vrighter

36 ms·

721.
by vrighter 1y ago
no. It is very clear that they will either spend it on bullshit, or pay it out to the ceo or something. And then selling out because "they're broke"
722.
by vrighter 1y ago
but.... but.... isn't it supposed to be immutable? It did something that the code allows for (because supposedly, code is law) which some others didn't like, so they agreed, out of band, to change stuff, instead of realizing that
723.
by vrighter 1y ago
yes they do. training code, dataset, and random seed. given all three, you can recreate the final model (the weights). The set of all three is the source code
724.
by vrighter 1y ago
I hate the "accessible to the layperson" argument. People who couldn't do art before, still can't do art. Asking someone, or something else, to make a picture for you does not mean you created it. And art was already acc
725.
by vrighter 1y ago
Except that a corrupt packet can easily be detected when compared to a valid packet (is the checksum valid?). There is an algorithm to execute that can tell you, with high confidence, whether a given packet is corrupt or not. In an LLM a to
726.
by vrighter 1y ago
only if you lower your standards as to what "quite well". The biggest con by the AI industry so far is convincing people that 90% is somehow "quite well". 90% is only enough for uninformed people to buy into it, and fuel
727.
by vrighter 1y ago
"hallucinations" CAN'T fade away. They are the one and only thing LLMs can do. If you remove that your output would be absolutely nothing. It is intrinsic to how they work. Anyone claiming they can eliminating has a bridge to
728.
by vrighter 1y ago
Yeah decimating them isn't enough (decimating really means eliminating 10% of something)
729.
by vrighter 1y ago
Talk about javascript, present code in typescript. This would put me (the interviewee) off
730.
by vrighter 1y ago
The game itself also contains code that might be reusable for a future project. Among other things, the game engine itself. They have no problem shipping that to people though? Why is server code any different?
731.
by vrighter 1y ago
Then you design for it up front. There is nothing that stipulates that the eol build has to be developed right at the end. Then if they go bankrupt, the job is already done. It isn't more work, it's just knowing a feature needs to
732.
by vrighter 1y ago
so are you telling me nothing breaks ever if you update python 3 minor versions? Because if so, I don't believe you. They also make breaking changes
733.
by vrighter 1y ago
it gives you an answer. Not a highly intelligent one. Just an answer. And if it doesn't know what it's talking about, it'll still give an answer.
734.
by vrighter 1y ago
So you suggest training a model for each individual student? Because LLM "inference" sure as hell isn't capable of tailoring anything to anything, or change in any way. And also, how is personalized bullshit better than gener
735.
by vrighter 1y ago
python is a really bad example. Code constantly stops working properly with language updates.
736.
by vrighter 1y ago
ah but neural networks are universal function approximators! (proceeds to ignore the size of network needed for an adequate approximation and/or how much data would be required to train it)
737.
by vrighter 1y ago
I haven't heard anything but negative comments about liquid glass from any apple users I personally know. Here goes everyone else rushing to do things just because apple did them. Not everything apple does is good. This is one of those
738.
by vrighter 1y ago
twice as fast, but asked to render 4x the pixels. Do the math
739.
by vrighter 1y ago
The deduplication in the product I worked on was implemented by me and a colleague of mine, in a custom format. The point of it was to do inline deduplication on a best-effort basis. I.e. handling the case where the system does NOT have eno
740.
by vrighter 1y ago
i saw the video, and it seems really similar to just a tree view with all the children collapsed by default. How is this different?
741.
by vrighter 1y ago
I used to work on backup software. Our first version did exactly that. It was a selling point. We later switched approach to a deduplication based one.
742.
by vrighter 1y ago
but don't you enjoy the memory safety of it all?
743.
by vrighter 1y ago
you bury your hand in the sand and pretend the 2nd llm magically lacks the limitations of the first. Which begs the question... why not use the 2nd llm in the first place if it is the one who actually "knows" the answer?
744.
by vrighter 1y ago
What are the strengths? And saving developer time is not one of them. This is assuming that reviewing and understanding the code (properly) takes as much time as actually writing it. (If it's not, then you're not saving work, you&
745.
by vrighter 1y ago
"Smarter" and "stupider" are completely subjective terms. It translates one language into another, while maintaining logical correctness and its original semantics. Some cases are easier to do than others, but it doesn&#
746.
by vrighter 1y ago
So software 2.0 is huge and inefficient. It is nondeterministic. It is not reproducible. It is not correct (with the same definition of correct used for usual stuff. I.E. If there is a right answer, it will give that one and only ever that
747.
by vrighter 1y ago
you really really REALLYYY don't want an llm involved anywhere near this stuff. Hardware bugs are notoriously expensive (see pentium fdiv bug).
748.
by vrighter 1y ago
A compiler is a language translator. To me it makes no difference whether it generates machine code (one type of language, machine executable), asm (equivalent language but which needs additional preprocessing into machine code), or some ot
749.
by vrighter 1y ago
We only give select and execute permissions. This has saved our asses a lot of times. Do not underestimate the damage a development team with a high turnover can do.
750.
by vrighter 1y ago
So how does it save me from having to read the man page? How do I know if the command is safe to execute when I can't trust LLM output? Also, for "safe" commands, how do I know that it is doing what I asked it to do (without

More ›