fock

19 ms·

151.
by fock 4y ago
MB is also famous for running Kubernetes. At a 1000 individual setups (all automated of course!). I had the pleasure of flamewaring someone who was very proud of this and couldn't really explain to me, why they would not just get two b
152.
by fock 4y ago
that doesn't look to bright if we look back towards Covid and vaccines. People are even bad at plain frequentist statistics!
153.
by fock 4y ago
The weirdest thing to me is that a person with a high personal, financial involvement in the subject went and took that quote from an old man (which to your and my understanding only states that these things are not linguistic "science
154.
by fock 4y ago
where exactly does the electron catch the proton?
155.
by fock 4y ago
https://en.m.wikipedia.org/wiki/Alpha_particle well ...
156.
by fock 4y ago
> A second mechanism, radiolysis, may generate the rest. As radioactive elements in the crust such as uranium and thorium decay, they emit beta particles, aka helium nuclei, What is this? Chat-GPT writing "sciency" fairytales?
157.
by fock 4y ago
so? translation is not the point of this models anymore.
158.
by fock 4y ago
well, then stop flying and the useless content-creation and let's get back to our "few GPUs" for consumption, when living, food and flying have transitioned to wind and solar. Then people can start setting up their text-based
159.
by fock 4y ago
supposedly the Fiverr-contractor soon just will be a prompt-tuner with a large cache of "good" images.
160.
by fock 4y ago
> we will certainly have some breakthrough as we have been seen for the past years. And things are speeding up. In the past years we have seen a tremendous scaling of a) workforce, b) hardware, c) data. For me a breakthrough would be on
161.
by fock 4y ago
Well, the number of parameters is directly related to the size of the machine to keep the model around. And so while you can host GPT-2 on a single consumer GPU, for ChatGPT you allegedly need 4-8-10k-a-piece GPUs. Which also consume 8-16 t
162.
by fock 4y ago
in some countries in Europe you actually can do that too from trains (and not only in Romania, where a family crossed the tracks (2 <4 y.o. kids, 4 pieces of luggage) to conveniently enter the parking lot, where their car was located)
163.
by fock 4y ago
> I suspect in hundreds or thousands of years we'll look back at age-related death culturally the same way we look at infant mortality during the medieval ages Currently our culture/civilisation can be expected with my life. We
164.
by fock 4y ago
I was shocked how (on my notebook) it sounds like there is something with the speakers.
165.
by fock 4y ago
Didn't we have those things already? Virtual-GL and Co. say hi. Also for most real GPU applications, you need to get the data in and out. I don't think splitting compute across a (insert any non-Infiniband-link) solves this
166.
by fock 4y ago
It's very good at talking something. But the something is ... well: how many people here think that carpooling with friends to go to a nature preserver is a good idea for a first date?
167.
by fock 4y ago
so what exactly is the difference between binaries and scripts executing arbitrary system calls?
168.
by fock 4y ago
> All rises in incidents of transgenderism are easily explained by a more tolerant society accepting it. The same appeared when we stopped. Explained (aka semi-solid empirical experiments - where are those?) or imagined (aka ideological
169.
by fock 4y ago
now a fairphone uses 2W during use, a dual-socket server 100W in idle (!). Assuming emissions scale with weight, I'd say for a typical server, this is very different. Still, I agree with you: design for efficiency and then mend and use
170.
by fock 4y ago
> One must use some digital scales for measuring the weight every day at precisely the same moment, because the weight varies during a day by much more than the difference between 2 successive days. or just do a moving average over some
171.
by fock 4y ago
looking at the time you seem to spend on this issue and the fact you're apparently only needing low double digits of those instances. Are you really sure you shouldn't just buy a bunch of machines (500cores/2TiB go for ~60k€)
172.
by fock 4y ago
https://news.ycombinator.com/item?id=32280756 as discussed here ;)
173.
by fock 4y ago
those papers are not "very visible" ML (NeurIPS and co.), but domainspecific "applications" and there's tons of it (as I am acutely aware).
174.
by fock 4y ago
not only social sciences. Except for the very visible areas of ML, the same happens in actual science and engineering... Example: the thousands of fraudulent XRD spectra of made-up compounds.
175.
by fock 4y ago
and then there is basecamp and you wonder whether maybe there's "experience" and ... experience. (though I agree that regulations and fitting to markets just requires people)
176.
by fock 4y ago
> Early cars were terrible too, but here we are. Were they? https://en.wikipedia.org/wiki/Benz_Patent-Motorwagen - as fast as a carriage, about the same stink. Carriages clearly had a usecase. Generated images now:
177.
by fock 4y ago
I'm sorry for reinforcing my bias publicly, unfortunately to late to edit. Thank you.
178.
by fock 4y ago
I'm not sure, what he claimed to do, but if he didn't run around praising this as the next grep: wtf is wrong with those people? I mean you have to start exploring things somewhere and why not share them?
179.
by fock 4y ago
> Setting up an IngressController is probably the job of a specialized Platform team who maintain your k8s cluster, or by the cloud platform you're using. What? I thought this was a solution solving problems, not making ones which y
180.
by fock 4y ago
> Add to this mix the USA trying to prevent China copying cutting edge AI technology ML-tech from the past century seems to work just fine for killing each other (have a look at Ukraine). I would be rather more concerned about the foe ha

More ›