? Firstly ADS systems are likely to be extremely effective against drones. Secondly tank armour is actually really really difficult to pierce. Drones are only effective because tanks have weak spots where it was considered to be too unlikely that an enemy could target, and it turns out that assumption was wrong. The next generation of tanks will likely not leave such weak spots, possibly by using lots of slat armour, requiring far more sophisticated—and heavier—solutions to destroy a tank using a drone.
Now for this to revolutionise warfare requires that your drone + payload can be mass produced cheaply, but everything above seriously cuts into that. You need sophisticated communication systems, battery, navigation systems, payload etc. if each unit costs 100,000 dollars instead of 1000 dollars, sending 70 to destroy 3 tanks is much less valuable a proposition.
When is talking about kinetic energy weapons it’s referring to armour piercing sabots, because that’s what needed to pierce tank defences. I don’t know how effective ADS would be against other kinetic energy weapons because there’s never been any need to try, they’re useless against tanks. These rounds are so heavy, and fly so fast, that’s it’s practically impossible to fire them from anything weighing less than a few tons. Not relevant for a drone/Javelin. Also notice how you’re creating epicycles upon epicycles here. A drone that fires a Javelin, that fires a railgun, to defeat an existing fairly straightforward defence. Each of those is going to be an impressive technical achievement, the entire package is going to take a while to iron out the kinks, and is going to be expensive. If drones are so powerful they’re going to completely replace existing armies, I wouldn’t expect all the epicycles.
A single javelin missile on its own costs more than an iron dome tamir interceptor, so becomes a valid target for existing SAM defences.
Sure it might be existing ADS defences aren’t enough to defend something like the phalanx, but there’s lots of implementations out there, and the trophies characteristics were chosen because it was sufficient to protect tanks. Could trophy be modified to protect more delicate equipment, or could something like the Iron Fist work? I don’t know, it’s never been tested because it’s never been necessary.
As for the drone you linked—it contains a turbojet. I cannot find any production turbojet with hundreds of kilos payload plus strong performance characteristics selling for less than a few hundred thousand dollars.
Finally I think all of this is mostly irrelevant. The phalanx consists of two parts—a relatively long range, delicate and expensive radar, and a pretty robust, shorter range, cheaper M61 Gatling gun + turret.
On a ship they’re colocated because that makes sense given limited space. But most land based SAMs separate the radar and missile launcher.
I expect that if drones ever become a serious threat will see the proliferation of lots of Gatling guns mounted on tanks and other vehicles, linked to a decentralised radar system combining lots of different radars of different specs. The radars will generally be deployed further behind the front line, (although some cheaper short range ones might be mounted on tanks) and will give targeting information to the guns scattered across the front line which will take out the drones.
The guns are much less vulnerable, and less expensive so don’t make good targets. The radars are very expensive but much further behind the front line, out of range of cheaper drones and well defended by both guns and missiles against more expensive solutions. And taking out a single radar just degrades performance, doesn’t take down the whole system.
This will be expensive and complex to develop but far quicker than your autonomous drone army, since all the pieces are already in place.
Finally you claim iron dome is out of reach of most countries, but most countries do have SAM systems of various sizes. Iron dome is unique in it’s ability to target SRBMs, and reflects the trade offs needed for that, but drones are much simpler to take down, and countries that deploy SAMs capable of taking down modern fighter jets could easily deploy ones capable of taking down drones. Tamir is just a good example since I know it’s cost and it’s not that expensive.
Like everything you have layers of defence. Phalanx takes out all drones in an area. Against any ATGMs you use trophy or an equivalent ADS.
Also once you have a javelin/anti armour carrying drone it’s going to set you back hundreds of thousands of dollars and be a suitable targets for iron dome style defences, which can cover a larger area and where each missile costs some 75000 dollars.
Why compare with a Javelin, and not e.g. a Kornet which exports for 25,000 dollars (similar to a top range GPU), and can be produced for much cheaper (as evidenced by the fact Hamas is perfectly capable of producing them).
As Gregor Samsa awoke one morning from uneasy dreams he found his wife transformed in his bed into a gigantic insect...
...but he concluded that since ontologically this insect was not his wife, his marriage vows no longer applied. He squashed it under his boot as he walked out.