Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
l33tman
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
16 ms
·
181.
▲
by
l33tman
3y ago
But there is really nothing that "normal" AI requires that is bound to CUDA. pyTorch and Tensorflow are backend agnostic (ideally...). Sure some hardcore people write some of their own cuda extensions but the vast majority don
182.
▲
by
l33tman
3y ago
Sure, but apart from the detail that you can make them pause by not feeding them words, you can't technically argue that they lack all those things. They are stateful in the sense that they see what they write, so they can keep their i
183.
▲
by
l33tman
3y ago
Yes, I came to the same conclusion. DDG doesn't feel as good as Google once was, but Google is almost unusable now, the entire first screen is almost always 5 paid placements and 5 SEO scam sites.
184.
▲
by
l33tman
3y ago
Just having the most popular product doesn't mean you're liable in an anti-trust case, you have to abuse the position as well (IANAL though). There are no viable competitors for a PC desktop OS for them to abuse and haven't b
185.
▲
by
l33tman
3y ago
Bankruptcy court can do clawbacks, I'm curious how that could apply for an investment scenario though.. it's not exactly paying a creditor or buying a Porsche just before a bankruptcy...
186.
▲
by
l33tman
3y ago
The analogue from stable diffusion would be ControlNet, where you can train a superimposed model on auxiliary data, this should be possible to do with chords for example, just like you can do with human poses, 3D depth maps etc in stable di
187.
▲
by
l33tman
3y ago
If the wavefunction can be prepared in the same way multiple times you can do quantum state tomography on it to reconstruct the wavefunction, this is essentially what they do in the paper but in a more efficient way than the state of the ar
188.
▲
by
l33tman
3y ago
A coder that can't draw, becomes a creative when s/he uses tools for drawing. I don't think the argument is in this, it's that the tools embody the collective creativity of all the world's previous-gen artists.
189.
▲
by
l33tman
3y ago
Whoever is simulating us might be so far ahead of us that they simply don't see or care about what is bad and cruel and tragic. It's like if we would simulate an antstack or a beehive. We would think it interesting but wouldn'
190.
▲
by
l33tman
3y ago
If you're referring to Feynman diagrams, sure in the modern treatment they don't mean what people usually think they mean, but it's also not accurate to state that particles don't pop in/out. It's a simplificat
191.
▲
by
l33tman
3y ago
In essence, you need to enumerate and sum up all possible ways everything can happen in a very little piece of the universe that corresponds to the "reaction center" of the experiment. All possible ways you can pop out particles f
192.
▲
by
l33tman
3y ago
In a sense you're right - it's not an electron that jumps around, the electron is its quantum amplitude wavefunction. Until you measure it, that is what you need to calculate the details of and time evolution of and interactions
193.
▲
by
l33tman
3y ago
Yeah it's weird, very often a venue on a street is simply not visible unless zoomed in TO THE MAX. And then it's not any weird obscure thing, it's a restaurant or coffeeshop that's been there for years.
194.
▲
by
l33tman
3y ago
If your feature actually has to perform texture lookups and stuff having effects outside the SM core (even if it just means flushing some cache lines), it will have some cost to run it in a float-blend kind of situation compared to simply &
195.
▲
by
l33tman
3y ago
I've been using telegraf + influxdb + grafana for my last projects, never really had to tweak anything after uncommenting the right sections in the telegraf.conf.. is prometheus and associated tools kind of an alternative to that stack
196.
▲
by
l33tman
3y ago
Auto1111 (latest on git) OOMed my 3080 running the base-xl on 1024x1024 unfortunately. Granted, my xorg takes up almost 950 MB on the precious VRAM. Did you get it to run using A1111 on the FTW3 without OOMing it?
197.
▲
by
l33tman
3y ago
The quoted paper yes, but the MoE concept and layers and training is old. Published as a conference paper at ICLR 2017 OUTRAGEOUSLY LARGE NEURAL NETWORKS: THE SPARSELY-GATED MIXTURE-OF-EXPERTS LAYER Noam Shazeer, Azalia Mirhoseini, Krzyszto
198.
▲
by
l33tman
3y ago
You shouldn't take the "mixture of experts" too literally, it's yet another architecture to use internally for a gradient descent optimized graph of ops. I obviously don't know how GPT-4 do it (or if it even does it
199.
▲
by
l33tman
3y ago
FWIW I recently had to switch to ZeroSSL from Letsencrypt which I normally use, while porting a backend from AWS (that used their certs) to a local server, as the IoT gear from 2016 we're supporting didn't have the root certs that
200.
▲
by
l33tman
3y ago
Which is weird because with the vision pro is not really any "first", they are incrementally doing a lot of things a bit better. My stupid guess is they probably worked for so long on this that they figured they have to ship somet
201.
▲
by
l33tman
3y ago
I think in the EU you're only allowed to call someone who you already have a customer relationship with.
202.
▲
by
l33tman
3y ago
Axis doesn't sell to customers directly (not even to businesses), they sell exclusively to distributors that in turn sell to well-stocked online computer stores from where you can buy them like any other tech product
203.
▲
by
l33tman
3y ago
I see your logic, but the apple hardware is super-expensive, a single Macbook is the same cost as a single RTX 4090 (not the MBA maybe, but definitely the MBP). So it's not that wide a stretch to say that the 4090 in a normal PC is als
204.
▲
by
l33tman
3y ago
The RTX 4090 GPU (a single PCI card) hits 82k GFlops in FP32
205.
▲
by
l33tman
3y ago
Also, it is starting to become the case that CUDA isn't that important anymore, both pyTorch and TF have numerous other backends and the programmer doesn't need to know what it runs on. And the GGML project has shown that you can
206.
▲
by
l33tman
3y ago
IIRC Jobs purchased NeXT when he took the Apple CEO job, to get a new OS base. As NeXT was his company, I'm pretty sure it was a decision based on bias and urgency and not because NextStep was the best there was. But yeah it's an
207.
▲
by
l33tman
3y ago
Unfortunately any case where the camera moves is not going to work, for a very large portion of the population (I don't know the percentages of people who get motion sickness in VR but they get that even from their own slow motion, ima
208.
▲
by
l33tman
3y ago
But I'm genuinely curious, why would the bigwig CEOs buy this if they didn't buy the Quest 2 or other previous headsets that could do the same things? You could do the cinema and virtual desktop and zoom calls with the Quest. Why
209.
▲
by
l33tman
3y ago
The "look at the search bar and speak" was pretty cool even if it's simple. Eyetracking is not available on most other VR headsets yet
210.
▲
by
l33tman
3y ago
Maybe the author's intent wasn't that, but the phrasing is a bit apologetic, like (exaggerating here...) "we're sorry to bother you, the average internet user with this discovery, but just so you know there ARE minoritie
More ›