pkd

4K posts

pkd banner
pkd

pkd

@pkdroux

wannabe Maxwell's demon

Paris Katılım Ocak 2022
765 Takip Edilen271 Takipçiler
pkd retweetledi
Tribune Populaire🌐
Tribune Populaire🌐@TribunePop23·
🇫🇷⚓️INSOLITE - Le porte-avions Charles de Gaulle a été localisé avec un traceur à 5 euros ! Un journaliste d’Omroep Gelderland a glissé un traceur Bluetooth dans une carte postale, l’a envoyée via la poste militaire à la frégate néerlandaise Evertsen (en mission d’escorte du Charles de Gaulle en Méditerranée) et a suivi le navire pendant vingt-quatre heures.. Pas de piratage. Pas de satellite. Une puce à cinq euros dans du papier cartonné.
Tribune Populaire🌐 tweet mediaTribune Populaire🌐 tweet media
Français
38
355
1.7K
129.6K
pkd
pkd@pkdroux·
@JFPuget This ! LLM are juste ranking engine for word search
English
0
0
0
45
JFPuget 🇺🇦🇨🇦🇬🇱
A common misconception about LLMs is that a LLM outputs the most logical word(*) following its input. This is wrong. A LLM outputs the probabilities of each possible word given the input. For instance, let's assume the input is: Paris is the capital of Some high probability continuations are: Paris is the capital of France Paris is the capital of fashion Paris is the capital of which country? An inference engine will pick one word, then get new probabilities from the LLM. And it will pick another word, etc. Depending on the pick, it would generate one of the above continuations. Picking the word with highest probability is one way to generate text. In general generation picks a word among top priorities, so that repeated generations yield different results. (*) I equate token with words for the sake of this discussion.
English
11
3
65
10.4K
pkd
pkd@pkdroux·
@snwy_me Just talk about matmuls
English
0
0
0
121
snwy
snwy@snwy_me·
god i wish there was some way to actually explain to non-ML people what an LLM actually is/how one actually works is there any way to convey what a hidden state is effectively without any jargon? and without using some tenuous simile
onion person@CantEverDie

my biggest pet peeve around LLMs is when people (usually those invested in its success) call it “intelligent”. it definitionally, how it functions on a base level, is not intelligent. the way LLMs are built, it can never hit real intelligence. it’s just predictive

English
125
30
1.6K
266.8K
pkd
pkd@pkdroux·
@AcerFur Why don’t you continue the math degree ? Undergrad math is pretty chill. And if Cambridge is really hard just get from a lesser uni
English
0
0
0
42
Acer
Acer@AcerFur·
1/ Seeking advice on this since it’s been on my mind for the past four months or so If I were to drop out of my undergrad Cambridge degree and join OpenAI fully, what options would I really have on trying to continue to postgrad maths at some point?
English
90
7
334
74.4K
pkd retweetledi
Cassie Pritchard
Cassie Pritchard@hecubian_devil·
@tszzl Tbf he did predict this actually
Cassie Pritchard tweet media
English
7
6
509
7.9K
pkd
pkd@pkdroux·
@TheDavidSJ @yoavgo I fear all continuous math IS an arbitrarily close approximation. Lookup how reals are constructed
English
1
0
0
78
(((ل()(ل() 'yoav))))👾
ln, exp, minus and 1 are a neat basic set. but did you know i can also do it with 0, 1 and NOR
English
5
2
51
7.4K
pkd retweetledi
pkd
pkd@pkdroux·
@DesheShai @davidbessis Simplifying FP units in GPUs ? Though I feel like much better optimizations have already been used
English
0
0
0
279
Shai (Deshe) Wyborski
Shai (Deshe) Wyborski@DesheShai·
@davidbessis It's a cute result but it's hilarious how everyone is losing their shit about it. It has no practical implication whatsoever.
English
21
0
162
19.5K
pkd
pkd@pkdroux·
I’m no différent than an LLM: when working on a hard problem I need to talk to myself for hours
English
0
0
2
52
pkd
pkd@pkdroux·
@ProfBZZZ Meanwhile France be like: > You did very good, 14/20
English
0
0
0
13
Brandon Zicha
Brandon Zicha@ProfBZZZ·
A student today at my elite university admitted to me today that she took a class so she could work on reading for more than 20 minutes at a time. She can't read. She mainly skims and summarizes, she says and still gets A's. This student is, by professional standards, illiterate. Gonna have high GPA when she graduates. This conversation was had after 6 of 22 students dropped my course because the maximum reading per week in one week was over 100 pages. What people aren't grasping is that this is literally *dangerous*. These people are going to be come doctors, engineers, etc. They are - by any metric - vastly less capable than prior generations. These effects are cumulative over a lifetime. This grade inflation is part of the problem, but not even close to the entirety. And the problem obviously starts in K-12. Students don't know history because, you can't actually become historically literate on the advice of 'never assign more than 30 pages a week'. You can't develop any of the skills that came with literacy. This is, quite honestly, a civilizational catastrophe.
Steve McGuire@sfmcguire79

79% of grades at Yale are A-range. Graduating summa cum laude requires a record high GPA OF 3.98.

English
526
3.3K
20.8K
3.2M
wordgrammer
wordgrammer@wordgrammer·
How to build a code sandbox for GPUs. 👇
English
27
6
185
14K
pkd
pkd@pkdroux·
@Dorialexander Fuck I thought i had the Monopoly on witt quotes
English
1
0
2
82
pkd
pkd@pkdroux·
@badlogicgames I have an hosted instance and its very lighweight and customisable
English
0
0
0
7
Mario Zechner
Mario Zechner@badlogicgames·
is there something like google docs, but for markdown? i need a cloud based collaborative markdown editor please.
English
242
8
720
337.8K