Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
svantana
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
18 ms
·
211.
▲
by
svantana
3y ago
It will be interesting to see how this plays out. Spotify announced that it would start to crack down on adblockers in 2019 [1], but I've been using ublock origin on their web player since then and haven't noticed any issues (apar
212.
▲
by
svantana
3y ago
They are indeed, and to drive this point home (subliminally?), one of them is wearing a t-shirt with a llama on it in the press image.
213.
▲
by
svantana
3y ago
That article says nothing about how they achieved those numbers. Meanwhile, in the mojo docs, the matmul results are 20-50x slower for some reason: https://docs.modular.com/mojo/notebooks/Matmul.html
214.
▲
by
svantana
3y ago
I'm writing this comment on a computer, not with pencil and paper. Does that mean it was automated?
215.
▲
by
svantana
3y ago
Edge just means that the computing is done close to the I/O data, so that includes PCs and such.
216.
▲
by
svantana
3y ago
Yes - look at the file tests/test-opt.c. Unfortunately there's almost no documentation about its training/autodiff.
217.
▲
by
svantana
3y ago
with this method you get the same score for A-minor and C-major, right? Do you have a strategy to choose between those?
218.
▲
by
svantana
3y ago
It's not just you. All tech is the most magical, exciting and scary somewhere near the midpoint of its sigmoid development curve, whether it's rocket launches in the 1960's, radio in the 1920's or steam engines in the 18
219.
▲
by
svantana
3y ago
The wording makes it seem that the small one outperforms the large one, all else equal. But PaLM is tested in the few-shot setting, specifically chain-of-thought prompting, which means: 770M model uses 80% of dataset 540B model uses 0.01% o
220.
▲
by
svantana
4y ago
It finds the minimum with regards to the training data. The double descent phenomenon is about error on unseen data.
221.
▲
by
svantana
4y ago
All the model links are empty, that doesn't exactly inspire confidence.
222.
▲
by
svantana
4y ago
Hindsight is 20/20. Before ChatGPT, chatbots had a lousy track record. Pretty much any such project run by bigtech had been cancelled due to PR issues ("Tay" etc).
223.
▲
by
svantana
4y ago
These 2021 "lost tapes" have a lot of humans in the loop - according to this article, they made use of tribute bands for the recordings [1] [1] https://www.thestudentpocketguide.com/2021/05/entertainment&
224.
▲
by
svantana
4y ago
Looking at the source paper, they are claiming a 2.2x speedup over cuFFT for convolutions, so it's not an earth-shattering gain, but still.
225.
▲
by
svantana
4y ago
Yes, Google TPUs are sold to consumers under the Coral brand. https://coral.ai/products/
226.
▲
by
svantana
4y ago
I'm very into TinyML - not because of IoT but because small and fast models work well in (clientside) webpages, in realtime systems and pretty much everywhere. All while consuming less resources, which should be celebrated. Also, it&#x
227.
▲
by
svantana
4y ago
I feel like this article is a bit backwards - solving quadratic equations is only easy if you have access to the square root function, which by definition is a solver of quadratic expressions. Without it, one needs to resort to iterative ro
228.
▲
by
svantana
4y ago
I don't think anyone expected LLMs to do this well on code problems. Just a few years ago they were struggling to get the syntax right. But since it does seem to work in many cases, it's only reasonable to explore its limits. But
229.
▲
by
svantana
4y ago
> Search [...] is practically unusable today I keep seeing this sentiment around here, which confuses me because I think gsearch is better than ever. I find what I need instantly in 90% of cases, and when I don't, other engines don&
230.
▲
by
svantana
4y ago
Honest question: why not? Those 97% should have been fully aware of this arrangement when they bought their stock.
231.
▲
by
svantana
4y ago
I think (and hope) you're right. Recently, someone coined the phrase "bach faucet" to describe the mundanity of an "AI" that can compose millions of novel bach fugues that are indistinguishable from the real thing [
232.
▲
by
svantana
4y ago
On an emotional level, I totally agree. But as a developer of creative tools, I've been trying to make sense of what's good and bad in this field. Arguably, all creative software is at least partially automating creativity (photos
233.
▲
by
svantana
4y ago
Right! I think most people are uncomfortable with totally outsourcing their decisions, however smart the service is. I want to do my own research, and feel like I've gotten a good overview of the market before I commit.
234.
▲
by
svantana
4y ago
Exactly. Search ads are pretty much the perfect internet business model - if my biz is having a sale on lawnmowers, who better to target than people googling for "buy lawnmower"? If this is done right, everyone walks away happy an
235.
▲
by
svantana
4y ago
> I guarantee this game wasn't built with Unity It looks like you're right - the announcement tweet credits threejs and houdini [1]. I measured 13MB downloaded, of which most are fairly bloated "assets" (here's a
236.
▲
by
svantana
4y ago
I am one of those people - an academic type that likes to over-engineer, and at least partially work on embedded systems. What I've learned after two decades of experience is that it's best to apply most of that sophistication to
237.
▲
by
svantana
4y ago
Has it been verified that they actually use these instructions in Accelerate.framework? I just benchmarked this on my 2019 intel i9 mbp, and got the following speeds for 128x128 matrices, 32 repeats: cblas_sgemm: 36 GFLOP/s vDSP_
238.
▲
by
svantana
4y ago
Seems like she independently discovered Hofstadter's Law: https://en.wikipedia.org/wiki/Hofstadter%27s_law
239.
▲
by
svantana
4y ago
> All analogue circuitry is characterized by non-linearity to a greater or lesser extent The author seems to view this as a negative trait, which is bizarre to me. All acoustic instruments including the human voice are highly nonlinear.
240.
▲
by
svantana
4y ago
Well, there is hydro, wind and photovoltaic. And in the fusion field there are startups working on aneutronic fusion, which can generate power directly from charged particles. LPPFusion is one that seemed promising a few years ago, but unfo
More ›