HNHacker News
TopNewBestAskShowJobs

bnegreve

2,040 karma · joined January 4, 2011

submissionscomments
bnegreve··on Big O Notation – Explained as easily as possible
> often enough it's not really clear what a step is

The steps are basic CPU operations such as load/store and basic arithmetic operations performed on fixed width type. I.e. things a CPU can do.

> And there we have it: with a fixed bit-width, this becomes a constant term.

It makes sense because that is how the CPU works: arithmetic operations on fixed width types are performed in constant time in the CPU ALU. Adding anything below 2^32 requires the same number of cylces.

It is only confusing when you confuse basic CPU operations with basic python instructions (an arbitrary precision addition is certainly not a simple operation from the cpu perspective)

bnegreve··on Machine Learning: The Great Stagnation
Discrete convolutions are a special case of matrix multiplication in the sense that a discrete convolution can be formulated as the product between a Toeplitz matrix and the input signal.

[1] https://en.wikipedia.org/wiki/Toeplitz_matrix#Discrete_convo...

bnegreve··on Drone Footage of Arecibo Observatory Collapse
The entertainment industry maintains the US hegemony just as much as the military.
bnegreve··on Mmhmm
>I don’t need an “argument” to charge a specific price or business model.

I disagree, you need to sell your business model just as much as you need to sell your product.

As a customer,

1. I need to know whether this is a fair pricing for the service, because I need to decide whether I should pay for it, or search (or wait) for an alternative. If it is overpriced, a cheaper alternative is likely to appear soon, and it is probably better to wait.

2. I want don't want to encourage business models that don't fit my usage, to avoid the proliferation of such business models across the industry.

bnegreve··on Apple Silicon M1: Black Magic Fuckery
> Programs often reserve a lot more memory than they actually use (zero hit in performance) so memory stats are misleading, and the OS is really good at swapping memory not touched in a while to the SSD without you noticing.

The stats are absolutely reliable because no physical memory page is allocated until it is actually used to store something. So allocating a large chunk of unused memory wouldn't show in the (physical) memory usage stat.

bnegreve··on “To be safe allocate 5 bytes more than you need”
> There is a big difference: in science you observe and try to understand something that already exists, the universe.

An important part of CS research is empirical. It's not because you have designed something yourself, that you don't need to observe it to understand it, or that you can answer any question about this thing analytically.

For example, the folks from high performance computing conduct loads of experiments to observe and understand which CPU design work best and why. Deep Learning people also run lots of experiments understand which model is easier to train and corroborate theoretical results from learning theory. etc.

bnegreve··on “To be safe allocate 5 bytes more than you need”
> However, it is entirely implementation defined what that 1 refers to.

Not entirely, there is a vague description of what a byte should be able to represent.

The C and C++ programming languages define byte as an "addressable unit of data storage large enough to hold any member of the basic character set of the execution environment"

https://en.wikipedia.org/wiki/Byte

bnegreve··on Apple's apps bypass firewalls like LittleSnitch and LuLu on macOS Big Sur
Oh come on! It is not because I spend most of my life inside a terminal that I don't prefer simple things over complicated ones.

Technical doesn't mean "unnecessarily complicated", it means "rich, expressive and built for users that are willing to spend some time to learn" (at least it should)

bnegreve··on Starting a physical product company? You’re gonna need a lawyer
Actually, the video suggests that the complicated part is not the baking process itself, but the handling of thousands of blocks of different types and colors.
bnegreve··on Animation of how bridges were built in Central Europe in the Middle Ages [video]
> As humans, we have a tendency to be pessimistic

Maybe, but it is not because we tend to be overly pessimistic in general, that serious problems cannot happen to us. Entire civilizations have diapered in the past, so we should remain vigilant.

After all, another bias that humans have, is a strong tendency to believe that what is true now, will remain true forever.

bnegreve··on Reinforcement learning is supervised learning on optimized data
The way I understand it, the two subproblems are supervised in the sense that they are trained using data sampled from a fixed distribution, instead of data sampled from a distribution that changes as you update your model, as it is usually the case in RL. This makes the training more stable.
bnegreve··on Nvidia Uses AI to Slash Bandwidth on Video Calls
> Instead, the points on the face are sent across the net and the face is actually generated.

This is not fundamentally different from other lousy compression algorithms.

The bits that are missing are reconstructed using a bias built in the compression algorithm. Deep learning based algorithms simply have a more realistic bias, so the artifacts that introduced to replace the missing bits are less noticeable.

bnegreve··on Computer Scientists Break Traveling Salesperson Record
Apparently, heuristics are still better than DL-based solutions for the TSP

From the paper you linked:

heuristics based solutions are still superior to deep learning based solvers regarding accuracy and execution time for larger problem sizes.

bnegreve··on XB-1 Supersonic Rollout
For downvoters: it's a fine project, but I don't think they can fly planes at higher speed, lower cost, and lower carbon emissions, without a major technology shift. It is hardly controversial.

Most likely: they do care, but it is not their main priority.

bnegreve··on XB-1 Supersonic Rollout
Neutral would be good, but I doubt that they are serious about this.

Otherwise, they wouldn't build their plane around gas turbines, which create trust from burning carbon compounds.

bnegreve··on XB-1 Supersonic Rollout
Actually, I find worrying that you can build a plane that creates trust from breaking long carbon chains into CO2 and, at the same time, publicly announce that it will be "carbon neutral" some day.
bnegreve··on Honda quits F1, invests in carbon-free tech instead
Because they want to reduce them more.
bnegreve··on Memristor Breakthrough: First Single Device to Act Like a Neuron
I don't think Relus are the proplem. Relus are good, but artificial neural networks can work with sigmoid activation fonctions instead. The neurones in an ANN with sigmoids are somewhat close to what is described in this article.

The main problem is that the neurons in this article are not trainable (you can't update the weights dynamically) and not differentiable (you can't use the usual gradient-based training procedures to update the weights). They could be used for inference only though.

bnegreve··on Tracking users via CSS
> I do wonder why you need CSS analytics rather than just using the server log from the request for the HTML file.

From the article:

Lots of automated traffic on the web, bots, crawlers and scrapers. So if there is a way that can remove most of the automated traffic without loading any JS, is that a win?

   body:hover {
        background-image: url("https://underjord.io/you-was-tracked.png");
    }
[...] This has a certain elegance because it actually requires mouse interaction.
bnegreve··on Why You Will Marry the Wrong Person (2016)
> The statistics about marriage failures are misleading. They are biased by the fact that people who divorce tend to have multiple marriages.

They are also biased because many unhappy couples won't divorce for various reasons.

bnegreve··on New academic journal only publishes 'unsurprising' research rejected by others
> then I find papers and people saying what we did would never work.

Seriously? What kind of scientific paper makes such claims?

bnegreve··on 2.5M Medical Records Leaked by AI Company
I'm reposting a dead comment by photon12 which was probably killed for using too many profanities. I agree with this part:

> It's a collective action problem that needs collective action. Individual incentive changes aren't going to cut it. Unfortunately many thought leaders in the industry have tried to build a culture of suggesting that the solution to collective action problems is individual iteration.

bnegreve··on A new design ship in the Philippines is partly powered by wave energy
> It’s honestly a shame that we didn’t keep pursuing sail technology.

We did, modern sail boats are much faster than they used to be and can almost sail against the wind (~30° from the wind direction). Fastest sail boats can cross the Atlantic in few days now. (speed record: 3days15h [1])

https://en.wikipedia.org/wiki/Transatlantic_sailing_record

bnegreve··on Don't close your MacBook with a cover over the camera
> Every laptop Apple has manufactured in the last ten years has an LED connected to the same circuit which powers up the camera.

At the very least, I need a citation or an official statement. Because clearly, this has not always been the case [1]:

We describe how to disable the LED on a class of Apple internal iSight webcams used in some versions of MacBook laptops and iMac desktops.

[..] our investigation of the iSight revealed that it is designed around a microprocessor and a separate image sensor with an indicator LED sitting between them such that whenever the image sensor is transmitting images to the microcontroller, a hardware interlock illuminates the LED. We show how to reprogram the microcontroller with arbitrary, new firmware. This in turn enables us to reconfigure the image sensor, allowing us to bypass the hardware interlock and disable the LED.

[..] iSight webcam [was] found in previous generation Apple products including the iMac G5 and early Intel-based iMacs,MacBooks, and MacBook Pros until roughly 2008

Whatever reason Apple had to design the camera system this way back in 2008, is probably still a valid reason (cost, hardware simplicity, spacial constraints etc.). It means Apple and others have incentives to build camera systems that are easier to compromise. It's enough for me to worry.

[1] https://jscholarship.library.jhu.edu/bitstream/handle/1774.2...

bnegreve··on Has GitHub been down more since its acquisition by Microsoft?
That's pretty much their conclusion

> [..] GitHub has been down more since the acquisition by Microsoft. But that could be all a part of coordinated effort to be more transparent about their service status, an effort that should be applauded.

bnegreve··on MIT Ends Elsevier Negotiations
Also:

It's difficult to know whether a publication is open or closed when you download it from a research institute or a university (access is automatically granted based on the ip address).

You only realize how much it cost when you try to download it from home.

bnegreve··on Show HN: Quake 1 movement physics reinforcement learning project
> It wouldn't be difficult at all if the optimal running technique was known before hand;

You're right, but this is a simulated environment. It means the physics in it are driven by a small set of known (and most likely deterministic) rules.

So I wonder if it would it be possible to analytically determine what is the best possible running strategy.

bnegreve··on Show HN: Quake 1 movement physics reinforcement learning project
How difficult would it be to encode the optimal running technique as a rule instead of learning it with RL?
bnegreve··on Why does `True == False is False` evaluate to False in Python? (2013)
I know how to make this work, but a bug like this would be hard to spot.
bnegreve··on Why does `True == False is False` evaluate to False in Python? (2013)
I find something like this totally plausible and yet it is completely wrong.

    >>> def check_parity(x, expect_odd):
    ...     odds = [ 1, 3, 5 ]
    ...     if x in odds == expect_odd :
    ...             print("ok")
    ...     else:
    ...             print("error")
    ... 
    >>> check_parity(5, True)
    error
Crazy!
← PreviousPage 4 of 25Next →