Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
hislaziness
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
6 ms
·
31.
▲
by
hislaziness
3y ago
From the fraudsters/spammers perspective large platforms provide the economies of scale.
32.
▲
Pretraining data enables narrow selection capabilities in transformer models
(arxiv.org)
65 points
by
hislaziness
3y ago
|
109 comments
33.
▲
by
hislaziness
3y ago
As the responses are non-deterministic I wonder how useful these tests are? Even if I did the tests with my own prompts, wouldn't I get different answers for the same questions at different points in time?
34.
▲
Google / DeepMind Launches SynthID
(deepmind.com)
3 points
by
hislaziness
3y ago
|
1 comments
35.
▲
by
hislaziness
3y ago
Watermarking of generated art. I think this technology should also help original artist in watermarking their works.
36.
▲
by
hislaziness
3y ago
Curious is this just for fun or there is some advantage in representing the clock this way?
37.
▲
by
hislaziness
3y ago
I was expecting something like this but better - https://www.nationalgeographic.com/science/article/trapped-a...
38.
▲
by
hislaziness
3y ago
The post is gone...
39.
▲
CS25 Stanford Seminar – Introduction to Transformers with Andrej Karpathy
(youtube.com)
45 points
by
hislaziness
3y ago
|
3 comments
40.
▲
Fast Transformer Decoding: One Write-Head Is All You Need
(arxiv.org)
4 points
by
hislaziness
3y ago
|
1 comments
41.
▲
by
hislaziness
3y ago
A more efficient way to infer with lower memory requirements.
42.
▲
by
hislaziness
3y ago
From the course page - Classes are in person on Wednesdays from 2:00PM - 4:00PM Eastern Standard Time in E14-526
43.
▲
by
hislaziness
3y ago
I do not see any of the AI processors like the Google TPU. Would they be cheaper?
44.
▲
by
hislaziness
3y ago
Does it mean DOS/XP are not supported anymore? That would be sad.
45.
▲
by
hislaziness
3y ago
Works fine in Asia.
46.
▲
by
hislaziness
4y ago
I believe the Title is typically not written by the author. I would guess the Title is optimized for clicks.
47.
▲
by
hislaziness
4y ago
Not to digress but we are not able to regulate / control the use of genetic engineering, how can we regulate the current GenAI hype?
48.
▲
by
hislaziness
4y ago
Geothermal energy should be enough to sustain life.
49.
▲
by
hislaziness
4y ago
Start building real-time data apps in minutes Connect any data source, combine them in real-time and instantly get low-latency Data APIs. All with just a simple configuration!
50.
▲
by
hislaziness
4y ago
Papers would be good next step. The concept and the algorithms do not seem very complex. The training and fine tuning is what needs effort and infrastructure.
51.
▲
by
hislaziness
4y ago
I really liked this series. https://karpathy.ai/zero-to-hero.html
52.
▲
by
hislaziness
4y ago
A bit too late I guess. The only way this will stop is when people realize their expectations were misaligned and things start failing spectacularly.
53.
▲
by
hislaziness
4y ago
Where can we find this type of data for other countries? The data here is from the Office of the Australian Information Commissioner (OAIC).
54.
▲
by
hislaziness
4y ago
If anyone is interested PAIP is downloadable at https://github.com/norvig/paip-lisp
55.
▲
by
hislaziness
4y ago
Fascinating story. Daemon domes from Maxwell's thought experiment.
56.
▲
by
hislaziness
4y ago
"For this new analysis, ATLAS physicists revisited its data collected in 2011 at a centre-of-mass energy of 7 TeV (corresponding to 4.6 fb-1, also used in ATLAS’ previous measurement). Researchers employed improved statistical methods
57.
▲
by
hislaziness
4y ago
It does not have to be "me". My point is we seem to have a different benchmark for Natural Intelligence vs Artificial Intelligence.
58.
▲
by
hislaziness
4y ago
For whatever reason we seem to have set a very high expectation from AI as compared to NI (Natural Intelligence). I remember reading "There is a considerable overlap between the intelligence of the smartest bears and the dumbest touri
59.
▲
by
hislaziness
4y ago
I remember reading this somewhere - "There is a considerable overlap between the intelligence of the smartest bears and the dumbest tourists.". Though I do not think GPT-4 is even close to AGI it can definitely claim to be better
60.
▲
by
hislaziness
4y ago
Copilot with GPT-4 model, chat and voice, pull requests, command line, and docs to answer questions.
More ›