Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
T-A
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
24 ms
·
361.
▲
by
T-A
3y ago
> 50 million that is absolutely outrageous. Not even the British and North Americans(now US and Canada)native extermination had such high numbers North America was far less populated than South America to begin with; 7-18 million vs mayb
362.
▲
South Korea launches digital nomad visa to boost remote work options in 2024
(techsauce.co)
43 points
by
T-A
3y ago
|
58 comments
363.
▲
Long-Context Retrieval Models with Monarch Mixer
(hazyresearch.stanford.edu)
3 points
by
T-A
3y ago
|
1 comments
364.
▲
by
T-A
3y ago
This seems topical: https://tvtropes.org/pmwiki/pmwiki.php/Main/NoHonorAmongThie...
365.
▲
by
T-A
3y ago
Pretty much the only "AI" in that demo is edge detection (against a uniform background) and pattern recognition (by direct comparison with a handful of simple geometric shapes). Other than that, the robot seems to be following a h
366.
▲
by
T-A
3y ago
von Braun liked Kiritimati [1], and who can blame him? But the logistics seem daunting. You'd have to build up something comparable to KSC (staff: ~13k [2]) almost from scratch (there is a runway at Aeon Field and a JAXA tracking stati
367.
▲
The Global Project to Make a General Robotic Brain
(spectrum.ieee.org)
70 points
by
T-A
3y ago
|
21 comments
368.
▲
by
T-A
3y ago
https://en.wikipedia.org/wiki/Cosmological_principle
369.
▲
Gen AI on RTX PCs Developer Contest
(nvidia.com)
2 points
by
T-A
3y ago
|
0 comments
370.
▲
I Was a Statistics Professor. I Used Sports Betting to Retire at 43
(newsweek.com)
7 points
by
T-A
3y ago
|
0 comments
371.
▲
by
T-A
3y ago
Apollo 13's emergency was caused by the spacecraft's service module, not Saturn V: https://en.wikipedia.org/wiki/Apollo_13#Investigation_and_re...
372.
▲
by
T-A
3y ago
https://www.astrobotic.com/update-4-for-peregrine-mission-on... Unfortunately, it appears the failure within the propulsion system is causing a critical loss of propellant. The team is working to try and stabilize this los
373.
▲
by
T-A
3y ago
From the article, describing the now-retracted paper: “In people with hearing loss,” the authors wrote, “hearing aid use is associated with a risk of dementia of a similar level to that of people without hearing loss.” They proposed “up to
374.
▲
Vulcan Centaur rocket 'go' for historic Jan. 8 moon launch
(space.com)
4 points
by
T-A
3y ago
|
4 comments
375.
▲
China removes official after video games rules spark turmoil
(reuters.com)
4 points
by
T-A
3y ago
|
0 comments
376.
▲
by
T-A
3y ago
> And it played out for music just fine, after a while. The RIAA fought tooth and nail. And according to this guy, it won: https://foreignpolicy.com/2023/12/31/artificial-intelligence...
377.
▲
Google, Yahoo announce new email authentication requirements for 2024
(csoonline.com)
4 points
by
T-A
3y ago
|
0 comments
378.
▲
by
T-A
3y ago
From the paper's abstract [1]: It stands out by requiring merely a 24G GPU for training and an 8G GPU or CPU for inference. Built upon Phi-2, TinyGPT-V couples an effective language backbone with pre-trained vision modules from BLIP-2
379.
▲
TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones
(github.com)
237 points
by
T-A
3y ago
|
37 comments
380.
▲
Remembering robotics companies we lost in 2023
(therobotreport.com)
1 points
by
T-A
3y ago
|
0 comments
381.
▲
The state of Generative AI and Machine Learning at the end of 2023
(cnvrg.io)
1 points
by
T-A
3y ago
|
0 comments
382.
▲
A Novel Estimator of Earth's Curvature (Allowing for Inference as Well)
(papers.ssrn.com)
2 points
by
T-A
3y ago
|
0 comments
383.
▲
The Space Review: SpaceX Starship in Lunar Development
(thespacereview.com)
3 points
by
T-A
3y ago
|
0 comments
384.
▲
Causal Reasoning with ChiRho
(github.com)
1 points
by
T-A
3y ago
|
0 comments
385.
▲
Efficient Large Language Models: A Survey
(github.com)
2 points
by
T-A
3y ago
|
0 comments
386.
▲
by
T-A
3y ago
https://en.wikipedia.org/wiki/Copyright_in_compilation
387.
▲
by
T-A
3y ago
> This article is absolutely pointless. Not at all, but maybe you didn't read far enough to get to it. > The statement was from 2009. Back then, neither side had any real evidence in favour of their argument. Deep learning and AI
388.
▲
Metaverse or Mindtrap: Inner Shift, Outer Shock Kindle Edition (Currently Free)
(amazon.com)
1 points
by
T-A
3y ago
|
0 comments
389.
▲
AdapterHub – 716 adapters for 78 text tasks and 97 languages
(adapterhub.ml)
1 points
by
T-A
3y ago
|
0 comments
390.
▲
by
T-A
3y ago
Also, their derivative is just a multiplication (if normalized) or two (if not). https://hannibunny.github.io/orbook/preprocessing/04gaussian...
More ›