pgorczak

While this font looks kind of weird up close, I found it great for creating plots. It’s my default choice in matplotlib rcParams.

It also syncs config and plugins

This reminded me of Elixir’s GenStage at first glance. Now I wonder how the underlying libs OTP and core.async relate conceptually and implementation wise.

I guess the issue is having to store the intermediate state somewhere. It’s true for the PID example that numerical integration is easier to compute - if you look at the comment with the closed form solution, you need trigonometric and exponential functions to evaluate it. It’s kind of fascinating that the iterative method approximates the same thing with just addition and multiplication.

> If a company decides to lay off, for instance, 40 employees, German law doesn’t prevent this.

At least this part is partially wrong. There is an entire law about how lay offs are only allowed if they are “socially justified” with definitions of acceptable circumstances. An employer can not fire you “at will” in Germany.

No, they could literally not send on the frequencies they listen on in case of FDD. Enabling this would require extra radio hardware. Also there would need to be a some kind of encryption key exchange between devices which is not needed in the centralized setup. They could not easily route to one another without adding extra stuff.

They are not entirely orthogonal. You need a central base station with scheduling, high dynamic range and power control to maximize performance in an OFDM system. In addition, most cells used to use frequency division multiplexing meaning that the base station and phones send and receive on different frequencies. So lack of point to point capabilities can at least in part be explained by the design goal of optimizing for throughput and user density.

You can cast video e.g. from the QuickTime app or a <video> tag in the browser too which won’t just mirror your screen. In fact the cast video won’t even show on your device’s screen but only on the receiver in that case.

“That means the conventional predictions are largely inference—and worse, they result in unquantified uncertainty.”

Wild claim given the fact that Gaussian process regression / Kriging was invented in the 1960s in geoscience to do exactly what the article claims only their models do: “quantify uncertainty, which in turn guides our data collection, as the most uncertain rocks often represent the most valuable ones to sample”

That’s one way to frame it. The other would be the candidate being able to understand that the “apart from money” part is implied in the question and the answer given the social context. This makes the money answer go from straightforward and honest to blunt and cringeworthy.

Maybe “an automated thing that moves to do stuff”. Generally there are manipulators and mobile robots like vacuums. An interesting edge case would be a CNC machine which has degrees of freedom similar to a pick and place robot but is seen as a whole static thing with moving parts. Also if the mobile platform transports people it’s usually not called robot.

The German federal government alone will pay 1.28 billion Euros to Microsoft until 2025 [1] so I’m pretty sure they do care about their government contracts.

[1] https://amp.zdf.de/nachrichten/politik/deutschland/it-open-s...

I don’t think commonly used LLM architectures have internal state that carries over between inference steps, so shouldn’t that be none? Unless you mean the previously generated tokens up to the context limit which is well defined.

You could also use a single receiver with a small antenna array (GPS wavelength is around 20 cm) to estimate the angle of arrival of the incoming signals.

I’m listening to some podcasts published through Acast and 95% of the time, my injected ads are about how I don’t have to listen to said ads if I buy Amazon prime.

I appreciate the build step in setups like Vite/Vue because the development server can automatically and accurately hotpatch the application when I make changes. I don’t think you’d want to change the standards to couple DOM and js in the way that makes this possible in Vue but it’s an iteration speed improvement nonetheless.

The subscriptions page can’t be filtered to show some subset/group of channels. I use PocketTube as a plugin-based alternative.

FWIW regex are also a DSL for finite state machines acting on strings, so in that light it’s a straightforward jump

I had similar issues with the Dask scheduler a few months ago. The docs say it encourages depth first behavior in the computation graph but in my case it kept running out of memory on a large ETL task by first trying to load all the input files into memory before moving on to the next stage.

I’m not familiar with the Python lib but it could be waiting for streams to acknowledge each message reception/persistence before sending the next one. Some clients allow transactions to run in parallel e.g. with futures.

From a European perspective: regulation is just catching up. Risk assessment and safety frameworks have been established and (non-autonomous) beyond line of sight operations are slowly becoming feasible to do on a regular basis. Operating many drones (+ manned vehicles) in the same airspace needs a traffic management solution (U-Space) which is also seeing some initial trial runs now.

If it’s just off the coast as in “territorial waters”, there’s the concept of innocent passage in international law (https://en.wikipedia.org/wiki/Innocent_passage) - although the US are not part of this UN convention. If the criteria for innocent passage are not met, “the coastal State may take the necessary steps in its territorial sea to prevent passage”.

Edit: I guess in more tricky situations e.g. the ship isn’t obviously armed or spying, it comes down to questions similar to probable cause.

Wouldn’t {excuses} be reverse engineering all of Lightroom’s adjustment tools to be able to develop the same picture from a given pair of raw and sidecar files?

Another possibly related difference is the training managers get. There are military careers where you know on day one which leadership responsibilities you will take several years down the line. That’s a stark contrast to starting out as an engineer and then transitioning as described in the blog post.

> Because simulating damped springs requires calls to potentially expensive trigonometric and exponential functions…

It doesn’t though if you go numerical and simulate with Euler or RK4. That approach also fits well to the application in a video game with discrete time steps. I wonder why the author chose to go with the closed form solutions.

> There is an observer and that which is observed, the thought. I am the observer, not the observed.

Why couldn’t you be both? The duality of physical and metaphysical self is beautifully illustrated in the Tree of Jiva and Atman from a few thousand years ago: https://en.m.wikipedia.org/wiki/Tree_of_Jiva_and_Atman

Yeah the handedness of a 3D coordinate system is a pretty arbitrary low level choice. You don’t even need to talk to aliens, just try implementing some geometric algorithm in Unity which uses a left-handed system.

Yes, for example there is srsue in https://github.com/srsran/srsRAN_4G (the predecessor of the stack used in TFA)

It can be used at small scale like home wifi (small cell / femtocell) and it would probably be on par with regular wifi. Cellular networks don’t really stand out in home-sized few-user deployments though because they’re designed for serving large areas or high user densities.

Nonlinear systems and control could be something that’s useful to get into regarding legged robots or exoskeletons. It’s not covered in depth by many general robotics textbooks but there’s a great lecture including fantastic interactive notebook material at http://underactuated.mit.edu/