blt

73 ms·

841.
by blt 8y ago
The NYC subway system is not autonomous. Every train has a driver. Autonomous subways do exist though.
842.
by blt 8y ago
This proof is missing a step showing that the problem exemplified by the simple two-node system applies to all system topologies. This website only proves: "there exists a system for which CAP is impossible." I do not think the
843.
by blt 8y ago
no marketing budget.
844.
by blt 8y ago
There's a lot of counterfeit stuff on Amazon. I ordered a Pulltaps corkscrew and the chrome-plated metal part [1] was instead made of crudely bent sheet metal that didn't look remotely authentic. After that I mainly use Amazon for
845.
by blt 8y ago
Yes I agree that the language semantics and the philosophical motivation are more important. But there is a reason academic papers have abstracts. The core ideas should be summarized crisply at the beginning of a technical document. A few s
846.
by blt 8y ago
show a code sample before your whole manifesto
847.
by blt 8y ago
I don't really understand this post. It reads like a neutral description of differences between kv/content addressed storage and file systems, but then concludes that you shouldn't use file systems for distributed systems. Fu
848.
by blt 8y ago
Yeah I read that proof as more of a reassurance than an emphasized part of the paper. If the method does not even work on convex problems, it's highly suspicious as a tool for nonconvex problems.
849.
by blt 8y ago
IMO, the main benefit of ROS is the message definitions. Having exactly one definition of "point cloud" that everybody accepts means everyone's code will be compatible. That isn't normally the case in library ecosystems.
850.
by blt 8y ago
There is no mechanism in this paper (or in a standard VAE, or GAN) to encourage that a single human-understandable semantic quantity should be captured in a single dimension of the latent code. So, in general, it won't happen. "Bl
851.
by blt 8y ago
Why did you choose an unsupervised method to solve a classification problem?
852.
by blt 8y ago
Yes there are certainly situations where the back button becomes nonsensical. I'm talking more about things like a 30-image slideshow generating a trail of 30 URLs in the history, or (mainly) outright "dark pattern" sites tha
853.
by blt 8y ago
As much as I hate bloated page sizes, breaking the back button is the absolute worst thing a website can do. It is a disgusting violation of the user's experience. It's like locking the door behind a customer when they enter your
854.
by blt 8y ago
I agree that statistical ML is only good for interpolation, but I also feel that the set of human behaviors that are reducible to interpolation is larger than we realize. Even high order human behaviors like academic research or art critici
855.
by blt 8y ago
It's visible in any Windows screenshot
856.
by blt 8y ago
Subpixel AA was never about CRTs; they do not have a consistent mapping between logical pixels and physical rgb holes in the shadow mask. It was always about LCDs
857.
by blt 8y ago
I never upgraded past Mavericks. It seems that the OS has been going downhill. Thinking about installing Ubuntu on my MacBook.
858.
by blt 8y ago
What is special / unusual about the splines? ("Check out those cut splines in the middle!") It looks like normal spline drive to me.
859.
by blt 8y ago
The new desktop site sucks too. The old one was great, information dense, no bullshit.
860.
by blt 8y ago
Your points are good. I did not intend to suggest that software developers should make the kind of decisions that lead to "mangled and indecipherable set of 'critical sections'". More that developers who understand their
861.
by blt 8y ago
It sounds like the services you are describing basically take the form of "evaluate a pure function on an input". I am not going to argue with your stance there. I am more referring to programs that run interactively with some lar
862.
by blt 8y ago
I always call it "death by a thousand cuts".
863.
by blt 8y ago
> But if performance work has been neglected for most of the product development cycle, chances are that when you finally fire up the profiler it’ll be a uniformly slow mess with deep systemic issues. Totally true, and I have observed
864.
by blt 8y ago
PC user here, but yeah it was magic. Installing a new ISA or PCI card was like entering a new world. I remember my first sound card, modem, and 3d accelerator vividly. Upgrading modern PCs is still fun, but it isn't nearly the same as
865.
by blt 8y ago
I like this model, but it puts a huge workload on the TA to help with questions about the homework, since the students don't get to hear the professor explain the material until after.
866.
by blt 8y ago
I agree, it is really unfortunate that even people within the software development profession take these issues so lightly.
867.
by blt 8y ago
It is different though. The derivative of the L2 loss function w.r.t. the linear regression parameters is a "flat" function that is easy to derive by hand. With neutral networks you have deeply nested vector-valued functions. If y
868.
by blt 8y ago
My 5 year old MBP is more than fine, it's great, but I am worried about what I will be able to buy to replace it. Eventually it will break; if not then someday I'll want something with more computing power.
869.
by blt 8y ago
Raw null terminated strings are just a bad idea. `std::string` and the bafflingly just-introduced `std::string_view` are the right way to handle strings. We can spare the bytes now.
870.
by blt 8y ago
I don't really think that characterization is fair, for example GANs, there is no data set of correct input output pairs for the function that is learned.

More ›