94 Matching Annotations
  1. Nov 2022
    1. Because empowerment is a function of a transition operator, there is a natural synergism between empowerment and OBEs: the OBEs create hierarchical transition operators, and the valence function can evaluate hierarchical empowerment change defined on these operators. The valence function can then be used for goal selection, wherein the agent chooses a policy sequence that realizes goal states which produce maximum empowerment gain. In doing so, the agent will seek freedom and avoid internal death-states that undermine its ability to control both external and internal states in the future, thereby exhibiting the capacity of predictive and anticipatory self-preservation.

      this is way too detailed for an abstract, and not detailed enough to be clear

    2. empowerment

      unclear because unclear what channel capacity is

    3. channel capacity

      what's that

    4. physiological

      what's a physiological model in this context

    5. state-time

      what's that

  2. May 2022
    1. Access to the model will be granted to academic researchers; those affiliated with organizations in government, civil society, and academia; along with industry research laboratories around the world

      oh so its not on github

  3. Apr 2022
    1. machine-translate English prompted instances into language X write prompts in language X for datasets in English

      how would either of these two work? Especially the latter seems counterintuitive.

  4. Dec 2021
    1. Your optimal city is one of the top 300 largest cities.


    2. Your optimal city is the city with the greatest density of people who share your worldview

      optimal for what function?

    1. iteratively applied Layer Normalization across layers can lead to very high gradients during the first iterations

      I wonder if there's a mathematical proof of this somewhere. How do we convince ourselves of this assertion?

    2. residual connections are crucial for enabling a smooth gradient flow through the model.

      why/how does a residual connection enable "smooth" gradient flow through the model? What is smooth gradient flow?

    3. general advantage of a wider layer instead of a narrow, multi-layer MLP is the faster, parallelizable execution

      I don't see why this is the case - it boils down to a matrix multiplication anyway doesn't it?

  5. May 2021
    1. otherwise the data should be normalized.

      How do we deal with outliers when doing this?

    2. or standardized with a zero mean and a standard deviation of one.

      what about input variables that are not distributed normally?

    1. In linea con questo principio una recente ricerca dimostra i meccanismi neurali che supportano la riduzione del dolore basata sulla consapevolezza

      ehhhh, dove? un citazione?

    1. There is no way, all ML models is not about phenomen understanding, it's interpolation methods with hope "that it works". Start with such questions confidence, robustnes to noise there is no answers.

      seems unnecessarily pessimistic

    1. Generally, the term “validation set” is used interchangeably with the term “test set” and refers to a sample of the dataset held back from training the model.

      I wish people would stop doing that

    1. Statistical-based feature selection methods involve evaluating the relationship between each input variable and the target variable using statistics and selecting those input variables that have the strongest relationship with the target variable.

      Equivalent to "Filter-type methods" as described on the Wikipedia page on Feature Selection

  6. Apr 2021
    1. vector field (or vector-valued function)

      Any vector field will be a vector valued function. Not all vector-valued functions are vector fields. In this case we focus on vector field


  7. Feb 2021
    1. This means setting aside time to have conversations about how much friendship you’re looking for — whether a mere running buddy or a BFF — while still allowing for the relationship to evolve

      This sounds like awful advice. I'd be weirded out if someone started this convo with me. "How much friendship", bruh what

  8. Jan 2021
    1. Claiming ownership of a subarea doesn’t lock you into that area, but it does make you more focused and conveys scientific maturity

      I think this is good advice, but it is probably not necessary for Masters applications (Yes, i know, this article is about PhD applications, just wanted to point it out)

    1. reading it daily

      That frankly seems a bit suffocating. I thought spaced repetition implied some spacing (usually increased spacing) between interations?

    2. methodical sending of messages from the conscious mind to the subconscious mind so that we begin to believe new things about ourselves and our capabilities.

      2 observations

      • Spaced repetition is a bit broader than that, and doesn't necessarily have to be used in this context
      • This kind of sounds like self-brainwashing -- that isn't necessarily a bad thing though, I suppose.
  9. Dec 2020
  10. www.youtube.com www.youtube.com
    1. Listening to this while typing some notes on vim


    1. Sound of Silver by LCD Soundsystem.

      This is a good album but it definitely isn't ambient and in general just doesn't feel like Kid A/Amnesiac/Vespertine/Substrata which is what OP seemed to be after :/

      The song 'Sound of Silver' SLAPS tho

    2. Moon Pix by Cat Power

      idk why but i wasn't expecting to find Cat Power here but that album kinda slaps

    3. Lift Your Skinny Fists

      Of course.

    4. Lost And Safe by The Books

      This was great

  11. Nov 2020
    1. describes the relationship between input and output.

      via differential equations

    1. string out the data 7 values for day1 followed by 7 values for day2 ,etc .

      I believe he means that they concatenate the data, eliminating the gaps...

  12. Oct 2020
    1. numerically evaluate the derivative of a function specified by a computer program

      I understand what they're saying, but one should be careful here not to confuse themselves with numerical differentiation a la finite differnces

    1. work on projects that seem like they'd be cool

      Interesting, so, work for startups to get in the habit of thinking of things that are missing, so that you can make your own startup.

    2. Well

      is PG suggesting that startups should go after niches?

  13. Sep 2020
    1. “I don't really believe in anonymous peer review,” he says. “I think it’s corrupt. It’s all a giant story of somewhat corrupt gaming, I would say. I think it’s sort of inevitable that happens with these very large systems. It’s a pity.”

      i wish he could go into more detail about this statement

    2. hard to expect physicists to comb through hundreds of pages of a new theory out of the blue, with no buildup in the form of papers, seminars and conference presentations

      yes. this is echoing my comment from earlier about him dropping the 448 page preprint

    3. sequestered himself

      seems like a recipe for echo chamber/confirmation bias

    4. To the contrary, Wolfram insists that he was the first to discover that virtually boundless complexity could arise from simple rules in the 1980s. “John von Neumann, he absolutely didn’t see this,” Wolfram says. “John Conway, same thing.”

      he sounds like he's obsessed with being the smartest person in the world or something, or the first one to come up with the theory of everything; it's almost like he's so obsessed with it that he's become oblivious of reality - delusional. He believes his own reality

    5. “It’s this sort of infinitely flexible philosophy where, regardless of what anyone said was true about physics, they could then assert, ‘Oh, yeah, you could graft something like that onto our model,’”

      in a way, sounds like astrology

    6. 448-page preprint paper

      how can he expect people to read that with scrutiny?

    1. μ not


    2. H not

      I'm sorry but this is kind of lazy from the author. Either write H0, \(H_0\) or H naught. H not sounds like you're saying H "not" (negation)

  14. Jul 2020
    1. I learned that I should never ever trust the feeling that I would understand my notes later.

      Your audience is you, but you from the future, so not really you. Write as if you were explaining it to someone else.

    1. “Essentially a friend of mine is head of procurement for NHS London. He has really struggled to get stuff to people in the right timeframe. So essentially what we decided to do was set up a little company and become sales agents for people in the UK who have got stock. And then essentially we would help in linking up the supply chains that are massively disastrous at the moment.”


    2. Singleton said he disclosed his business to superiors in the NHS, in accordance with the rules, and was told there was “unlikely to be a conflict”.

      I can see 3 different explanations for this response:

      • He offered a cut of the profits to the superiors
      • He was unclear / lied about what exactly the business was about
      • The superiors are incompetent
    1. Welcome to my website everyone! Yes, I am a hypothesis user.

  15. Jun 2020
    1. At the end of the day, TikTok’s flawless algorithm will keep people hooked.

      Byte's flawed algorithm, or lack of one, will keep people hooked to TikTok.

    2. Byte, in contrast, flaunts its origins in the United States and emphasizes privacy. “Explore what’s loved by the community, handpicked by our human editors, or just served up at random,” its description reads.

      I feel like this is a naïve approach. If you decide to be in the business of social media, you need to fully commit to the standards of it, at least to start with. It's naïve to believe that you will be able to compete with such an "honest" approach, when everyone else isn't.

    3. Another issue people have with TikTok is its dependence on artificial intelligence (AI). Using AI technology, TikTok can figure out exactly what the user wants to see based on likes, comments and time spent on a video. Based on what the user appears to enjoy, AI can determine the user’s age, location, socioeconomic status and more. This allows the app to push more desired content. However, since the app pushes such specific content to each user, it’s addictive and invasive.

      Okay, but this is what ever successful social media app (facebook, instagram, youtube, twitter) does. It is "the algorithm" and is certainly not an issue unique to TikTok

    4. a host of concerns have sprung up around national security. TikTok has denied all allegations of espionage, but the United States government is still investigating the app.

      It blows my mind that this is the only concern. How are people not concerned about their PERSONAL privacy, other than national security

    5. It is full of bright colors, moving graphics and bumping music.

      Kind of like...TikTok

    1. It's really a shame that vine ended. They left a hole which has unfortunately been filled by the privacy monster that is TikTok. Hopefully Byte will be able to reclaim some of the market but I am skeptical.

    1. the "determiners" and "primary words" are equal

      Wait, what makes the determiners and primary words in the examples below "equal"? They're similar in meaning at best no?

    1. -e-When the plural of the first noun adds an –e-.Die Hundehütte (der Hund -> die Hunde)- er-When the first noun is either masc. or neu. and is pluralized with-er-Der Kindergarten (das Kind ->die Kinder)-n-When the first noun is feminin and is pluralized –en-Der Birnenbaum / the pear tree (die Birne -> die Birnen)

      I mean, this just sounds like you pluralize it in these cases.

    1. If your compound word is supposed to be used for more than one thing each time or usually produces more than one thing or has to do with more than one thing in general, the first noun is to be put in the plural form.

      I need to investigate exactly which plural form (Nominativ | Genitiv | Dativ | Akkusativ) to use here

    2. Firstly you want to find out the main function and focus on that.

      Function: What is the purpose of the word you intend to create?

    3. You can make up cool German words yourself, easily.

      If I can, then so can an algorithm...interesting.

    1. Google has freezed hiring for ML researchers. Uber laid off the research half of their AI team.

      woah, wait a minute - source? Could this not be covid related?

    1. second step in calculating self-attention is to calculate a score. Say we’re calculating the self-attention for the first word in this example, “Thinking”. We need to score each word of the input sentence against this word. The score determines how much focus to place on other parts of the input sentence as we encode a word at a certain position.

      The output is a vector of scores, each score corresponding to a a word in the sequence

    2. What does “it” in this sentence refer to? Is it referring to the street or to the animal? It’s a simple question to a human, but not as simple to an algorithm.

      In the italian education system, we have something called "Analisi Logica", basically a logical analysis of sentences. This sounds like a very basic version of that :')

    3. each of them flows through each of the two layers of the encoder

      each of them flows through each of the two layers of EACH encoder, right?

    1. It made it challenging for the models to deal with long sentences.

      This is similar to autoencoders struggling with producing high-resolution imagery because of the compression that happens in the latent space, right?

    1. it’s paramount that each person — regardless of being for or against the result — individually pledges support out loud in the meeting. “Go around the room and ask each one of them to support the decision one at a time,” he says. “Commitment meetings are really important, because when you pledge to support a decision in the presence of your peers, you're much more likely to support it.

      Damn. Not gonna lie this sounds a little bit culty

    2. think of the last time you were handed a decision that someone else made but for which you had to execute and usher to success. How did that feel? I’d guess it made you feel frustrated, powerless or disengaged. We want to avoid that. That’s why the decision maker is both accountable and responsible. It’s more fulfilling and empowering.

      Plus, if you're engaged with something, it makes it no longer a chore, and more of pleasure to do the work related to it.

    3. people understand the when, and the ‘why’ of the ‘when.’”

      This is so important. I hate being told (not) to do something without knowing why I'm (not) meant to do it. Transparency is key and a lack of it can lead to a lack of motivation for me. Adding meaning to deadlines.

    4. forward-thinking companies still gravitate to consensus as the way to make decisions. It turns out that for important, difficult choices, that approach is often ineffective and impractical.

      kind of like how big FOSS projects have a BDFL for ultimate decision-making

    1. they had already sent hundreds of thousands of people to this country to spread that virus

      Hard to say if they explicitly "sent" these people to spread the virus. I do agree that it is heinous that they didn't disclose information about the virus for the first month or so, and that might've been on purpose

    1. it seems that word-level models work better than character-level models

      Interesting, if you think about it, both when we as humans read and write, we think in terms of words or even phrases, rather than characters. Unless we're unsure how to spell something, the characters are a secondary thought. I wonder if this is at all related to the fact that word-level models seem to work better than character-level models.

    2. As you can see above, sometimes the model tries to generate latex diagrams, but clearly it hasn’t really figured them out.

      I don't think anyone has figured latex diagrams (tikz) out :')

    3. Antichrist

      uhhh should we be worried

    1. We only forget when we’re going to input something in its place. We only input new values to the state when we forget something older.

      seems like a decision aiming for efficiency

    2. outputs a number between 000 and 111 for each number in the cell state Ct−1Ct−1C_{t-1}

      remember, each line represents a vector.

    1. crowdfunding campaign on Sci-Hub to buy additional drives

      wow wait a minute, does that mean the sci-hub database is a server run locally by alexandra?

    2. I did not tell Science how credentials were donated: either voluntarily or not.

      Hahahah - Surely a donation can't be involuntary

    1. Can strongly say that this is the case for myself aswell. Another detail that wfh proponents seem to forget.

  16. May 2020
    1. Hypothesis/Brave login now works seamlessly there, and will on any other sites I unblock.

      I'm ngl I'm not a huge fan of either of the proposed solutions for this. Just because I'm happy to allow third party cookies provided by hypothesis.is on example.com, doesn't mean I'm happy to allow all the other third party cookies on example.com, which is what the more conservative of the two solutions suggests.

      Maybe there's some form of whitelisting we can do so that Brave automatically unblocks all hypothesis.is third party cookies, regardless of URL?