HN user
I feel like the fact that ML has no good explanation why it works this well gives a lot of people room to invent their head-canon, usually from their field of expertise. I've seen this from exceptionally intelligent individuals too. If you only have a hammer...
Don't we all experience this from time to time? When I'm focused on solving some mathematical problems I'm not thinking in words, but in concepts. When you are thinking of words you also think of a concept, the only difference is that sometimes there are no words associated to it. Im my opinion, words, sentences are just a label to the thinking process, a translation of what is really going on inside, not the driver of it.
This exists and does work to some degree, e.g. Detecting hallucinations in large language models using semantic entropy https://www.nature.com/articles/s41586-024-07421-0
My experience is mostly with gpt-4. Act like it is a beginner programmer. Give it small, self-contained tasks, explain the possible problems, limitation of the environment you are working with, possible hurdles, suggest api functions or language features to use (it really likes to forget there is a specific function that does half of what you need instead of having to staple multiple ones together). Try it for different tasks, you will get a feel what it excels in and what it won't be able to solve. If it doesn't give good answer after 2 or 3 attempts, just write it yourself and move on, giving feedback barely works in my experience.
There were a lot of people who were just reciting the best practice rules they'd learned from blog posts, without really having the experience to know where the advice was coming from, or how best to apply it
This is exactly my experience too. Also, the problem with learning things from youtube and blogs is that whatever the author decides to cover is what we end up knowing, but they never intended to give a comprehensive lecture about these topics. The result is people who dogmatically apply some principles and entirely ignore others - neither of those really work. (I'm also guilty of this in ML topics.)
WPA2 also had an exploit (KRACK) while the handshake algorithm itself was "proven to be secure". Formal verification is a powerful tool but it does not guarantee bug-free code: it merely guarantees that the particular bugs you checked for are not possible.
Based on response times it probably just uses caching for some answers and has limited token/second toward chatgpt so it gives an error sometimes if you ask a question that needs to be forwarded.
I'm running models locally on my 3090 and it's fast enough, although for example building a vector database can take a while. I can run LoRa training but I haven't done anything meaningful with it so far. I chose 3090 because of the cable issue of 4090 (also, no nvlink, although I'm not sure that matters) but it's debatable if my fears are justified. I need to leave the gpu running while I'm away and I just don't feel comfortable doing that with a 4090. I rather take the lower performance.
One caveat though, my asus b650e-f is barely supported by the currently used ubuntu kernel (e.g. my microphone doesn't work, before upgrading kernel + bios I didn't have lan connection...) so expect some problems if you want to use a relatively new gaming setup for linux.
One possible explanation is that Putin wants to prevent a possible rival from winning over oligarchs by promising to make agreement with the eu. Many oligarchs are losing a lot of money on this war and it is likely that a coup that replaces him with a more friendly leader could ease up the sanctions. Without the gas leverage this option is less likely. There are clear signs that Putin is very paranoid of being replaced in the near future.
It would even decrease the value of their property to no fault of their own. That sounds like something the manufacturer should be responsible for.
The entangled photons can't detect the interruption alone, you need the message too. If you are interested how I suggest watching Veritasium's video about Bell's theorem as a start. The cryptographic method is called Ekert protocol and it uses Bell's theorem to detect if the quantum spin was measured in an incorrect "direction" (called base). Just write down the expected outcome in 60 degrees tilted base and you will see how it changes compared to being measured in the correct direction. Add enough photons and possible directions and you can detect eavesdropping with a very high certainty.
Quantum entanglement cannot be used to transmit information but can be used to check if the message was interrupted, so it's still important for quantum cryptography. (I'm just a programmer but had a course quite some time ago on quantum computing thought by a physics professor.)
Learning algorithms are also affected by experience (or data in case of supervised)
While the framework of differentiation between intranational and international interests fits, it is certainly not "for the good of the people" in this case. Orban antagonized the EU because it was unlikely that he would be punished for it. This is textbook right wing populism: make an enemy out of some group that is unlikely to hit back and you can appear strong. This is often done against minorities but the EU is perfect for it as it is powerful enough to actually look like a threat but has no means to punish small "violations".
The hungarian leading party (fidesz) has already declared they will not veto Ukrain's adoption to the EU so I personally really doubt they would veto Finland's entry to the NATO. They are playing a double-game (sorry, this probably doesn't make sense in english, it's a hungarian expression): Orban had a famous quote he said to EU leaders "don't listen to what I say, only to what I do", which in practice means alienating the EU in rhetoric but not so much in practice. Of course this is not so simple as other EU members also care about the rhetoric too but it still summarizes Orban's geopolitics fairly well. This is not unique either, Merkel had similar "two-faced" approach to Russia and even toward Hungary, Orban is just significantly less popular in the west due to his populist anti-democratic tendencies.
Videos usually had about 10% of the number of votes compared to views, so depending on how you define (vast) majority it is correct.
I remember watching some video and looking at dislikes: why is this downvoted this much? Pretty run of the mill video with nothing outrageous or even particularly controversial. Oh it got posted _there_...
Still, I don't think this is the reason. Visible votes simply modify what other people vote on and youtube wanted full control over what users watch instead of the few who actually votes. It is simply not a good metric most likely. Was it improved by removing the dislike bar? I would be surprised if was...
I just want to point out the irony in falling back to good old centralized fintech companies to actually use the supposedly decentralized currency.
Sorry, I know, off-topic.
Source? I've seen this claim many times but is it just a feeling or is there at least a whistleblower?
These were my exact thoughts about CS vs Quake back in the days. Quake had no downtime, all action while CS had time between rounds or just tense moments without anything actually happening. I think this dynamic was an important point behind the success of battle royal games too. An all action game is like an action movie without a moment's pause - it just doesn't work for most people.
Well, if you collect commits on a feature branch you can just create a new one with a slightly different name - all the commits on your branch are rewritable before you push it in and create a pull request (deleting and pushing it again with different history does not work for me). I often do this for rebasing, I doubt anyone would have ever noticed if I changed the authors.
I call this argument "totalitarian democracy argument". Yes, NGOs and media are not democratic, nor family, corporations, etc... Not everything is and it is fairly obvious that if everything was controlled through the goverment (so in some sense, democratic) it would lead to one party rule almost immediately, which unfortunately seem to be the point in this particular case.
It is also very ironic that in Hungary these independent (i.e. from state) institutions are replaced with ones who's mandate last longer than 2 terms of election - fairly obvious that public control is not the goal, but the exact opposite.
I think the book is still useful with all it's flaws, mainly because "overengineering code" is like exercising too much: sure, it happens to some and can be a big issue, but for the vast majority of people it's the opposite that is the problem.
If the recognition is done by autoencoder network as I suspect, then generative adversarial network could break it easily.
My guess would be that they trained an autoencoder network for facial recognition and they use the encoder part for fingerprinting and check images by the decoder part.
I'm developing driver assistance functions for vehicles and I have experience using simulink, although not anymore, fortunately. It's usage is converging slowly but steadily to zero in the company I work for, although we still use it for controllers, but mostly only legacy "code". Why is it?
- First is the most serious, debugging: this is a deal breaker alone. Once your code will go into production, and once, bugs will be found in it. How do you debug it? A typical variable in simulink looks like this: `Simulink_Sum_Node_4521` (something similar, I don't remember exactly). Yes, that is a variable in the c code generated from simulink graphs. It's index start from 0, but you have quite a lot of it - just count the operations in your normal code, and don't forget that (a+b)/c already contains 2 operations! Not sure if the original programmer was lazy to name them all, but not sure if it is possible to meaningfully name them either. If you want to get (a+b)/c^5, (a+b) might not have a meaning on it's own. So you stuck debugging a 10k+ lines long file containing variables like this.
- Reading is natural: we read from early childhood, we write from early childhood. Did you draw computational graphs as a kid? In school maybe? Yeah, me neither. When I want to tell my colleague a formula, I'm not going to draw him/her a computational graph, I'm going to write it down. Hell, you are reading right now, and not looking at a graph consisting words and arrows that represent relations! Imagine how that would look like...
- Graphs take up much more space on the screen than text. Grab a pen and draw a computational graph of a Fourier transformation! It takes up a whole screen. As a formula, it takes up a tiny fraction of it. Our state machine used to take up about 2m x 2m on the wall behind us.
- Also the other reasons already written in this thread...
Working with simulink convinced me entirely that graphical programming has no future whatsoever.
Yes, it is sarcastic: "As a seller (only) myself, I am one of the many victims of this scam."
Maybe the problem is not where we search for it. There is a theory that willpower is a limited resource [1]. If this theory is true, it means productivity hacks cannot work, because the problem is not in our weekends, it is in our weekdays: in our work that demands all of our willpower, in our relationships, social interactions, or in our inability to relax.
I read an article some time ago that I cannot find. It stated that extreme procrastination is the result of a form of drug abuse: we need our daily dose of fb, youtube, etc to keep us functioning, but that takes all the time, and when we realize we wasted another day, it makes us sad, and the cycle begins anew. It is also no secret that these sites are designed to be addictive.
This article might help you too (well, the original, but I couldn't link that because, most likely, the author of the original essay procrastinated the renewal of the web page for too long, and it cannot be reached): https://www.theguardian.com/lifeandstyle/2012/sep/07/change-...
[1] https://www.apa.org/helpcenter/willpower-limited-resource.pd...