wyager

FYI it looks like some of your comments are getting auto-flagged by the HN moderation system and marked as dead

99.9% of normies have a router NATing all their traffic

It takes active effort to expose a camera publicly

Very exited about WASM/WCM as a portable format for capability-secure applications.

I had a spec file sitting around for an OS project idea I had, where the kernel would just be the WASM compiler + a few small shim drivers, and everything else (including e.g. PCIe device drivers) would be WASM modules with WIT interface specs. I handed the spec off to Fable and it seems to have made a working proof-of-concept. Has a maximally-WASM OS running on browser/QEMU/Orange Pi. https://eo9.org

I'm curious if people have a good story for why WASI will succeed where Java failed

The entire US lunar effort cost only $330B in current USD, commensurate with the amount AI companies have raised on private markets alone, and there was also a cold war

> Solstice (100k NVIDIA chips, currently spec'd to be Vera Rubins) in the next five years

Is this supposed to be impressive? Five years for the equivalent of, what, Colossus 1? What a joke

EU countries continuing to ensure the conditions for their future economic competitivity

> i wish govt would fund these labs and make it free and opensource.

It would be impossible for the govt to allocate this much capital towards such a moonshot, and even if they could, they would do it in a way that would get 90% frittered away to fraud and waste

That would be a valid explanation if they hadn't totally oversold and underdelivered on "apple intelligence". In reality, this explanation is just cope

Apple has completely dropped the ball on every single detail of AI rollout for the last 5 years - why do you think they will suddenly stop now? My prior is that the new siri stuff is just as vaporware as the previous "apple intelligence" rollout

Isn't the entire point of this order to prevent filling low-paying jobs with cheap foreign labor, in order to increase demand for domestic labor? "Rural district schoolteacher" sounds like exactly the kind of job where the H1B program has very low public support

> By the end of next year you’ll be running most of your AI on device.

I expect I'll probably keep paying for whatever badass high IQ model is running on inference servers at that point

> If you believe it is dangerous, you should be dedicating yourself to STOPPING others from making it

I don't think anyone has been more successful in promulgating AI safety

There are groups like MIRI who tried what you're sugesting, where they make no AI and just push for AI regs, and they have been relatively much less successful

> It is not illegal, at this point in time, for teachers to oppose AI for political reasons.

No, but that would make it a "political coalition thing", which is why I asked

Yes the a priori most likely reason for the TU to be "against AI" is political. If you know much about TUs this is pretty obvious

> Yeah, I work at a CSU and the Teacher's union is against AI.

Is this a political coalition thing or is there a real teacher-related reason they don't like it?

> Where's the trillions, or hundreds of billions worth in improved quality of life?

Starlink and Claude are both awesome and huge QoL improvements for me!

You don't need to make this judgement; it's fixed by the colorspace you're working in.

First, figure out what colorspace the processing needs to happen in. Usually this is linear RGB.

Then, figure out what OETF and EOTF your input/output format use. This will be something like PQ or HLG. This will exactly specify the meaning of each integer value.

This fixes the choice of representation and conversion.

> there will be hardware capable of running frontier models

The current frontier? Sure. The frontier then? No - obviously that frontier is going to keep consuming available datacenter compute capacity, which will be better

This description falls apart for two reasons

1. It only accurately describes pre-training 2. It ignores the existence of generalization

Next token prediction is just a training task, not "what the model does internally" in any meaningful sense

> They are a highly sophisticated statistical model designed to mimic the distribution of programming

Are we really still doing this?

You're already writing Rust in a very different style if you're writing the type of code that gracefully handles allocation failure. It's to Rust's immense credit that this type of coding is actually fairly well-supported (unlike in Go), but you're already a bit off the beaten path for stuff like error handling.

It seems bizarre to model the end goal of this research as "grow more birds". Obviously that's not the point

Every time I've ever seen a tech worker's union, it's always some sort of political experiment rather than legitimately advocating for the interests of the workers it nominally aims to represent. E.g. the Google AWU-CWA union just did a bunch of political stunt stuff, no salary negotiation or anything useful to the modal Google worker.

Is it more moral to allocate goods to the people who are willing to pay the most for them, or to the people who have good scalping bots/happen to get lucky

Mercury has been awesome, I've been using them for my business account for years and recently started using them for personal as well. I didn't know they used Haskell until well after I started using them, but it definitely tracks. The quality of their exposed software surface is at least a couple stddev above median.

There are two countervailing effects when you choose a more theoretically advanced programming language. On the one hand, your hiring pool shrinks. On the other hand, the quality of the remaining hiring pool goes way up, which acts as an excellent recruiting filter (for both employer and employee). Jane Street made a similar play with OCaml.

> You can do this in basically any language.

You can do it in Assembly. That doesn't mean it's cost effective.

The final frontier of display tech (as far as being able to elicit any physiologically possible eye response) is a pair of tunable lasers. You really can't go much farther than that for emissive displays! We're almost saturated (no pun intended) on useful resolution, so I expect color to be the next area of focus.

The bottleneck isn't research funding, it's getting past the FDA