Szpadel

27 ms·

1. 301.▲
   by Szpadel 3y ago

   Of course :) as always autocorrect twist my words :)
2. 302.▲
   by Szpadel 3y ago

   i run the latest lts kernel on CentOS 7 in production and do not have any issues so I assume that it's not really a big deal as Linux kernel is fairly good with backward compatibility
3. 303.▲
   by Szpadel 3y ago

   is it only me, or wording is kind of weird > Apple Silicon Macs now natively supports Unreal Engine 5 while technically correct, it sounds to me like this is an achievement of apple silicone like apple managed somehow to now support UE5
4. 304.▲
   by Szpadel 3y ago

   the thing is that you need additional CPU power to compress and decompress that data, maybe that is not much by today standards, but when accumulated it could be a significant number
5. 305.▲
   by Szpadel 3y ago

   you can just enable the camera and show your ID and someone on the other side will check if that's you /s
6. 306.▲
   by Szpadel 3y ago

   This could allow use some training methods similar to alpha go zero. I'm thinking if training ai to output edits instead of next token would be enough. Training would be similar to currently used: just provide some sentences with missi
7. 307.▲
   by Szpadel 3y ago

   AFAIK there is no limit of depth in modern filesystems (at least in design), there is usually limit in path length (in bytes) but that's just that you have to limit somehow implementation to prevent pathological instances. Eg imagine b
8. 308.▲
   by Szpadel 3y ago

   I think the main issue is in a way that gpt like models generate output. eg. when you write some serious document or write code you don't just start from the first line and write the whole application in one go (line by line) you write
9. 309.▲
   by Szpadel 3y ago

   > We would like to automatically construct datasets, but we don’t currently have any good approach to doing so. wouldn't fine tuning an existing LLM model like GPT4 could be enough to filter our automatic dataset from "garbage&
10. 310.▲
    by Szpadel 3y ago

    my analogy would be that you have a house, in that house you have rooms and in those rooms you have humans why you cannot have houses inside humans? while analogy to a filesystem would be: you have a box, and in that box you can have anothe
11. 311.▲
    by Szpadel 3y ago

    I really wanted to repurpose my old phone as a security camera , but android power management and lack of sensible way to automatically restart the app for an IP camera made it not very practical. this could be a promising project but devic
12. 312.▲
    by Szpadel 3y ago

    Thanks, apparently I'm just blind
13. 313.▲
    by Szpadel 3y ago

    I didn't find this in the article and I think this would be a nice context to have: what was battery life before?
14. 314.▲
    by Szpadel 3y ago

    The question is if they limit abnormal voltage, or just lock to "safe" ones. this might not be a solution, this could be just a workaround basically crippling oc AFAIK none of them said anything about what is happening nor what th
15. 315.▲
    by Szpadel 4y ago

    I never noticed that before, now I cannot unsee it
16. 316.▲
    by Szpadel 4y ago

    I'm surprised that there are many cases as to enable sideboarding on Android you basically have to go to settings and enable option that basically tell you that this isn't a good idea if you don't know what you are doing and
17. 317.▲
    by Szpadel 4y ago

    assuming that this failure was caused by failed engines, I think this is a very important lesson, engines can fail during real missing and if their algorithms couldn't account correctly that, this is a very important thing to improve.
18. 318.▲
    by Szpadel 4y ago

    keep in mind this is still a prototype rocket, they have no guarantee that they won't learn anything important about how this should look like or what issues it could introduce they used something they knew to work so there is one less
19. 319.▲
    by Szpadel 4y ago

    I use btrfs + bees ( https://github.com/Zygo/bees ) for such cases
20. 320.▲
    by Szpadel 4y ago

    I'm also not an expert by any means, but wouldn't this be used for free computation? like a CPU that can compute without consuming any power? if anyone knows better I would like to know why or why not this is an possibility
21. 321.▲
    by Szpadel 4y ago

    you are referring to the dolly model? I think the training set could achieve similar performance if we would fine tune similarly sized model
22. 322.▲
    by Szpadel 4y ago

    AFAIK current models can run even with 64GB, but I would assume that we will very likely have bigger models very soon so I guess the answer is as much as you can afford
23. 323.▲
    by Szpadel 4y ago

    Not sure about you, but I also have a huge backlog of projects that don't exist yet and I would like to create them. but usually they're too complex to do over the weekend and I don't have that much time to spare. I already t
24. 324.▲
    by Szpadel 4y ago

    does anyone know how such a plugin definition is passed to gpt model? I tried using eg. langchain but it didn't work well with gpt3.5 and this looks like it's more capable anyways.
25. 325.▲
    by Szpadel 4y ago

    I personally would be more excited to get any access to the large context gpt4 model.
26. 326.▲
    by Szpadel 4y ago

    If you see generating model weights in the same way like generating executable binary from source code, then sure. But AFAIK this is just the first step to get initial weights and later you need much more work to fine-tune this to get usefu
27. 327.▲
    by Szpadel 4y ago

    I very recently compared the usefulness of gpt4 vs bingai assuming that theoretically bing should be gpt4 with search. I assumed that for codding tasks access to up to date documentation would be very beneficial for that model. unfortunatel
28. 328.▲
    by Szpadel 4y ago

    I guess they could autonomusly generate spam in comments, but i dont believe they could work in any coordinated manner to achieve some high level goal. (At least until we discover some ground breaking technique to boost performance in task
29. 329.▲
    by Szpadel 4y ago

    I think we are way off having ai that's able to replicate itself anywhere, and currently all sensible agents are based on OpenAI API, so you can just kill internet/account/etc and agent is dead therefore I would worry much ab
30. 330.▲
    by Szpadel 4y ago

    is uncertain, as with codding you need white room methods to prove that new code is not contaminated with patented implementation, as it might be here, so basing anything on an existing model could be also copyrighted.