Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
torginus
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
8 ms
·
151.
▲
by
torginus
2mo ago
I don't think its an inherent quality - a lot of older models had a much more natural feel to them. I guess it has to do with the low temperature (low randomness in final word choce - something all vendors seem to have converged on fo
152.
▲
by
torginus
2mo ago
Most of what you write about falls outside of software engineering or computing, with the exception of Rust - which is something that you could argue is a genuine improvement over SOTA.
153.
▲
by
torginus
2mo ago
I think there's the problem of pretending to be something that one is not - when people have always tried to associate themselves and the things they did with the cool 'in-group' of genuinely inspiring people, to the point wh
154.
▲
by
torginus
2mo ago
I think the parallels with AI art are uncanny - nowadays people with zero artistic skill, vision or effort invested are capable of creating stuff that would've previously taken a master artist a solid week of work - doesn't mean a
155.
▲
by
torginus
2mo ago
I had a validation script in the background that checked the generated graph - constraint based PCG often fails (it gets into an unsatisfiable state) anyways, and you have to (partially) retry it.
156.
▲
by
torginus
2mo ago
Imo 'AI ethics' in this context was code for 'how to make our platform advertiser friendly'
157.
▲
by
torginus
2mo ago
I think this is the solution to a lot of AI 'alignment' issues. We have laws, and LLMs should NEVER break them unless the user states its fine with some qualifying condition. Yes, every country has a different legal system, but I
158.
▲
by
torginus
2mo ago
LLMs are a PCG system. This is one of my pet peeves, I've been working on a procedural generation constraint system for quite a while (way before LLMs) as a hobby, that came with a domain specific language, where you could specify cons
159.
▲
by
torginus
2mo ago
I get that, but a lot of C interpreters use switch, for example Lua: https://www.lua.org/source/5.5/lvm.c.html#vmdispatch
160.
▲
by
torginus
2mo ago
The first one, in that it expresses intent, not implementation, which is why most languages you have `filter` and `map` methods that abstract over recursion, and most people use `foreach` instead of `for(int i=0;...` in procedural ones
161.
▲
by
torginus
2mo ago
Bet none of them have a chief ethicist tho.
162.
▲
by
torginus
2mo ago
What practical patterns are enabled by TCO in C? My impression is that every tail call can written as a loop much more naturally. Tail calls are important in functional languages where you don't have mutable loop variables. And imo the
163.
▲
by
torginus
2mo ago
I recommend you install https://learn.microsoft.com/en-us/sysinternals/downloads/pro... And take a look what DLLs have been loaded into your explorer.exe. Look for non Microsoft-signed ones.
164.
▲
by
torginus
2mo ago
Deferred renderers work differently from compositors. They still build up the entire displayed frame in screen space, except they dont write into a color buffer like directly, but produce a bunch or intermediate buffers, called G-buffers.
165.
▲
by
torginus
2mo ago
I'm pretty sure you have a plugin installed for Explorer that makes it crash - while Explorer is far from a perfect piece of software, its not even close to this bad, but it's pretty easy to extend it with COM objects that then ru
166.
▲
by
torginus
2mo ago
One thing shared between old software rendered desktops and modern videogames is that they shunned these intermediate 'composited' screens. On the desktop, there used to be the screen, and basically windows used to draw on top of
167.
▲
by
torginus
2mo ago
Counterpoint: Nautilus is both slower and less stable than Windows Explorer on large folders, even though the Linux filesystem is much faster. These super-indexed desktop linux search functions are also dog slow, while `grep` and `find` in
168.
▲
by
torginus
2mo ago
These apps have codebases going back to the 2000s when having 256MB ram total was a big deal, yet these apps could contend with that (while offering pretty much identical functionality). I remember one of the big arguments for GTK against Q
169.
▲
by
torginus
2mo ago
Haven't read the article, but multiple times during my life I have had to research how to do things in a domain, and have come up against academia (both literature and people), who seemed to have formed an opinion on how to approach ce
170.
▲
by
torginus
2mo ago
> Nobody _wants_ it, but in practice if all cars/TVs/smartphones break every two week. Certainly not true for me. One example is Linux. I've been sticking with Ubuntu LTS, and delaying upgrades for as long as I can get awa
171.
▲
by
torginus
2mo ago
This isn't true. I don't think you or me would want a car/TV/smartphone that's cheap and flashy but breaks every two weeks. Sometimes people are strapped for cash that's why they cheap out on stuff, or really a
172.
▲
by
torginus
2mo ago
They do. Most economically viable software already exists today, most people's job is to make .1% improvements, and features on top. For this work it's obviously pivotal that any change should not worsen the existing software, as
173.
▲
by
torginus
2mo ago
In that case they should clearly label that this is a new benchmark.
174.
▲
by
torginus
2mo ago
Are you saying they are... tasteless?
175.
▲
by
torginus
2mo ago
I think the choices of these workloads are deliberate, considering this is a large core count CPU linked to a FP-monster GPU with a high speed, low latency datalink. The former implies per core memory bandwidth is probably not great, meanin
176.
▲
by
torginus
2mo ago
There's no honor in being stupid, yet I must admit I am stupid, as there's even less honor in being stupid and pretending otherwise. A lot of people were super hyped about OpenAI's 10 discoveries, but I still don't under
177.
▲
by
torginus
2mo ago
I'm just saying that this Task[T] pattern is completely alien to Go. If you want to have one-off ansynchrony, then just write it synchronously and start your work on a goroutine. If you wanna do batch processing, use channels. For all
178.
▲
by
torginus
2mo ago
Imo that's a rather nihilistic take. My personal opinion is that despite ungodly amount of money invested, the web has changed very little from the 'static website +maybe jQuery' days - 90% of the content people consume is s
179.
▲
by
torginus
2mo ago
> That car can be full time serving 4-10 units daily The problem is rush hour. You still will need some kind of minibus that can serve 4-8 people efficiently, and take up way less space. An added benefit is that if you cut peak traffic v
180.
▲
by
torginus
2mo ago
I am not an expert on LLMs, but what's preventing you from treating the context as 'virtual memory', and using the attention matrix to 'blank out' tokens which have very low weights and will not contribute much to t
More ›