XTXinverseXTY

4 ms·

1.
by XTXinverseXTY 5d ago
Scaling laws project that a model with more parameters trained for longer on more data yields predictably better performance, and that generally you want to scale these factors commensurately. More of the compute budget is being spent on RL
2.
by XTXinverseXTY 5d ago
Puts the onus on the AI companies to provide a specific replacement mechanism, no? Unless I'm unfamiliar with something else he's written that proposes something more specific and constructive To Tao’s credit he obviously identifi
3.
by XTXinverseXTY 8d ago
If they could declare with certainty that Buckminster's and Alpoge's usage data had been totally excluded from training, would that set a worse precedent and reflect poorly on their de-identification process (and data access safeg
4.
by XTXinverseXTY 10d ago
Apparently this post was prompted by a scary-sounding headline in The Information[0], that Astra is a looped transformer, implying CoT monitorability may be less reliable. The day after the report, Jakub tweeted[1] that he "wanted to p
5.
by XTXinverseXTY 12d ago
unnecessary condescension
6.
by XTXinverseXTY 2mo ago
So you admit that chapter 7 does not read almost as poetry?
7.
by XTXinverseXTY 5mo ago
I noticed that Clawdbot’s initial acolytes seemed to skew towards solo founders and hustler/grifter types. The Mac minis were likely to spam leads over iMessage. The single top downloaded skill was for Twitter. The fastest way to monet
8.
by XTXinverseXTY 9mo ago
I cannot tell, from the article, how to perform the Buteyko method. From the "Medical Evidence" section, it seems I'm not missing much.
9.
by XTXinverseXTY 10mo ago
Seems like this signals Yann Lecun's direction now that he's leaving Meta The EMA teacher model still seems like black magic to me (present in DINO and JEPA series but gone now)
10.
LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristics(arxiv.org)
5 points by XTXinverseXTY 10mo ago|2 comments
11.
by XTXinverseXTY 1y ago
This is an extremely common use case. Reading your comment history: are you an LLM? https://news.ycombinator.com/item?id=44531907 https://news.ycombinator.com/item?id=44531868
12.
by XTXinverseXTY 1y ago
Forgive my indelicate question, but why would someone buy a PO franchise?
13.
by XTXinverseXTY 2y ago
Friedman's thermostat Analyst visits his lumberjack cousin one Christmas at his cabin. Notices the cousin puts an amount of fire in the fireplace, which is correlated with the outside temperature, while the inside temperature remains c
14.
by XTXinverseXTY 2y ago
> I must've missed the part where the models began iterating under their own power He's probably referring to AlphaGo/AlphaZero in that sentence
15.
by XTXinverseXTY 3y ago
More intelligent people can afford to take on greater risks than most people. They make better inferences OP shouldn't try to imitate his friend, thinking he'll find similar success. It's an unpleasant conclusion but it'
16.
by XTXinverseXTY 3y ago
Guys like him get to do that, because he has the sort of brain that allows him to get a PhD in physics The world is a whole lot more dangerous and confusing for everyone else
17.
by XTXinverseXTY 3y ago
That's extremely pertinent information, why didn't you say so? An MS probably doesn't have that much more of a causal effect on your ability to do good ML work, as opposed to an equivalent amount of diligent self-directed stu
18.
by XTXinverseXTY 3y ago
This sounds like a bad idea. You are very smart to have asked before proceeding. I work as an MLE at a growth-stage startup with ~20 MLE/MLS folks. I was unemployed for 6 months before getting this job with 4YOE as an MLE and DS. There
19.
by XTXinverseXTY 3y ago
I've heard great things about bacopa monnieri[0], but many people complain about decreased motivation as a side effect Horizontal eye movements have consistently been shown to improve memory retrieval, dependent on handedness [1] [0]:
20.
by XTXinverseXTY 3y ago
"Emergence of life in an inflationary universe"? https://news.ycombinator.com/item?id=30047650
21.
by XTXinverseXTY 3y ago
don't rub it in
22.
by XTXinverseXTY 3y ago
I got a disappointing score on mensa norway[1] and boy, that made a lot of things suddenly click for me. I have yet to exploit this information for my benefit, as far as I'm aware. I think this concept is called an information hazard[2
23.
by XTXinverseXTY 3y ago
I think he just means that it can be an incredible pain to install.
24.
by XTXinverseXTY 3y ago
> according to George Hotz In this video from June, George Hotz says to go with "3090s over 4090s. 3090s have NVLink... 4090s are $1600 and 3090s are 750. RAM bandwidth is about the same." Has he changed his recommendations sin