xipix

Why is it that every audio waveform renderer looks super aliased? I also recently made a fast audio waveform renderer using the best bits of Wasm and WebGL to create an animatable waveform that looks a bit less 1990s: https://play.google.com/store/apps/details?id=com.parabolare...

Happy to open/share the code, just haven't had time.

PS (yours didn't work for me in Safari)

That's data size, not code. There's no fundamental reason that a program that can smoothly render unicode at 4k needs a GB download when kB could suffice.

Quite simply: it lets you build portable apps with near-native performance. Same source for both your desktop browser and your phone. I really love using it to make audio apps like this one: https://bungee.parabolaresearch.com/bungee-web-demo, also it's great for other real time rendering / visualisation in the browser.

There's also a web audio stretcher: https://bungee-audio.pages.dev/bungee-web-demo

Yes, also newer iGPUs (11th Gen onwards) can use same software

Best to ignore such messages. Call Google's bluff. I do, my apps keep paying out.

Nice. I'd love to see how this changes when you have SIMD and multiple cosines to compute in parallel. Also when you have to compute sine and cosine simultaneously which is often the case, and then you may be more interested in polar error than cartesian error.

A large system, so much higher chance of something breaking.

What happens if it loses a node or a link? Or some memory becomes unreliable? This thing needs some sophisticated fault tolerance.

I always thought this would be a cool thing to do if I ran Evil Corp...

A = <totally random bits>

B = <personal data> XOR A

Store A in USA

Store B in EU

The data is not stored in EU, and it's not stored in USA either. It's not stored elsewhere. But Evil Corp still has it!

It's so funny to see Apple's "combined memory architecture" praised. When Intel did this a decade+ ago with its integrated GPUs, it was roundly slated.

Why are people still using mono fonts in 2023? There are very few reasons not to love proportional fonts in your IDE.

Strategies for dealing with ageism would be useful indeed. But otherwise, besides knowing shortcuts to increase zoom/font size for changing eyesight, age brings no special needs.

Perhaps (a) there do exist bigger models that weren't better or (b) this model isn't better than somewhat smaller ones. Perhaps the CEO has seen diminishing returns.

Something not mentioned here is that software itself is rarely the subject of a patent. Instead, it is a "computer system running software". This, I think, is a trick to get around software itself not being patentable in some jurisdictions.

I really like this proposal but it lays bare the fact that the invention lies entirely within the software itself. When the patent industry has been pretending for years that the "invention" is a hardware thing that could, optionally, use software.

yup and, as ever, Apple definitely weren't the first with this.

Matrix math units are a smart use of silicon for many ML applications, true. But Von Neumann compute isn't going away. Neither are fixed function modules for shading, ray tracing and massively parallel GPGPU type applications. Also not video coding.

My point was that if training could emit an efficient video codec as a network of logic gates, rather than as a dense array of neural network weights, it might just produce something practical for video playback on mobile devices and video processing at hyperscale.

The thing about video compression is that it's already very compute intensive. Most devices have dedicated hardware cores for encoding and decoding in order to save power. They work exclusively with integer values, chiefly unsigned 8 and 10 bit values with sequential and block memory transactions.

Humans not only design for compression quality, we also really focus on cost. And we design direct to what is cheap in hardware: integer adds, shifts, binary logic.

I do believe a different paradigm of ML is necessary for this particular application. The neural primitives used need to be more closely related to transistors than to matrix multiplications.

"Technological advancements" is a term that is open to interpretation. To many, it means advancements in terms of reducing CO2 emissions.

Prices of the chips may be rising, but power usage is falling and in the end, that's often the bigger cost.

Better than RTO, an annual or six monthly team meet up. Its enough to catalyse new and boost existing relationships without the pointless grind of daily commutes.

The implicit point here is that speed limits should be based upon safety, or they should be abolished. This article fails to mention that there are important reasons for speed limits other than safety. For example, noise, pollution and climate. Mobility and freedom of those without a car, especially the young and very old. Speed limits are a tool that can be used to help boost the health, fitness and quality of life for people who live and work near roads (that's pretty much all of us).

Edible mushrooms do benefit from being eaten. Their total biomass is today much greater than if they'd remained naturally-selected, growing in the wild. That said, the evolutionary feedback loop of incrementally more intelligent humans farming and eating them does seem tenuous.