Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
cs702
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
24 ms
·
301.
▲
by
cs702
1y ago
In theory, quantum algorithms could solve problems that have a combinatorial nature faster than classical algorithms. In practice, today, those problems are faster, cheaper, and easier to solve with classical algorithms on rented cloud hard
302.
▲
by
cs702
1y ago
The OP claims progress in AI is stagnating. The author is, no surprise, Gary Marcus. Even if you tend to disagree with him, as I do, the OP worth reading. My two cents: Progress in AI seems inevitable but its timing is unpredictable . Th
303.
▲
by
cs702
1y ago
The announcement is meant to make corporate buyers believe that IBM's quantum magic + AMD's chips could somehow leapfrog Nvidia's GPUs, hoping to slow down its sales. I never wrote Nvidia is working on quantum anything.
304.
▲
by
cs702
1y ago
> Today, IBM (NYSE: IBM) and AMD (NASDAQ: AMD) announced plans to develop next-generation computing architectures based on the combination of quantum computers and high-performance computing, known as quantum-centric supercomputing. AMD
305.
▲
by
cs702
1y ago
Yeah, I agree with that. When I first saw the announcement, my immediate thought was something like "huh, the Nobel guys sure want to make sure they've given an award for something related to AI which they can describe as foundati
306.
▲
by
cs702
1y ago
Whatever the facts, the OP comes across as sour grapes . The author, Jürgen Schmidhuber, believes Hopfield and Hinton did not deserve their Nobel Prize in Physics, and that Hinton, Bengio, and LeCun did not deserve their Turing Award.
307.
▲
by
cs702
1y ago
The OP is spot-on about this: If AI technology continues to improve and becomes capable of learning and executing more tasks on its own, this revolution is going to be very unlike the past ones. We don't how if or how our current ins
308.
▲
by
cs702
1y ago
Hmm... I wonder if the same pattern holds for people. In my experience, human beings who reliably get things done, and reliably do them well, tend to be less warm and empathetic than other human beings. This is an observed tendency, not a h
309.
▲
by
cs702
1y ago
Yeah. Anyone with even a vague awareness of history is aware of the historical parallels. Let's hope saner heads will prevail in these times of rapid change.
310.
▲
Restaurant chains feel the pinch as US consumers tighten their belts
(ft.com)
3 points
by
cs702
1y ago
|
1 comments
311.
▲
My ancestors fought in WWII. Hiroshima is plagued by shallow reading
(washingtonpost.com)
18 points
by
cs702
1y ago
|
37 comments
312.
▲
by
cs702
1y ago
Link to many pairs of photos showing the same individual(s): https://www.peterfunch.com/portfolio/42nd-and-vanderbilt/ In each pair, the photos may be years apart! What I find most fascinating is how consistent
313.
▲
by
cs702
1y ago
I haven't had a chance to read the preprint carefully or play with the code yet. Best place to follow what's happening is by looking at the github repo, specifically open and closed issues and pull requests.
314.
▲
by
cs702
1y ago
Yeah, the results look incredible indeed. That's why I and many others here have decided to download, review, and test the code published by the authors.[a] If their code doesn't live up to their claims, we will all ignore their w
315.
▲
by
cs702
1y ago
The authors' code is at https://github.com/sapientinc/HRM . In fields like AI/ML, I'll take a preprint with working code over peer-reviewed work without any code, always , even when the preprint isn&#
316.
▲
by
cs702
1y ago
Yeah, that was pretty much my reaction. I will need time on a computer too. The repo is at https://github.com/sapientinc/HRM . I love it when authors publish working code . It's usually a good sign. If the code
317.
▲
by
cs702
1y ago
Based on a quick first skim of the abstract and the introduction, the results from hierarchical reasoning (HRM) models look incredible : > Using only 1,000 input-output examples, without pre-training or CoT supervision, HRM learns to so
318.
▲
by
cs702
1y ago
Yeah, I cracked up when I read that, and couldn't resist posting it here.
319.
▲
by
cs702
1y ago
Has this been tried with reinforcement learning (RL)? As the OP notes, it is plausible from a RL perspective that such a bootstrap can work, because it would be (quoting the OP) "exploiting the generator-verifier gap, where it is easie
320.
▲
by
cs702
1y ago
The Jheri curl is the inspiration for "Soul Glo" in the movie "Coming to America:" https://www.youtube.com/watch?v=CGrasobHcKA https://www.youtube.com/watch?v=961x0NmyHKE According to h
321.
▲
by
cs702
1y ago
The question is: How do we get LLMs to have "Eureka!" moments, on their own, when their minds are "at rest," so to speak? The OP's proposed solution is a constant "daydreaming loop" in which an LLM is does
322.
▲
by
cs702
1y ago
I've only skimmed the paper, but it looks interesting and credible, so I've added it to my reading list. Thank you for sharing on HN! --- EDIT: The hierarchical composition and routing aspects of this work vaguely remind me of ht
323.
▲
by
cs702
1y ago
The results are entirely un surprising: > Countries with the highest levels of active transportation generally had the lowest obesity rates. > Europeans walked more than United States residents (382 versus 140 km per person per year)
324.
▲
by
cs702
1y ago
TL;DR: The OP believes that if we train large AI models via RL to duplicate the behavior of existing software (for example, train them to duplicate the behavior of an existing spreadsheet, an existing command-line tool, or an existing appli
325.
▲
by
cs702
1y ago
According https://en.wikipedia.org/wiki/Curtis_Yarvin , Yarvin: * has said that American democracy is a failed experiment, * has said he wants to replace American democracy with an "accountable" "techno-m
326.
▲
by
cs702
1y ago
Shouldn't the bureaucracies be penalized, instead of the poor engineers? The engineers built the 90-degree layout specified by their clients ! I wouldn't be surprised if there's a paper trail documenting the engineers'
327.
▲
by
cs702
1y ago
Relevant background on Brian Acton's funding of Signal, after leaving Facebook: "WhatsApp Cofounder Brian Acton Gives The Inside Story On #DeleteFacebook And Why He Left $850 Million Behind" ( https://www.forbes.com
328.
▲
ISS Astronaut Captures Photo of Rare Lightning Sprite Shooting Upward
(petapixel.com)
9 points
by
cs702
1y ago
|
0 comments
329.
▲
by
cs702
1y ago
Just, wow. I've never seen anything like that. Thank you for sharing it on HN!
330.
▲
Nvidia becomes first company to reach $4T in market value
(ft.com)
4 points
by
cs702
1y ago
|
1 comments
More ›