I was excited to hear some details about how LLMs are changing big companies' workflows. This is a fairly straightforward problem and good on them for discussing it. My first gut reaction is that there is no way I would let LLMs write a significant amount of my codebase. And then I remembered that not invented here (NIH) is a problem I have. If I let it decide what code I ship, I am making the same mistake as writing all my code from scratch. When I find myself worrying about alignment, that is what rigorous computer security policy is for. If it's not catching serious bugs written by LLMs it's not catching bugs written by people. Treating my own code as if it was written by someone else is just another part of computer programming.
Obviously yes. While I have the privilege of earning money from services instead of products, I still think that producing creative works is important and should be done whether the motive is profit or not. Many things are not profitable. Should we leave them unwritten? Leave them to those who have the time to spend. For those who have chosen the life of producing products that are easily copied, it is part of _reality_ that those things will be copied when copying benefits the copier. That doesn't mean I think all books should be free. So many books I buy because I don't have another choice or because I want to support the author. But expecting everyone to be in the same situation as me is nonsense.
Yeah art has objective value and subjective value. Every once in a while you'll find something with a lot of subjective value. Finding something with both is also a thing but it's not easy to find them for a low price.
The reason that expensive art exists is because there's a market. The fact that the market is weird and in decline doesn't change the fact that wealthy people find art to be a worthwhile thing to buy.
I'm surprised no one in the comments has mentioned overfitting. Perhaps this is too obvious but I think of it as a very clear bug in a model if it asserts something to be true because it has heard it once. I realize that training a model is not easy, but this is something that should've been caught before it was released. Either QA is sleeping on the job or they have intentionally released a model with serious flaws in its design/training. I also understand the intense pressure to release early and often, but this type of thing isn't a warning.
In what way are we not ready? If you don't have the necessary critical reasoning skills to doubt Joe Rogan telling you about complete bullshit, no one needs technology to trick you into a scam. If you have critical reasoning skills, it's pretty hard to convince you to part with your money. Who exactly does this scam target?
This helps answer a serious argument I had on whether Puerto Rico and/or Washington D.C. should become states. Some people argued that the the flag could stay the same. I think the most clever answer would be to make a flag with a star randomly placed to annoy flag-obsessed nerds. More seriously, I think the reason that a 51st or 52nd state can't be added is that the urgency of the matter is lacking. Doing something about statehood would require it to be exceedingly popular and even then the odds are poor.
I used emscripten to make physics and math software written in Fortran available to my JavaScript program. I would've used wasm, but it was in poor shape back when I wrote the software. There are a lot of things besides games and graphics written in C/C++ that can be used in JavaScript as a result of emscripten and wasm.
You ignore the fact that there are perverse incentives among the participants. It's possible to implement, and I'm doing it myself. If I had more time to spend on it, we could end spam. Instead I am fine as is: most of the spammers have given up.
Can you give any advice on how to get in the right mindset to give significant part of your income to charity? Or did you have no trouble deciding that? I understand how fortunate I am, but I cannot convince myself to give.
Step 3: Find (or geek out in) a hobby that you can obsess about and that your friends and family can accept.
Step 4: If you can't find out a path toward self-actualization, ask your employer or boss about ways that you can improve the business mission goals. In that path, you should find a path toward self-actualization and go for it.
Step 5: Accept the excellent result you get because outcomes are very dependent on luck.
Rock Band 3 was pretty close to that game for me. It convinced me to get real drum training and to learn the keyboard.
I want to play a game where all the NPCs are AI trained with Seq2Seq neural networks. I have been trying to write the game off and on for a while, but it's not easy to write. There are some things that come pretty close, but are not quite there.
MysteryTwisterC3 is an excellent way to learn RSA, ECC, and so on through the process of cryptanalysis. https://mysterytwister.org/home/welcome/
To learn how to implement stuff correctly, I recommend trying to work through an algorithm (say SHA-1 for example), and try to find people that explain it. Once you're ready to implement it yourself, go through the list of flaws in common implementations (side-channel, power analysis, diff, linear, etc) and try to prove whether your implementation is vulnerable or not vulnerable. Once you've done this more than once, you can do it for a lot of things.
I don't have a lot of experience with IMPALA, but the sequence of events you describe should be very easy for an end-to-end system. Assuming you don't have an end to end system, just getting a gradient would result in rapid learning of that sequence. I'm surprised that at 2.5e9 frames you're not done. Perhaps there is a hyperparameter issue. Sorry I can't help but it sounds like you are in the same place I am with ML project. Good luck.
Here I have a little bit of experience. When a person has shown themselves to be a pathological liar, you can't judge their ideas on their own merits because the cost of judging far outweighs the cost of producing garbage ideas. That is why attention is so valuable, there are plenty of people out there who do not make it their trade to espouse nonsense.
The emulator doesn't work in Firefox because of a syntax error, await in the top level isn't supported. Remember to test your code in more than one browser.
Is your game for sale or available? I'm releasing my first game now and am curious to see examples of negative results -- since all I ever see are positive results.
Your questions are answered in the thread, which is a pretty significant feat to overcome (I ditto the pomodoro method, exercise, and higher expectations). That said, laziness is not all bad. Keep some of it in case you need it. Given sufficient motivation, I recommend aligning your productivity with meaningful outcomes. If you can't, you will be back here looking for another solution in not much time with a bigger problem. But then again, there are worse things in life than being lazy.
It's been a while since I visited Chicago, but I stayed in a hotel downtown. Not only was public transit excellent and easy to use, but walking was a delight. I don't remember there being a grocery story anywhere, but Google Maps shows quite a few. Chicago's downtown may have a lot of problems including being expensive and a high crime rate, but poor access seems inaccurate.
Do you think that static analysis is a valuable tool for security research? Do you recommend static analysis software to a single developer with a limited budget or an amateur?
It now is a lot more clear what's going on here. The discoverer of this issue is basing his argument on the fact that when you verify a fingerprint, you are now confident that your end-to-end encryption won't transparently send your encrypted data to someone with a different keypair. The other side of the argument is that if WhatsApp actually did what you expect, data would be lost when a person switched phones in the middle of someone sending them a message. As a person who doesn't switch phones very often, I would prefer an end-to-end encryption to never send data to a different public key than the one I've used before. I would rather lose data than divulge it to a third party who has the ability to spoof the recipient's phone. This would only come up whenever someone switched their phone when I was sending them a message, so it's pretty rare.
To me the trade off is a no brainer, and apparently to Facebook and Whisper Systems the trade off is a no brainer in the opposite direction.
No, by itself the number of vulnerabilities patched is only useful in telling us which software has been tested and found to be lacking in the past. Along with other metrics (severity, static analysis results, code quality, complexity, reports by an independent auditor, availability of a testing framework, and competitor quality), this can be used to decide which projects need to be replaced or improved upon.
The reason that many vulnerabilities have been found in Chrome is because it is a very large and complex project. The bug bounty only gives people the necessary additional motivation to work on it during business hours. Other projects that lack bug bounties have found similar numbers of bugs (Wireshark and ClamAV to name a few) due to their complexity.
Thank you very much for testing this exploit. The vulnerability is in gssapi.c which is only compiled in if HAVE_GSSAPI is defined. This is an optional configuration parameter, so it sounds like the configuration you tested did not have GSSAPI/kerberos enabled. That's good news for users who have a similar setup, it will give them ample time to switch to a different IPsec implementation.