blackbear_

32 ms·

1. 421.▲
   by blackbear_ 4y ago

   In the general case no because the final metric we care about (eg, accuracy or AUC) is not differentiable (ie, we cannot compute its gradient) with respect to the hyperparameters, especially the discrete ones. However a recent work at this
2. 422.▲
   by blackbear_ 4y ago

   I'm one of those suckers and I'm glad I wasted years of my life being paid to study things I found interesting, on my own terms, without pointless meetings, impossible deadlines and unreasonable bosses, dropping projects and colle
3. 423.▲
   by blackbear_ 4y ago

   I believe that a lot of this cynicism comes after years of teaching and seeing that most students do not really care about the topic and do not want to put in more than the minimum effort required to pass. You don't realize how easy
4. 424.▲
   by blackbear_ 4y ago

   Neat idea and great website! I thought about having some periodic reminders, maybe using something in the spreadsheet (e.g. columns) as prompts or questions to be answered, but have not implemented anything yet. I am pretty disciplined myse
5. 425.▲
   by blackbear_ 4y ago

   I wrote a Telegram bot to save messages into a Google Spreadsheet [1], with some sime parsing rules to split the text into columns (essentially, #hashtags for column names and everything in between as content). The idea is then to use pivot
6. 426.▲
   by blackbear_ 4y ago

   I was really disappointed when they removed support for argument unpacking. It was just so useful, especially in lambdas, for example Mdist = lambda (x1,y1),(x2,y2): abs(x1-x2)+abs(y1-y2) p1,p2=(1,2),(3,4) Mdist(p1,p
7. 427.▲
   by blackbear_ 4y ago

   Disagree: great (non-fiction) writing should provide information in an efficient and structured way, so that readers can quickly understand the key points and if reading further is worth their time.
8. 428.▲
   by blackbear_ 4y ago

   No, because spam filters are pretty good. But that may change, too.
9. 429.▲
   by blackbear_ 4y ago

   > have never been shown a concrete piece of scientific evidence demonstrating Vim's claimed superiority according to any real-world metric I wonder if you've ever seen scientific evidence of the contrary? Or any kind of scienti
10. 430.▲
    by blackbear_ 4y ago

    I don't understand this line of thinking. The purpose of those regulations is to reduce e-waste. How is the EU's fault if some companies produce cheap shit that doesn't work very well? And do you really think things would get
11. 431.▲
    by blackbear_ 4y ago

    It's not so simple in the digital domain where CTRL+C/ALT+TAB/CTRL+V is all you need to have your very own copy of things.
12. 432.▲
    by blackbear_ 4y ago

    I think that the red line is definitely crossed whenever a third party is involved. It's okay for a shop to optimize what they sell and recommend things based on customer data. What is not okay is for a shop to share customer data with
13. 433.▲
    by blackbear_ 4y ago

    > General counsel Erich Andersen wrote to staff that a “misguided plan was developed and carried out by a few individuals within the Internal Audit department this past summer”, adding that those involved “misused their authority to obta
14. 434.▲
    by blackbear_ 4y ago

    Hm, I cannot decide whether this advice is useful or not
15. 435.▲
    by blackbear_ 4y ago

    I can imagine the material used for lenses is also rather cheap, and yet somehow they end up costing undreds of euros each... I wonder why...
16. 436.▲
    by blackbear_ 4y ago

    Furthermore, fusion does not produce highly radioactive, long lived nuclear waste. “Fusion produces only low level radioactive waste — more than fission does — but this low level waste does not pose any serious danger,” said González de Vic
17. 437.▲
    by blackbear_ 4y ago

    I like tmux a lot more once I started using ` as prefix
18. 438.▲
    by blackbear_ 4y ago

    Sorry but this is just wrong, using only fully connected layers would result in pretty bad performance on images, text, audio, etc., or at the very least require much more data to perform well. At least use the right type of architecture fo
19. 439.▲
    by blackbear_ 4y ago

    > they might have a big base of predefined response "templates" This could the result of overfitting on the human feedback that it was trained on [1]. [1] https://arxiv.org/pdf/2203.02155.pdf
20. 440.▲
    by blackbear_ 4y ago

    Would be a great comment if it ended with '/s'
21. 441.▲
    by blackbear_ 4y ago

    So just like lobbying, just from the other side?
22. 442.▲
    by blackbear_ 4y ago

    Not really, that is just a lower bound. Companies are incentivized to pay the least amount possible for labour regardless of productivity, so long as it profitable. If a company pays somebody 100k a year, it means that (1) the value produce
23. 443.▲
    by blackbear_ 4y ago

    What you say is true. But I urge you to consider Facebook's business model and what may be pushing them to work on these things. In other words, how can facebook use this to increase user engagement among each other and with ads? The p
24. 444.▲
    by blackbear_ 4y ago

    They will just do the same shit that's done on android smartphones, locking the bootloader and showing you the middle finger.
25. 445.▲
    by blackbear_ 4y ago

    Business is not the only sector where language is such a hindrance. I am a foreigner and came to Germany to do a PhD, I work 60 hours per week for a miserable salary and I do not have the time nor energy to learn German after work. About ha
26. 446.▲
    by blackbear_ 4y ago

    There's an app that constantly reminds me to exercise in a similar way: "Your body needs energy! You haven't exercised for 567 days" - at this point it's just funny to see the number go up and up
27. 447.▲
    by blackbear_ 4y ago

    Automatically writing cover letters could indeed be a good use for all the recent advances in AI text generation
28. 448.▲
    by blackbear_ 4y ago

    Sounds quite similar to a terminal, no?
29. 449.▲
    by blackbear_ 4y ago

    There are many forms of bullying that only involve words, and not violence. For example: > I think it’s best if you don’t contribute to this discussion any more to avoid embarrassing yourself further.
30. 450.▲
    by blackbear_ 4y ago

    From the article: > Here’s why: As researchers feed data into AI models, the models are expected to become more accurate, or at least not get worse. However, our work and the work of others has identified the opposite, where the reported