Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
iNic
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
13 ms
·
181.
▲
Preventing the Collapse of Civilization / Jonathan Blow [video]
(youtube.com)
4 points
by
iNic
3y ago
|
0 comments
182.
▲
Squiggle: A simple programming language for intuitive probabilistic estimation
(squiggle-language.com)
1 points
by
iNic
3y ago
|
0 comments
183.
▲
by
iNic
3y ago
`mdfind` and `networkQuality` seem very useful!
184.
▲
by
iNic
3y ago
We know that models like GPT-4 go through phase changes, not in train/test loss, but for loss on specific tasks. This suggests the gradient descent is learning one thing after another, possibly in order of "difficulty".
185.
▲
Mummy Brown
(en.wikipedia.org)
2 points
by
iNic
3y ago
|
0 comments
186.
▲
by
iNic
3y ago
JAX's use of pytrees is great! They implemented a lot of useful utility functions, namely `tree_map`, that makes working with these objects easy and intuitive. I recommend looking at their neural network example library "stax"
187.
▲
Top Salmon Exports and Imports by Country
(worldstopexports.com)
1 points
by
iNic
3y ago
|
0 comments
188.
▲
by
iNic
3y ago
Having also watched the hearing I was pretty surprised at all the negativity in the comments. My view of Sam Altman has improved after watching the hearings. He seems to sincerely believe that he is doing the right thing. He owns zero equit
189.
▲
by
iNic
3y ago
You clearly know nothing about the alignment field since you are throwing together two groups of people that have nothing in common. The stochastic parrot people only care about "moral" and "fair" AI, whereas the AI saft
190.
▲
by
iNic
3y ago
I recommend people read the actual paper [1]. [1]: https://www.anthropic.com/index/constitutional-ai-harmlessne...
191.
▲
by
iNic
3y ago
Spontaneously means for something to happen surprisingly or without an apparent cause.
192.
▲
by
iNic
3y ago
This is why you have to choose your adversaries well.
193.
▲
ChatGPT for Me
(rewind.ai)
2 points
by
iNic
4y ago
|
0 comments
194.
▲
Lipoker.io
(lipoker.io)
2 points
by
iNic
4y ago
|
0 comments
195.
▲
What does the weather percentage mean?
(nicthinksaboutthings.substack.com)
1 points
by
iNic
4y ago
|
0 comments
196.
▲
Will AI pass the Turing test by 2029 Jan 1? (Market on Kapor vs. Kurzweil bet)
(manifold.markets)
18 points
by
iNic
4y ago
|
21 comments
197.
▲
Hitler's Math
(tabletmag.com)
2 points
by
iNic
4y ago
|
0 comments
198.
▲
Russian Mountains
(en.wikipedia.org)
2 points
by
iNic
4y ago
|
0 comments
199.
▲
by
iNic
4y ago
I believe distributed compute *could* be one of the few applications where crypto might make sense? You could earn tokens by supplying compute and pay to get it, and the token price is a natural market price for compute.
200.
▲
by
iNic
4y ago
Originally, from here[1]. Large Language Models seem to be better than humans at text prediction (at least on their trained data set). So it seems like the limiting factor for AI is more the goals we give it rather than the data or the size
201.
▲
Language Modelling Game
(rr-lm-game.herokuapp.com)
1 points
by
iNic
4y ago
|
0 comments
202.
▲
by
iNic
4y ago
As a mathematician I can say that all math articles I have read so far in Quanta Magazine have been excellent and they deserve praise! This one article might have missed the mark, but the topic is inherently difficult to write about to a la
203.
▲
Sniffnet: Cross-platform application to monitor your network traffic with ease
(github.com)
3 points
by
iNic
4y ago
|
0 comments
204.
▲
Deadweight Loss
(slowboring.com)
2 points
by
iNic
4y ago
|
0 comments
205.
▲
FTX, RIP
(thediff.co)
2 points
by
iNic
4y ago
|
0 comments
206.
▲
Book Review: “Status and Culture”, by W. David Marx
(noahpinion.substack.com)
2 points
by
iNic
4y ago
|
0 comments
207.
▲
by
iNic
4y ago
If the AI has sufficient "real life" access, think a robot body, then it can do wire-heading. Assuming it is smarter than us, and we have thought of this idea - it will also think of this idea. Now it doesn't have a reason an
208.
▲
by
iNic
4y ago
What I love is that this exact behavior is one of the minor problems of advanced AI. How can we make sure that a sufficiently advanced AI doesn't just edit it's reward center? Usually called wire-heading or reward hacking. Humans
209.
▲
Futuure: Write Now, Reveal Later
(futuu.re)
1 points
by
iNic
4y ago
|
0 comments
210.
▲
by
iNic
4y ago
I don't understand the advantage of this over just having a digital NN and then pushing the output to some electric motor?
More ›