Share this page!

Enter URL or ID to Unroll

You can paste full URL like: https://x.com/threadreaderapp/status/1644127596119195649
or just the ID like: 1644127596119195649

How to get URL link on X (Twitter) App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Quantіan Profile picture
@quantian1
Jun 23, 2020 • 4 tweets • 2 min read • Read on X
It’s important to state for the record that I’m not cancelling Scott Alexander for some rambling 25,000 word antifeminist blog nobody read, I’m cancelling him because he’s a neural free energy crank like half the “cognitive science” quacks on here.
iN FrIsTonIAn TeRMs, ImageImage
To be clear: huge swaths of cognitive science are worthless for the exact reason huge swaths of econ and quant bio are: unfalsifiable "theories" are constructed by cribbing cool-sounding concepts from physics and then cherrypicking data to "validate" them.
You grab some sort of "insight" from physics- free energy, Ising model, field theory- that nobody in your field is familiar with, and then you write a bunch of papers and do simulations that would be a problem set in a grad physics course, and BOOM, you have tenure and a SFI post

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Quantіan

Quantіan Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @quantian1

Quantіan Profile picture
Dec 12
A zero employee single manager fund runs somewhere between $500k and $1mm per year in expenses like audit, admin, legal/formation costs nowadays, so $50mm is the minimum for *you* to draw a paycheck, let alone employ a team, without eating performance fees for ~all of comp.
Once you have launched, institution interest evaporates because giving money to a nonexistent fund is way easier than a fund w/ a 4 month track, so you need to tough out 3 years minimum to get calls, and then they’ll bitch about scalability because you were managing HNWI money
In practice, if you want to acquire the accoutrements of institutional hedge fund life- an office, a team, a good PB, ISDA, etc., you need at least $100mm of fee paying AUM, which translates to more like $150 or $200 because various people will want deals and discounts to invest
Read 5 tweets
Quantіan Profile picture
Mar 23
Not to pick on this guy in particular, but there is a great deal of misinformation floating around re: DWAC. Fortunately for you all, I spent today reading the S-1, and as a SPAC expert I can confidently answer the question of “can Trump sell Monday” with a firm “idk, maybe?”
Here is their lock-up agreement, I’ve attached a screenshot of the relevant sections. Formal disclaimer that everything that follows is neither legal nor financial advice, I may have positions in securities mentioned, do your own research, etc. sec.gov/Archives/edgar…
Image
This is pretty boilerplate, as far as these things go, which it of course is because SPACs mostly just use a standardized template. The three major conditions are as follows: can’t sell until the EARLIEST of: six months, five months of the price is above $12, or they get bought Image
Read 8 tweets
Quantіan Profile picture
Nov 11, 2023
I just attended a talk on the mathematics of diffusion image generation models and it blew me away. Takeaways:
1. The math is very “simple”
2. In fact, it’s so simple that this method *shouldn’t really work*
3. The fact it does says something very fundamental about the universe
The basic idea is you define a stochastic process X_t, where dX_t = a(t,X)dt + b(t,X)dW_t. Let X_0 be random noise and X_1 be, eg, vector of RGB values for an image. Interpolate and add noise to generate a sample X_t, do this for many images, and then you can “learn” a() and b()
To generate a new image, you simply pick a new vector of random noise and evolve forward your newly discovered SDE until you reach t=1, at which point you have a perfectly sensible image if you’ve done your job right, modulo various finicky fine-tuning bits.
Read 9 tweets
Quantіan Profile picture
Aug 7, 2023
Since it's Sunday and I'm bored, I figured I would finally, definitively answer the question of "why do people think the economic is bad (and is that the case?)". This is because it is easy to answer, and the answer is funny and will make many people mad.
First of all, we will ignore everyone on Twitter and look at objective data, in this case the University of Michigan's consumer survey asking what people think economic conditions are. Spoiler: People think the economy is very bad, worse than any time other than 1980 and 2008! Image
Ok, now let's think of all the variables that might cause them to think that and build a model. I picked the following:
Inflation rate
Inflation rate change
Unemployment
Unemployment change
Housing prices
Real wages
Dollar strength
Interest rates
Stock prices
and went from there
Read 14 tweets
Quantіan Profile picture
Jun 26, 2023
Junior data scientist interview question: Assume you generate points X = N(0,1); Y = N(0,0.1). Rotate the (x,y) dataset 45 degrees, so they look something like pic below (line is y = x). If you were to calculate the OLS regression y = b1*x + b0, what is E[b1] as n->infinity? Image
(this question and the answer were previously buried in my replies, don't cheat or i will bully you for it)
The responses to this were bifurcated between “this is a hard question no one would get” and “this is a trivial question everyone should get ”. Surely if “data scientist” means anything, it means you should be able to answer basic linear algebra and mathematical stats questions!
Read 4 tweets
Quantіan Profile picture
Mar 6, 2023
As I strive to provide #content for my followers, I typed up my ratings from this weekend. Of the 200+ wines I tasted, ~150 were producer samples of guaranteed quality, and ~100 of those were reds (I don't drink enough white burg to have an opinion). Editorializing to follow:
The scoring method was just A+ to F, since it needed to be simple enough to be robust to me getting smashed over the course of the tastings, since I didn't spit. The older wines were Friday, all the 2020s were Saturday.
It is probably not surprising that Dujac, Hudelot-Noellat, Salon, etc. had the best wines. However, I was very impressed with the Robert Chevillon NSG wines, and the Didier Fornerol regional appellations punched way above their weight. Denis/Arnaud Mortet was also great.
Read 5 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us!

Send Email!

:(