you're trying to say because a decoder maps n -> 2^n that it's comparable to a QC. lol. my friend you clearly don't understand interference and entanglement.
btw the circuits in quantum circuits clearly aren't just combinational since they evolve in time.
No I’m not, all I’m saying is that by OPs logic, I can claim to have a computer with only a part of it.
“It’s not very useful but it can make computations” is a very low bar to pass, and very basic discrete (classical) logic can clear that without problems.
pls note a difference between a 'logical qubit' and a 'physical qubit'.. currently they don't have even 1 logical qubit, and for quantum computer to be of any use it should have >10k logical qubits...
wut? different QEC produce differently sized logical qubits and there are absolutely machines with enough physical qubits to amount to a logical qubit:
one thing i wish for in a paper reader: hover (or click) on a reference and get a preview of the cite or fig. that way i don't lose my place in what i'm reading before landing on a cite that's just an authors self cite of their 14th paper on the topic.
You can middle click on a reference to quickly go to the reference. You don't lose your place in the document because you can press backspace to get back to where you were (you could also leave a right-click highlight before you click so when you come back you know exactly the line you were reading).
lol doesn't this create perverse incentives - i.e. you're incentivized to actually make Julia slower sine it'll lead to being able to charge more for compute :p
if you're wondering why lots of these differentiable pipelines are tasked with learning physics (and what that has to do with google) the answer is that this is "compute oriented development". by "compute oriented development" i mean that since google has access to unlimited compute they can use this compute to run physics kinematics solvers (ie pde solvers) that are then used to generate training data for RL models. what's the point of the RL model if the physics model already exists and gives you high fidelity simulations? well it's clearly an easy paper to write... but other than that, some people claim the RL models are faster than the physics solver. i guess that's true if you don't take into account the millions of hours of compute spent on the solvers themselves.
Good point. This is why robotics researchers do not take deep RL papers seriously unless they have some real world robotics results. I'm looking at you, people who only show mujoco results and claim their algorithm is useful for robotics.
Simulators are useful though for real world robotics. You can prototype your environment and algorithm, and also attempt sim2real transfer. For example, use the simulator to generate a lot of image data, and train image based controllers. Add enough domain randomization and maybe your controller trained on the simulator can transfer to real images.
(disclaimer: work on RL, have trained models for simulated tasks)
I'm fairly sure that people work on control because general algorithms for control would be very useful (e.g., robot that can skin a cat and drive a car by holding the steering wheel). Such a robot would exist in our 3d physical world, so simulations of of our 3d world are used for training. If this could be done with radically less compute, it would be.
sure but it doesn't hurt that you have infinite data too (i.e. the thing most other ML research is bound by). like you can't argue that it's not a very comfortable corner to be in wrt being able to publish.
>I think I'm missing what the problem with that is.
I'm complaining that publishing endless papers on your methods that are trained on endless amounts of synthetic data is more about paper churn than contributing something novel. like the person below says: no real control system uses an RL controller (e.g. boston dynamics uses only classical controls).
Oh I see. I would have guessed that that was because this way relatively new. If it really doesn't translate to anything real then I definitely get your point.
> Their discussions are largely of limitations, catalogues of failure. Their conclusions can be brutal.
pov: me writing my first published paper and being honest about the shortcomings and then having my advisor tell me that i should sell my results more enthusiastically.
I was described as "too scholarly" by my PhD advisor when writing up my first couple of papers. I'm in industry now.
People doing a PhD/post-doc need to understand that there is no escape from "sales". Lots of academically leaning people, including myself, set out with a goal of avoiding becoming a seller and wanting to work in a space where data does the persuading rather than smooth talk. Unfortunately no such place exists. Even in the "purest" science-focused workplaces, you still need to sell your ideas to managers or funding agencies. To transcend "regular researcher" and become widely respected in your field, having thousands of Twitter followers will be more helpful than a paper in Nature.
The way to sell a scientific paper should be with precision and honesty about what conclusions you get. Academics should have the expertise to reject dishonest descriptions of the paper's object. That not happening on practice shows a failure on the education institutions.
"Should be" is the operative part of your statement - I totally agree there, and that the salesmanship requirement indicates a systemic failure. I'd say it goes much further than just the educational institution being at fault - they are responsible for playing the game, but funding sources do nothing to drive them more towards academic integrity.
This even extends to the hardest problems in mathematics and science. Solve the Riemann Hypothesis for real? Good luck getting the community to actually take a look at your paper and award you for it. Us problem solvers like these "pure" fields because we idealize a world where the work indisputably speaks for itself. Unfortunately without the sales and connection bullshido, no one will ever look at your profound work to see that it is world changing. Yeah, fuck everything about this. That's why Coq and Lean and all the other formal systems are ideal. I solve a problem, the system validates it. Now listen Clay, give me the motherfucking million!
That's not my experience at all. To be fair, I've never solved the Riemann Hypothesis, but in my experience it is possible to find professional academic mathematicians who were willing to meet me (as an outsider) to discuss their work and my ideas in a serious way.
Obviously, you have to bear in mind that high-profile mathematicians do get contacted by cranks and they have to weigh up what's worth spending their time on, but I have no doubt that if I did chance upon a genuine proof of the Riemann Hypothesis, I'd be able to find a decently respected mathematician to look at it and help me to publish it in a form acceptable to the academic community.
On the other hand, I agree with your comment about formal proof systems, and I'd love to use them in my (pure mathematical) research, but I've found the usability isn't there for me yet.
I have mixed experience publishing in math and statistics. I've had some of the best and worst experiences there compared to more applied fields. There's just been much more variability.
The comments and feedback, handled by the right editor, have been some of the most thoughtful, logical, and rigorous I've had of all my papers. The sort that, even when I disagree, think are very good questions to ask of the submission. Those that catch errors, and those that improve the paper in new, thoughtful directions.
At the same time, my worst peer review experiences have also been in more "pure" math and statistics. When they're bad, they're bad. The worst corruption and obvious jealousy have been in publishing in these areas, so convoluted that it would almost take an entire blog post series to explain. Reviewers can get really caught up on missing the forest for the trees and not understand the applied utility of something, even if it's technically correct, if it's not part of standard procedure. Othertimes it's been obvious (in the sense that if I published the entire review publicly I'm confident public opinion would come to that conclusion) reviewers have been jealous, and have tanked submissions in journals by piling up small criticisms that have nothing to do with the primary theses being argued. Usually in these cases, the failure ultimately comes down to an editor not wanting to ruffle the feathers of someone prominent.
These things all happen in applied fields too, but it seems like there, there's a softening around the edges of sorts, so the problematic behaviors aren't as extreme or obvious, and people sort of expect that fuzziness, so everything is taken with a grain of salt. I think the rigorousness of things like pure math can cut both ways, in that when it's correct, and everyone is behaving rationally, and the process is working with integrity, it produces very rigorous, solid work at the end. However, when it's incorrectly applied, and people are behaving irrationally (these are humans, it is inevitable), and the process is tainted, it can stifle novel work, or lead to really misleading conclusions.
I suspect that, if it's not already happening, I think one of the next phases in documentation of the reproducibility crisis and academic problems is in the areas of math, stats, and computer science. These fields have a veneer of rigor which can be true, but can also lead to false assumptions about the process that produces results. I think you're already seeing this a bit on HN with posts about reproducibility of AI findings, and with things like Taleb's criticism of academic models (FWIW Taleb has his own problems, but I think as a public figure he's correct to point out these things), but there's a lot more going on behind the curtain.
Yes, certainly things like corruption, jealousy and ego games do exist in pure maths, but I just wanted to contest the other comment's assertion that pure mathematics is completely closed to outsiders, let alone ones who could resolve the Riemann Hypothesis.
Furthermore, I could believe that the above issues and, as you also mention, reproducibility, are worse in pure mathematics or theoretical CS than in other fields because so little is at stake. Which is not to say I think most maths papers are false, but just that they're not all _entirely_ true either. Their audience is so small and the readers/reviewers "know what they mean" anyway, most of the time. And others are just false and no-one noticed because they couldn't be bothered to check the details, because the result was inconsequential to their own work and the author was a nice enough chap...
Why is my impression of academics drifting from an image of intellectual curiosity, deep thinking, and excellent communication; to a scene from a bad movie where I'm supposed to like the car salesman.
Maybe it's good I didn't go back to grad school...
P.S. this article appeared in Nature, so I guess I should take it with the required amount of salt.
Because all the good ones are flocking away to industry or other less ego-intensive careers. In my field it is gross, and if you are not talking on twitter you are not "active". And you have the same kind of citation rings you could (and still see) in papers but now with retweets... It is all about making it look-like science. We are really deep into the cargo cult... And at many levels, the grants, the papers, the presentations...
Yes many do. And the students as well. And it really make groups, especially in conferences where it is another live communication media. The other thing is that then they start talking about their lives and food choices and other mundane things like any other. Give it a look.
>Krav Maga is the martial art that I've always thought of as actually useful for self-defense
i'm not big on martial arts or even mma but for some reason i've seen this argument in various places often enough that i've also seen the appropriate response: if that were true then all of the most successful mma fighters would be krav maga experts. in actuality not a single one practices krav maga. the proof is in the pudding.
I'm similarly not into MMA, but I wouldn't think it's a good system to determine real-world performance of a martial art. Most people aren't going to be pitted against high-level fighters in a ring; they'll be out on the street. It comes down to why you're learning the martial art, and to some degree, what your threat model is.
Krav Maga minus the parts that are illegal or irrelevant to UFC basically _is_ MMA, which is unsurprising since the whole concept of the method since its inception in the 1940s is to collect effective and intuitive techniques from any martial art you can find. (Sort of a _mix_ of different martial arts, one might say...)
There's at least one successful mixed martial artist with a Krav Maga background (Karolina Kowalkiewicz, UKM Expert level 2), but she's definitely an outlier rather than the rule. Krav Maga tries to be a no-rules self-defense method that can be trained in as little time as possible. If your aim is to be a high-level competitor in a specific sport, and you have the required time and athletic ability to train appropriately, then your best bet is to skip the middleman and train in that sport directly.
Maybe because MMA is a sport and not an actual life or death combat? I'm sure if they allowed things like throat jabs, eye stabs, crotch shots and immediate joint dislocation upon grapple we'd see a lot more krav maga techniques.
This is a popular argument but it doesn't work if you think about it for a while.
Put a good striker in a fight with someone who isn't a good striker. The skilled fighter would evade most of what the unskilled fighter threw at them. There isn't even much need for blocking and covering if most of what is coming at you isn't making significant contact in the first place. Now take out the rules. If the unskilled fighter couldn't land a punch on the skilled fighter before, why would you expect them to suddenly be able to land an extremely precise strike to a vulnerable area now? And why wouldn't the skilled fighter also be able to do the same but better since they still have more power and accuracy and the unskilled fighter still has weaker defence?
It's the same with grappling. It's true that a joint lock can be devastating and that's why we usually disallow most of them in combat sports, but to be applied reliably you need some control over the area around the joint. Put a good grappler in a fight with someone who isn't a good grappler. The good grappler is going to control the position better. Now take out the rules. If the unskilled fighter couldn't get a good position to control the skilled fighter before, why would you expect them to suddenly be able to apply an effective joint lock now? And why wouldn't the skilled fighter also be able to do the same but better since they still have a better position to work from and the unskilled fighter still has weaker defence?
MMA is a sport and an MMA match isn't like a "real" fight for all kinds of reasons but simple, efficient, effective techniques such as you see in MMA matches are a good foundation for "real" fighting too, even if you might choose to add other types of training as well for use outside the controlled environment.
A traditional martial artist who has seriously trained in sparring would also be able to deal with an unskilled fighter fairly easily.
The difficulty arises when the rules of a sport cause a fighter to ignore situations that can arise in an actual fight. For example, I train in Dog Brothers Martial Arts. Grappling is common and encouraged, but at the same time most fighters conceal training knives on their person and headbutts are common. That requires an adapted form of BJJ from that used in MMA. Overhooks become more valuable because they allow you to control a limb that is holding a weapon. When you are in the bottom of guard you need to keep your opponent away from you so that they can't headbutt you.
Yes, this is the sort of distinction I had in mind when I commented before that MMA is still a sport and for "real" fighting you might choose to add other types of training as well. The possible presence of weapons is one huge difference. The open environment where you don't have an octagon wall to limit your opponent's movement and a conveniently flat floor is another.
I'm having trouble following your reasoning. It seems predicated on the assumption that Krav makes you a less skilled/unskilled striker/grappler when compared to say MMA or others, but I don't see any reason why this would be the case?
That aside, the main thrust of your argument seems to be, "If you can't land a non-lethal blow, then there's little or no chance of your landing a lethal blow." I don't think anyone disagrees with that. I think the disconnect is that once again your argument comes down to, "Krav automatically makes you less skilled than _______ (insert style here) and thus unable/unlikely to land any attacks."
But, once again, I don't think that follows unless there's a provable weakness/deficiency with Krav itself and/or you really do have a BJJ black belt vs. a level 1 Krav student, in which case I don't think anyone is going to disagree about the outcome.
Anyone who practices Krav on a regular basis and learns the techniques well can become a skilled striker, grappler, etc., in which case they will have an advantage over a straight MMA fighter of the same level simply because the Krav practitioner will be trained in very harmful attacks that MMA disallows (i.e., groin shots, eye attacks, etc.).
It seems predicated on the assumption that Krav makes you a less skilled/unskilled striker/grappler when compared to say MMA or others, but I don't see any reason why this would be the case?
So where are all the successful Krav-trained strikers and grapplers in the MMA world? If the striking and grappling are as effective as other styles that have stood the test of time, it's remarkable that there are still few if any high-level MMA fighters with a Krav background.
Anyone who practices Krav on a regular basis and learns the techniques well can become a skilled striker, grappler, etc., in which case they will have an advantage over a straight MMA fighter of the same level simply because the Krav practitioner will be trained in very harmful attacks that MMA disallows (i.e., groin shots, eye attacks, etc.).
Leaving aside your questionable claim that groin shots and eye attacks are very harmful, all of this would only be true if the time the Krav fighter invested in training those other techniques was not at the expense of the basic striking and grappling skills.
However the big practical problem with techniques like joint locks and targeting soft spots isn't that they can't work, it's that training them to the level where they can work effectively often requires much more time to build up much more skill than just punching the other guy in the face as hard as you can or executing a basic takedown and hold on the ground. If you'd just spent that much time practising the simple, reliable striking and grappling techniques, you might get a lot more bang for your buck.
So where are all the successful Krav-trained strikers and grapplers in the MMA world? If the striking and grappling are as effective as other styles that have stood the test of time, it's remarkable that there are still few if any high-level MMA fighters with a Krav background.
Professional fighters aren't going to study a system where 50%+ of what they learn is disallowed in competitions. However, Krav _is_ used by/taught to various special forces and law enforcement groups. To me, this more than makes up for a lack of "professional fighters" using the discipline.
Leaving aside your questionable claim that groin shots and eye attacks are very harmful
Try going to any martial arts lesson while not wearing a cup and LMK what your thoughts are afterwards :)
all of this would only be true if the time the Krav fighter invested in training those other techniques was not at the expense of the basic striking and grappling skills.
However the big practical problem with techniques like joint locks and targeting soft spots isn't that they can't work, it's that training them to the level where they can work effectively often requires much more time to build up much more skill than just punching the other guy in the face as hard as you can or executing a basic takedown and hold on the ground. If you'd just spent that much time practising the simple, reliable striking and grappling techniques, you might get a lot more bang for your buck.
Your comment makes me wonder if either you've not actually seen proper Krav Maga or you've seen it as practiced by a McDojo type of place. I attend classes led by a Krav Maga Worldwide certified instructor. Each lesson begins with 10-15 minutes of combatives (basic kicks, palm strikes, hammer fists, etc.). It's the whole "practice 1 kick 1000x" thing. The lesson then focuses on basic skills like escape choke from the front, escape choke from the side, escape bear hug, etc. It's all very practical and focused on primal attacks/counterattacks ("punching the other guy in the face as hard as you can" like you said).
Joint locks and targeting soft spots are just tools that _can_ be used, but the mantra is "closest weapon, closest target." Hit where you can as hard as you can and don't stop until they stop. With only a couple of exceptions, no weapon disarms happen until your opponent has stopped fighting.
Bottom line is that Krav in its true form is reliable striking and grappling techniques combined with "OK let's practice how to respond in this specific scenario." It doesn't get fancy and tries not to employ techniques that require any fine motor skills.
Well if your Krav classes include live sparring (not point or light sparring) then it's going to be more effective than the classes I have observed. And that's kind of the issue with Krav, maybe you'll learn some useful techniques and maybe your classes involve sparring, but there's no guarantee. And if you do have sparring it just ends up looking like poor boxing and BJJ. Why not just learn the real thing?
At the end of the day, of course techniques matter, but it's far more important that your martial art incorporates LOTS of live sparring. It needs to become second nature to apply your techniques against live/resisting/struggling humans that are trying to apply their techniques on you. That is why wrestling/judo/BJJ/boxing/MT are so dominant in MMA, NOT because of the "rule-following" techniques they use.
If a guy who has spent his training learning "street" techniques that you can't actually spar with (eye/groin/throat attacks) had to actually fight someone who has spent his training applying "non-street" attacks live for hundreds of hours, it wouldn't even be a fair fight.
Krav is widely taught to armed/special forces and law enforcement, who absolutely need a style that works in the "real world." To me, this speaks volumes more about its practicality and real world efficacy than its use by tournament/competition fighters.
Additionally, Krav has no competitive/tournament side to it, unlike most other martial arts. According to my Krav instructor, this is because the minute you add this aspect to a style, you then need rules around what is and isn't allowed. This goes against the core philosophy of Krav, which is "win the fight and get back home alive by any means necessary." It's about survival, and when fighting for survival you can't impose rules around what is and isn't allowed.
Much of what Krav focuses on (i.e., elbows, knees, groin/face attacks, etc.) isn't allowed in competitive fighting, so why would a professional fighter spend their time learning a style that they can only use say 50% of? Much better to learn something that was created with tournaments/competition in mind, where you can use 100% of what you learn, which is what I think most (all?) of them do.
So I don't think the "very few/no professional fighters learn Krav so it's no good" argument holds up, because 1) professional fighters would be wasting a lot of time learning Krav if professional fighting is their goal and 2) its widely used by many organizations outside professional/tournament fighting who arguably have an even greater need than professional fighters do for an effective self-defense methodology.
Krav is widely taught to armed/special forces and law enforcement, who absolutely need a style that works in the "real world."
Law enforcement have very little need for street fighting. If things are getting physical, the job is usually to detain a suspect as safely as possible. It's unlikely that the rules of engagement will allow causing the suspect severe harm or death, and if a LEO really is in that kind of situation, their priority is probably going to be disengaging and either transitioning to a weapon or deploying some sort of protective and restraint gear as quickly as possible.
Additionally, Krav has no competitive/tournament side to it, unlike most other martial arts.
The trouble with all martial arts that don't have any focus on competitive training is that you can never be sure they actually work unless you're in the rare position of having to use them for real and finding out the hard way. If you have ever been in that position then you have my greatest sympathy because the result is usually horrible whether or not you "win".
My Krav instructor was a sheriff's deputy for 20+ years and teaches Krav to law enforcement. From what he's said, the curriculum for them is slanted toward weapon disarms, choke hold escapes, take downs, etc., so Krav can be and most definitely is taught to law enforcement.
I like other disciplines/styles and plan to study BJJ and Muay Thai later on. I just don't think it's valid to say no tournaments = untested. After all:
- US Military using Krav >= competitions/tournaments
- IDF using Krav >= competitions/tournaments
- Law Enforcement using Krav >= competitions/tournaments
>have to prove themselves superior to classical ones, at least on designed tasks.
i think calling them "designed tasks" is laundering what's going on. call a spade a spade - it's simulating completely random circuits. so what's the point?
lol isn't this just the most ironic claim? i'm not excited by augmented reality exactly because i'm not bored with life (i.e. unaugmented reality).