Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
tehsauce
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
10 ms
·
241.
▲
by
tehsauce
5y ago
haha it’s a shame the level of effort in his paper reviews has really gone downhill, this one was especially so.
242.
▲
by
tehsauce
5y ago
Is anyone else bothered by the close/minimize/maximize buttons in the upper left corner which are there for no apparent reason other than to associate with macos’ design?
243.
▲
by
tehsauce
5y ago
It’s true you can start small and scale up, but it still would probably take many years to accumulate enough batteries to make a significant difference. Secondly, it doesn’t make 20 years to build a modern nuclear power plant, 5 years is a
244.
▲
by
tehsauce
5y ago
realistically building batteries or thermal storage at that scale will probably take as long as nuclear plants
245.
▲
by
tehsauce
5y ago
Their most famous project can easily be tricked into saying quite coherent horrible things which is a huge liability. They have to protect their brand
246.
▲
by
tehsauce
5y ago
very interesting take, thanks for sharing
247.
▲
EleutherAI One Year Retrospective
(blog.eleuther.ai)
142 points
by
tehsauce
5y ago
|
43 comments
248.
▲
by
tehsauce
5y ago
really cool survey of the topic, nice work!
249.
▲
by
tehsauce
5y ago
> I'm pretty sure that structure from motion + mesh reduction + normal mapping can reconstruct the original images almost pixel-perfect This is actually incorrect, NeRF is the SotA method for these novel view synthesis benchmarks. I
250.
▲
by
tehsauce
5y ago
It’s not even close. Banks provide so many services. your bank could be using a cryptocurrency as their backend, and you wouldn’t even know. the technology has almost zero impact to the user experience or efficiency. sorry
251.
▲
by
tehsauce
5y ago
It's even better than that, the stock split three times.
252.
▲
by
tehsauce
5y ago
I think it’d more accurate to say all interactions with the stock market exist on a spectrum between gambling and creating real value, depending on your strategy.
253.
▲
by
tehsauce
5y ago
we need rational thinkers. over optimism can be just as dangerous as pessimism
254.
▲
by
tehsauce
6y ago
For anyone looking for cheap gpus (quite a bit cheaper than this even), I highly recommend the marketplace on vast.ai I’ve spent hundreds of dollars there and will spend many more.
255.
▲
by
tehsauce
6y ago
often these companies will not hire their own janitors, but contract a another company that doesn’t give them the same expensive benefits as the regular employees
256.
▲
by
tehsauce
6y ago
“it can detect patterns among an O(n) input size without requiring an O(n^2) size neural net” This might be misleading, the amount of computation for processing a sequence size N with a vanilla transformer is still N^2. There has been recen
257.
▲
Why does an electric kettle get loud then quieter? (2016)
(thenakedscientists.com)
2 points
by
tehsauce
6y ago
|
0 comments
258.
▲
by
tehsauce
6y ago
I understand the frustration with badly asked questions, but not sure I’d want to work with someone who views conversations with coworkers as “offering little value”
259.
▲
by
tehsauce
6y ago
The monster's AI is amusing to read, and its behavior has a surprising amount of depth. Really explains its scary, clever and unpredictable nature. It has over 15 different behavior states: Idle, GoHome, Wait, Patrol, Investigate,
260.
▲
by
tehsauce
6y ago
Gpt-3 is just one of openai's many projects. Theres ~20 authors on the paper, and they all almost certainly did not spend a full year on this project. So 300m is completely wrong.
261.
▲
DawDreamer – Digital Audio Workstation Without a GUI
(github.com)
2 points
by
tehsauce
6y ago
|
0 comments
262.
▲
by
tehsauce
6y ago
I'm not sure what your engineering experience is, but I'm pretty certain that >80% of engineers are not in a position to refuse to work and quit their job without a severe impact on their career. And the same goes for managers
263.
▲
by
tehsauce
6y ago
"Engineering ethics" is a fallacy as it implies that engineers have high level influence on the goals set by organizations. If lockheed martin trained all of its missile engineers in ethics would they suddenly start building bombs
264.
▲
by
tehsauce
6y ago
Yes, this is exactly how it will go down I think. Somewhere I read a great analysis which said this was the opposite of "disruptive innovation", but instead this type of innovation will primarily reinforce the positions of the inc
265.
▲
by
tehsauce
6y ago
I'm more curious to see what happens when single developers are able to launch products which are vastly better than existing solutions, using just a couple days of effort.
266.
▲
by
tehsauce
6y ago
Agree. Snap used to be cool but Instagram sailed past it.
267.
▲
by
tehsauce
6y ago
You're right,cuda is only officially supported to run on an nvidia gpu, not cpu or anywhere else.
268.
▲
by
tehsauce
6y ago
2048 neurons per layer isn't really an accurate description, what he means is 2048 dimensional embeddings at each layer. The actual multihead attention layers in a transformer are not just feed forward 2048*2048, but actually have many
269.
▲
by
tehsauce
6y ago
A while ago I made an interactive rendering of hydrogen orbitals that is fully volumetric. Fun feature, the electron pdf is calculated without using any look-up tables so the algorithm can theoretically be scaled to any energy level you lik
270.
▲
by
tehsauce
6y ago
Why is the human brain so inefficient? It takes years just for it to compute the sha-256 of this media file.
More ›