Nick
@renegadesilicon
Followers
905
Following
380
Media
263
Statuses
4,062
AI/ML OG. Classics. Math. Beauty. Complexity. Athletics. Justice. ⚔️ 🏴☠️🏺
Joined May 2022
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Telegram
• 1407731 Tweets
Hezbollah
• 160545 Tweets
Lebanon
• 94058 Tweets
#MarkTuanxVALORANT
• 90210 Tweets
嫌いボタン
• 70369 Tweets
#FreePavel
• 59341 Tweets
Mark Tuan Superpower Stage
• 58817 Tweets
#VALORANTChampions
• 58460 Tweets
GULF DHEVAPROM FANCON
• 42568 Tweets
新潟2歳
• 25285 Tweets
Christoph Daum
• 23188 Tweets
ビシエド
• 18685 Tweets
BINI PLEASE PERFORM PANTROPIKO
• 18053 Tweets
SB19 RED CARPET PREMIERE
• 17725 Tweets
アルコス
• 17600 Tweets
#SB19PagtatagDocuPremiereNight
• 16664 Tweets
サトノレーヴ
• 16220 Tweets
キーンランドカップ
• 12010 Tweets
モレイラ
• 11440 Tweets
ナムラクレア
• 10460 Tweets
Pinned Tweet
“Αναξίμανδρος (Anaximander)….set in motion the process of rethinking our worldview….The scientific quest for knowledge is not nourished by certainty, it is nourished by a radical lack of certainty. Its way is fluid, capable of continuous evolution, has immense strength and a
0
0
2
@rod_mallo
Wait til you invest in hedge funds where you are in two funds that are taking opposite sides of the same outcome (say, Icahn vs Ackman). You pay two sets of fees on a net zero outcome.
12
31
876
@levie
That’s not the thing that’s bad. You’ve missed the point.
The collection, aggregation and profiling that allows the targeting is bad. That’s what the EU (rightly) takes aim at.
24
6
766
@UpdatingOnRome
Step one toward happiness: block / filter the word “physiognomy”
Step two: enjoy a small increase in sanity.
3
4
314
TIL "there are ~1.2M (!) duplicate images on the ImageNet-21K dataset, out of which 104K pairs belong both to the train and to the val splits (this amounts to 20% of the validation set)" 🤔
9
31
255
@jockowillink
@SDFD
Your gym burns down? GOOD. You get opportunity to rebuild an even more inspirational one.
0
0
177
This is such a fantastic podcast: how to create the ideal work and study environment by
@hubermanlab
. First class. Notes in the link!
2
14
147
@honeyNonABG2024
Ugh. Such cope. No guy you actually “want want” will be thinking this at all. Just simps.
12
3
119
@triketora
It's not just female founders who get this treatment. Humans will human; many (romantic partners or otherwise) resent the success of others.
8
1
108
@HamelHusain
@stripe
I think you need to sort this
@patrickc
- Hamel is one of the most important voices in ML and this looks extremely bad. Makes me think twice about using stripe for sure.
2
2
107
@datepsych
It doesn’t “predict” anything. It’s just an association / correlation — and even then, from unverified from the people who *wanted* to take the survey…which immediately biases it.
I’m not saying the observation is wrong. But this isn’t a statistical prediction.
3
1
101
A thought-provoking guide to detailed LLM prompting for programmers / system designers:
1
26
90
@signulll
This is exactly the kind of woman that is entirely inaccessible (and intimidating) to the vast majority of men. She’s smart, strong, fun, confident, social and seems to aggressively pursue her goals. How many men can say the same?
She states it up front; she’s looking for
4
4
90
For those who want to really know math better or have ever struggled or just want their kids to have the best math start possible, I can’t recommend
@AoPSNews
Richard Rusczyk’s work more highly.
1
10
85
Good times are here again.....
3
12
83
Absolutely psyched to be coming 2nd in my first ever Kaggle competition...and ahead of some true heavy hitters. Won't last long I'm sure -- but glad to be here!
Full credit to the awesome teaching of
@jeremyphoward
and the whole
@fastdotai
crew and framework; it's a joy!
6
2
71
@TurnerNovak
Some people definitely won't get the stock split ref and will promptly shit themselves
1
0
64
I need your help, friends.
I've begun my "math from the foundations" tutorial series. As well as (what I hope) will be inspiring, intuitive and beautiful explanations, I will create companion Jupyter notebooks so one can "play" with expressions to aid intuition building, but...
I've actually been thinking about writing a long, on-going series of narrative blog posts as a companion to the AoPS curriculum for those wanting to learn ML / AI from the foundations but who have never really learned or intuited the math (or anyone who feels lost with math)
1
0
11
2
6
51
8th place so far in the
@kaggle
Playground S3E3 comp to predict employee attrition from tabular data.
These things are so addictive! I blame
@radekosmulski
's new YouTube channel and his recent Grandmaster promotions to inspire me to start doing more of these!
7
4
43
This is such an under-appreciated but deeply powerful point in today's world. Thank you
@edfrenkel
:
0
7
36
@ylecun
@karpathy
@eladgil
@patrickc
By definition, your example is only a handful of paid jobs of all possible AI/ML jobs. And many of the hiring criteria for those jobs is “exclusive” by definition.
I think
@jeremyphoward
and the
@fastdotai
community have been / are great at surfacing a lot of talent that would
3
0
35
@snowmaker
@tszzl
This is not unique to Silicon Valley.
But believing it’s unique to Silicon Valley is unique to Silicon Valley.
1
0
35
@Carnage4Life
This is also the top tier of tone-deaf, mini-baller, thousandaire, moralistic self-aggrandisement.
0
0
34
@dvassallo
@dennishegstad
So we should ban Kara Swisher, whole magazines that are based on reviews, review websites, star rating systems, likes etc etc?
Because otherwise you’re saying there’s some arbitrary line where you (or someone else) gets to be the judge of what’s ok.
If he says / publishes
0
0
33
@sentantiq
Very much agreed.
@EmilyRCWilson
‘s notes in her new translation gave me a new perspective on Achilles.
I think the constant pressure of pursuing glory aggravates the moral bankruptcy and betrayal of Agamemnon in stealing Briseis. That act of immoral theft breaks his
0
3
33
I can’t help but think that ChatGPT is going to be the most disruptive technology in quite some time.
For the last 2 days, I’ve gone straight to ChatGPT instead of Google — now, without even thinking about it.
Imagine once it knows the same about us as Google does.
7
6
32
Friends, go read this post from my friend
@Angelina_Magr
on tools for data scientists — in particular the phenomenal nbdev!
We met at the recent
@fastdotai
unconference. Not only is she super smart and interesting but producing some great content, too!
1
9
31
@hardmaru
Preferably you must also be calculating the gradients for your private alpha GPT9 in your head.
0
0
28
Ok. This is actually very cool. Once these code models actually become more than superficially accurate, this could become a real game changer.
Imagine having a synthetic
@jeremyphoward
or
@JeffDean
or whoever actively reviewing your code. You can learn new styles / paradigms
Now that AI can read your code on Replit, your errors, and can even eval in the REPL it can send you fix diffs 🤯
13
37
353
2
4
30
@spencer_chubb
@levie
Glad you asked. People are typically unaware of the extent to which each of those activities goes on around their data and usage.
For obvious reasons, Europe has a very long memory and deep suspicion of the deep collection and synthesis of personal information for any kind of
6
2
30
@SmokeAwayyy
@tomgoldsteincs
I think this is exactly right. The horse has bolted. It will be like trying to identify which pen brand the ink on a page came from (and then trying to find the pen and match its fingerprints to an individual). Edge case useful — but largely pointless.
1
1
28
I’m considering buying a new GPU workstation for prototyping. Am I better off with 2x RTX 4090 or 4x A5500? Use case is powerful prototyping and running parallel experiments before shuffling off to the cloud. However, has to be “home friendly” — noise, power, HVAC etc. Thoughts?
9
2
26
@amasad
Strong disagree. Of course, depends on definitions. But, still, huge call. Good vs outstanding parenting can create very lasting distinctions and changes over time.
1
0
20
@Austen
People don’t want to face the inescapable banality that it’s typing out recipes that usually represent mathematical constructs to push bits from silicon to meat and encourage a return trip of same.
Finance had the same curse in the early / mid 2000s
0
0
17
"Sooner or later every generation is shocked by the behaviour of interest rates." -- Sidney Homer
2
5
19
@HamelHusain
It never seems to be consistent. They must have some dynamic availability.
I suppose it’s also changing as a result of consensus that it’s now the best.
Note to everyone reading: stop telling everyone Claude is now the best option! 😂
3
0
18
@iScienceLuvr
@UCDavisGrad
@UCDavisBMEGG
Can you just hurry up and cure all chronic illnesses now, please, Dr. Abraham? 😁 Throw in invention of a perpetual motion machine if you get bored before 40.
On a serious note: congratulations! Very exciting indeed.
1
0
17
@Miles_Brundage
I think this misses the point; some of the results are impressive. What I want to understand is performance *on average*. When these algos are making “live” decisions at scale, cherry picking won’t be possible; real people have to live with the average result as it arrives.
2
0
17
@hardmaru
I asked GPT3 to answer an poetry task from my daughter’s “talented 10 year old” national extension exam. Question circled on the left image and involved detailed instructions on style and concept. GPT’s answer on the right. I was genuinely taken aback
2
4
17
@jeremyphoward
Yes,
@inkdrop_app
supports this. You can drag in (or link) an image to Inkdrop's markdown editor and then share the note on the web with images etc intact via inkdrop's server. Cross-platform, tons of plugins, encrypted. I love it.
2
2
16
@Carnage4Life
Fair point — but inflation growth can get pretty scary pretty quickly. The USG will struggle to finance itself and avoid default with rates much north of 4 pct. If that scenario pays out, they have to print money — and then that’s 50 percent at least off the value of your cash
1
1
16
If you’re trying to understand some of the limits / flaws in ChatGPT (and the nature of how they occur) this discussion by Stephen Wolfram is a great entry point. Highly recommended.
On my weekly Science & Technology Q&A for Kids & Others livestream someone asked about ChatGPT ... and I gave some explanation & thoughts...
3
9
98
2
2
15
@Yampeleg
It’s all about structural (versus muscle) support to minimise barrel movement. If you use muscle support (as you inevitably would standing “naturally”) you’ll get slight tremor which progressively worsens as you fatigue from trying to hold the position with muscle.
In air
0
0
15
@PrefShares
@TurnerNovak
Anything from Carl Icahn. Especially when he did impromptu stand up in the 80s…genuinely very funny and a great insider take
1
0
14
This post on disingenuous PE valuation practices / volatility laundering is so good and so excoriating I set up an individual bookmark folder just to keep it fresh, unspoiled and in perpetual posterity.
“Meals for a lifetime”, too.
Congrats on one of the stupidest most mendacious posts yet in this ongoing ridiculous debate that one side (mine) wins easily on the merits and the other side (yours) wins easily on the economics as the investing world keeps thoroughly enjoying your
#volatilitylaundering
.
Your
48
38
524
1
1
14
Goedel's incompleteness theorems are perhaps my favourite, near-mystical ideas in math.
This great excerpt from
@edfrenkel
and
@lexfridman
describes some wonderful properties of the theorems for programmers -- especially AI engineers / researchers.
0
3
13
It's almost beyond comprehension that decades of experience, wisdom, highs, lows and the general blood, sweat and tears of bringing a vision to life can be encapsulated in a book. But
@tfadell
has done it with "Build"
Immensely valuable and filled with "meals for a lifetime" 1/2
1
1
13
Australia’s financial future is in much darker shape than people realise. Nowhere (major) on earth wrote more variable rate loans than Australia in the recent past. Our banking system is a literal racket. And it’s all subsidised and condoned by government.
1
1
13
@wasimlorgat
Probably the best combination of content and visual design is the
@WolframResearch
documentation...especially the "Applications" and "Neat Examples" sections. Nice feature is all examples can be run in nb
Check out the docs on their automated classifier:
1
0
13
Great job,
@rasbt
! Another first class article -- you're on a roll.
For the rest of you -- read up!
0
9
12
@ESYudkowsky
@ylecun
Personally, I’m thankful for both of you. And I’m thankful for the debate.
None of us have well calibrated probabilities on this topic because it’s more or less impossible to induct beyond the next few weeks. Debates like these are essential.
Any personal views on content
3
1
12
but in SymPy we have this solution -- which, to me, seems easier to map concept -> narrative -> code:
3
0
11
Yeh. Why not just put everyone in Europe on trains and send them to little work camps to become as “productive” as you’d like them to be? 🙄
Europe is immensely valuable and, in many dimensions, far more sensible. Economic productivity isn’t the be all and end all of existence.
3
0
11
@Carnage4Life
Github I get. There are non-trivial / non-standard swatches of code being reported as lifted verbatim from *license* protected repositories.
Diffuser libraries? Much harder call in the legal abstract. They really are novel works - unless a head has been custom tuned to an artist
1
1
11
I've actually been thinking about writing a long, on-going series of narrative blog posts as a companion to the AoPS curriculum for those wanting to learn ML / AI from the foundations but who have never really learned or intuited the math (or anyone who feels lost with math)
1
0
11
@bhutanisanyam1
@radekosmulski
@fastdotai
@jeremyphoward
@MadeUpMasters
Posting it here to save some googling for the world!
0
0
10
@nickcammarata
Ever-increasing personal responsibility, sense of detachment from ego, increasing compassion. Standing before things far greater than myself. Whole-food-plant-based eating.
0
0
10
@nntaleb
And, perhaps inherently short gamma, the door is dynamically sized. The more people start moving / running toward the exit, the smaller and further away the door becomes.
0
2
10
@tweetzachsmith
@rod_mallo
Yes. Unless you have the time, money, available brains (yours or others) to create some unique system to repeatably find edges that no-one else has found and you’re capable of (sustainably) detecting, sizing, exploiting and managing it / them, essentially into perpetuity.
0
0
10
@shidcoiner
@VCBrags
Also known as "the military". Or jail. Just institutionalisation in general.
1
0
10
@SchoenhoffKurt
Aside from it being a joke, it’s still non-trivial even on some distros like Fedora (within last releases). And, yes, if you have to mod config files and reboot to a non-graphical target to install the proprietary drivers, that’s a pain in the ass for anyone; not just noobs.
2
0
9
@mikeharrisNY
3 years? That’s a nearly 640% compound annual return or, assuming a 2% risk free rate and 18pct historical equity vol, a sharpe of roughly 35.
This would, in every conceivable quant or actuarial sense, be the definition of luck not skill. Entirely unrepeatable.
6
0
10
@Carnage4Life
She must have graduated from the same school as these two:
Want to give *the* killer presentation? Not sure how to put together a deck? Don’t go to Steve Jobs. These two have the answer:
@JaxonFairbairn
@Lachlansfilms
😂😂
0
0
5
0
1
9
@soumithchintala
@awscloud
@HelloPaperspace
absolutely destroys both. Great selection of GPUs (higher end options at reasonable prices). Brilliant pre-wiring of frameworks like
@fastdotai
Great UI to remote connect your instance to your local VSCode session. Persistent storage. Highly recommend.
0
3
9
@Carnage4Life
Manipulating pixel values in arrays / tensors to produce different visual effects based on the pixel values of all known works related to a concept is just the same as any of us learning a style - be it Annie Liebovitz or Van Gogh
0
0
9
@TNWJackson
@Plinz
If I squint, the face in the sun actually does somewhat resemble
@stephen_wolfram
!
2
0
9
@YoucefZYX
Virtually no one who gets this joke is thinking of NVIDIA in a gaming sense. All deep learning.
1
0
9
@CliffordAsness
Ahh, the heady days of JDS Uniphase (JDSU). Remember that one, Cliff?
One of my first forays into trading when one could do no wrong. I do wonder if this will end (at least) as badly…
2
1
8
@hardmaru
Oh that needs to become a T-shirt or something; one of the best ones you’ve created so far!
0
0
8
@sentantiq
“I have relied for the past several years on a panoply of somniferous consumables, though I now think that I have wasted thousands of dollars on sleep aids which could probably be replaced by a regular dose of Hanson’s soporific writing” 😂😂😂
0
1
8