dimatura

27 ms·

1. 181.▲
   by dimatura 4y ago

   With discrete logic you'd be committed to the model for whatever its lifetime is, which might be acceptable depending on the application. With FPGA you would still be able to update it after it's deployed.
2. 182.▲
   by dimatura 4y ago

   Only skimmed the paper, but I'd say one potential issue is that they show good results in tabular data domains, where it's likely easier to get away with minimal models than in other domains, such as visual/lidar/audio d
3. 183.▲
   by dimatura 4y ago

   This thing was announced May 2022, and since then the shipping date has been pushed back several times. Any indication it's finally going to be available? Regardless - once this one comes out, I'd consider getting one, maybe on th
4. 184.▲
   by dimatura 4y ago

   I read the article, pressed play on the song and thought "wow, the author did a great job at sounding like the decemberists", then scrolled back up and saw the author... doh!
5. 185.▲
   by dimatura 4y ago

   Yep, I forgot the name of the settings but my Sony mirrorless has some modes which are in the same spirit of the typical smartphone camera - sort of like an enhanced Auto mode. Though In my experience, definitely less aggressive than what
6. 186.▲
   by dimatura 4y ago

   I think that description applies to Riffusion, one of the earlier models in this area that was a pretty straightforward to adapt image-based diffusion models to making music, since you can treat spectrograms as images. But this model uses &
7. 187.▲
   by dimatura 4y ago

   My day job is in ML, but I also enjoy music making as a hobby (on a very amateur level - mostly making 4-bar loops on a handheld tracker or knob twiddling on 90s synths I couldn't afford as a kid). I see an interesting mix of curiosit
8. 188.▲
   by dimatura 4y ago

   Oh, there's already all sorts of work mashing together GPT text with text-to-speech and speech-to-text. There were some viral videos featuring "Steve Jobs" talking about AI and one being interviewed by "Joe Rogan".
9. 189.▲
   by dimatura 4y ago

   I believe parts of this model could be adapted to accomplish this. The "mulan" network generates embeddings from music and text so that they "match" each other, i.e. audio segments and text captions that go well together
10. 190.▲
    by dimatura 4y ago

    I understand what you're saying, although it could be argued that at least for some types of image tasks one would prefer something like an SVG output with layers to make it easier to edit. For music, I think it's partly an academ
11. 191.▲
    by dimatura 4y ago

    Gets confused with multiple exclamation or interrogation signs in a row; seems to forget about the bottom dots.
12. 192.▲
    by dimatura 4y ago

    You might enjoy a bit of Camus. https://en.wikipedia.org/wiki/The_Myth_of_Sisyphus
13. 193.▲
    by dimatura 4y ago

    I'm in the same boat, but using trackpoint + vim shortcuts for most things (including window manager, browser, and of course, vim). Works really great for day-to-day coding and browsing. If I have to do something that's easier wit
14. 194.▲
    by dimatura 4y ago

    Binary neural networks, where weights and/or activations are just 0/1s, are an active research area. In theory they could be implemented very efficiently in hardware. But in contrast to FP16 (or to some extent, int8), just quantiz
15. 195.▲
    by dimatura 4y ago

    Yes, that totally makes sense - you can think of a function as an "infinite-dimensional" vector, which you can express in the (orthogonal) basis of a bunch of sines and cosines of different frequencies.
16. 196.▲
    by dimatura 4y ago

    I'm not sure what the main point is here. The paper is definitely sketchy on details, and the main idea is definitely simple enough to resemble a lot of other work. I wouldn't be surprised if someone (maybe a certain Swiss researc
17. 197.▲
    by dimatura 4y ago

    Yeah, no. Reading the paper I don't really see anything but a superficial resemblance to stacking. Hinton was active back when Wolpert introduced stacking and I'm fairly sure he is aware of it. If anything it much more closely res
18. 198.▲
    by dimatura 4y ago

    "it's impressive for what it is, but ..." is probably a good way to describe chatGPT's performance for a lot of tasks - but what's interesting is the vast diversity in these tasks. I've asked it write me a few
19. 199.▲
    by dimatura 4y ago

    The paper/article this refers to has more context. Lots of nice diagrams! https://transformer-circuits.pub/2022/toy_model/index.html .
20. 200.▲
    by dimatura 4y ago

    Not just industry. Linux has been my daily driver OS for ~20 years now, but I just installed windows 10 on an old laptop so I could run some old music production-related software on it, such as editors for the Virus TI and Nord Modular synt
21. 201.▲
    by dimatura 4y ago

    You can change the chunk size on restic, for what it's worth.
22. 202.▲
    by dimatura 4y ago

    I suspect a free modular synth app developed for an 8-bit fantasy console is not really aiming to replace a $500 professional DAW.
23. 203.▲
    by dimatura 4y ago

    I'm an FM junkie and own various pieces of gear based on old yamaha FM chips, as well as actual yamaha FM keyboards. I think the various emulations out there are actually pretty good, and I have to admit I'd probably have a tough
24. 204.▲
    by dimatura 4y ago

    Yeah, though the dev boards (the SD-1) are relatively hard to find - just recently got a hold of one. Tiny! The chip itself seems like it might be easier to procure.
25. 205.▲
    by dimatura 4y ago

    As others have said, these are essentially scalpers - the authorized retailers still have at the original price, but as a result are also sell out quickly whenever there's any stock. https://rpilocator.com is a tracker of a
26. 206.▲
    by dimatura 4y ago

    Data transmission over audio is still found in some more modern gear. The newer line of Teenage Engineering's Pocket operator synths/samplers, released in the late 2010s, also uses an audio-based format for import and export of it
27. 207.▲
    by dimatura 4y ago

    thanks for the pointer, I'll definitely give deeplake a closer look
28. 208.▲
    by dimatura 4y ago

    I've been observing and trying out all sorts of different solutions in this space (dvc, git-annex, git-lfs, quilt, just rsync, etc), and so far haven't been super happy with any of them. So it's nice to see more alternatives.
29. 209.▲
    by dimatura 4y ago

    I work in a startup that uses ML for computer vision and I get the impression that there's significantly less commercial tooling to handle computer vision (or more generally, computer perception)-type data than tabular or text data. I
30. 210.▲
    by dimatura 4y ago

    You would want to tune the pack size, https://restic.readthedocs.io/en/stable/047_tuning_backup_pa...