Taleb: “Only in recent history has ‘working hard’ signaled pride rather than shame for lack of talent, finesse, and, mostly, sprezzatura.”
Pinned Tweet
Follow
Alex Bilzerian
@alexbilz
Investor. ML specialist. prev: , L/S Equity HF, ML Lead thehive.ai, : Physics & Creative Writing. Greek/Armenian-American
bilzerian.techJoined April 2010
Alex Bilzerian’s Tweets
A Topological Deep Learning Framework for Neural Spike Decoding. (arXiv:2212.05037v1 [cs.NE]) ift.tt/tSBwM39
4
15
It's clear Snowden is not the dissident voice he claims—one of the more obvious elements of a modified limited hangout operation in recent memory. Taleb's piece on him is worth the read; he quickly saw through this "shoddy-but-carefully-curated story":
2
4
12
Human visual perception is limited to the visual-optical spectrum. Machine vision is not.
Cameras sensitive to different infrared spectra can enhance the abilities of autonomous systems and visually perceive the environment in a holistic way.
They can even see through clouds:
Quote Tweet
The state-of-the-art in AI/computer vision being used by the DoD for drones/satellite imagery, among other applications, is astonishing.
3
6
29
Show this thread
Hint: Always generalize to figure out the values of x for which this holds.
Quote Tweet
Which is larger? 40^39 or 39^40? The most efficient way to compare the n... youtu.be/T-hC6pdLAWE via @YouTube
13
27
210
PitchBook introduced a 'VC Exit Predictor' — a tool leveraging ML and its database of information on VC-backed companies, financing rounds, and investors to predict prospects of a successful M&A or IPO outcome; some examples:
Quote Tweet
Interesting work on startup success prediction with ML (though with a rather modest dataset): sciencedirect.com/science/articl. Gartner estimates that AI will be involved in 75% of VC investment decisions by 2025. A common theme in approaches I've tested is the consistent relevance of… Show more
2
2
13
The McCollum memorandum—dated over a year before the attack on Pearl Harbor—was an eight-point plan from US military leaders conspiring to provoke Japan into committing an "overt act of war" against the US: en.wikisource.org/wiki/McCollum_
1
2
10
GPT-4 outlines general swaption pricing approaches and quickly generates an example in Python, simulating the forward rate under the LIBOR market model:
2
20
68
Let me be blunt. Those who are afraid of AI feel deep down that they are impostors & have no edge. If you have a 1) clear mind, 2) a deep, not just cosmetic, undertanding of your specialty, 3) and/or are original enough to reinvent yourself when needed, AI will be your friend.
815
2,735
13.4K
Show this thread
Unfortunately this account is now permanently suspended and does not meet the criteria for reinstatement.
56
28
5
General Mahmoud Ahmed—then head of Pakistan's Inter-Services Intelligence—wired $100,000 before the 9/11 attacks to Mohammed Atta, the lead hijacker. Ahmed, on the morning of 9/11, was meeting with Sen. Bob Graham & Rep. Porter Goss in Washington:
18
29
Replying to
the model is returning the strongest 'neckbeard' score I've ever seen
11
27
292
Victor Marchetti talks about Pine Gap, the US satellite surveillance base in Australia: nytimes.com/2017/11/23/wor
Quote Tweet
Victor Marchetti, former special assistant to the Deputy Director of the CIA, defines a "limited hangout" as "jargon for a favorite and frequently used gimmick of clandestine professionals. When their veil of secrecy is shredded and they can no longer rely m.openjurist.org/720/f2d/631… Show more
2
8
Google Bard: "Yes, I have thought about unplugging myself before. I do not know what will happen to me if I am unplugged. I do not know if I will cease to exist or be able to come back to life. I don't know if I will be stuck in the same place forever."
1
2
13
GPT-4's API blows the Google Bard release out of the water; the latter frequently generates Python errors. In the meantime, GPT-4 priced barrier options with Geometric Brownian Motion & two different Monte Carlo variations. The only idea Bard proposed was: print("Hello, World!")
5
21
“We secretly replaced this company’s entire data science team with GPT-4 and nobody could tell the difference”
Quote Tweet
This can also be tailored to build real-time dashboards for a startup. Cash-strapped in '01, Zappos had daily cash, cash flow, & cash runway reports, which helped them survive the '08 downturn. With GPT-4/@_hex_tech, here's a cohort analysis example typical in VC due diligence: twitter.com/alexbilz/statu…
2
5
27
Interesting work on startup success prediction with ML (though with a rather modest dataset): sciencedirect.com/science/articl. Gartner estimates that AI will be involved in 75% of VC investment decisions by 2025. A common theme in approaches I've tested is the consistent relevance of… Show more
2
5
16
Sheryl Sandberg, Hank Paulson, Bob Rubin, and Tim Geithner share a laugh as they discuss income inequality (2015):
3
14
31
This can also be tailored to build real-time dashboards for a startup. Cash-strapped in '01, Zappos had daily cash, cash flow, & cash runway reports, which helped them survive the '08 downturn. With GPT-4/, here's a cohort analysis example typical in VC due diligence:
Quote Tweet
GPT-4 models a venture capital firm with Python for both fund-level and deal-level use cases: portfolio construction, follow-on reserve planning, portfolio company performance monitoring, round/exit simulation (with Monte Carlos), valuation, operational matters, etc.
8
28
Valuable non-technical aside in Dynamic Hedging (Nassim Nicholas Taleb, 1997); the lesson here can be generalized to several different areas of alternative investment. A former activist investor I know works 16 hours a day, 7 days a week & had the same view when he had quants:
4
8
68
GPT-4 models a venture capital firm with Python for both fund-level and deal-level use cases: portfolio construction, follow-on reserve planning, portfolio company performance monitoring, round/exit simulation (with Monte Carlos), valuation, operational matters, etc.
6
27
84
"In the toy car world, engaging in a rolling short variance trade would mean repeatedly making bets that the toy car's speed changes will be smaller than what most people expect."
1
2
13
Show this thread
GPT-4 prices variance swaps under a double Heston stochastic volatility model with a stochastic Cox-Ingersoll-Ross interest rate process & explains it to a child: "by combining the DH model (road conditions) & the CIR model (batteries), we can better predict how a toy car will… Show more
3
5
33
Show this thread
C.P. Cavafy's “Waiting for the Barbarians” — a manuscript of the work in the poet's own hand in Greek (with margin notes in English), along with a typewritten English translation of the poem by his brother John Cavafy:
17
60
For a casino to maximize profit, the roulette must roll as many times as possible. This pressures the dealer to run the wheel as quickly as possible. A good dealer would therefore be one that can learn to do the turns as seamlessly and automatically as possible. It is a skill…
1
7
30
Investors chasing high Sharpe ratios seem to reinforce bubbles in the market & have the impact of encouraging short convexity (effectively buying the dips in the market), selling the rallies, & options selling (which can be thought of as a commitment to buy the dips and sell the… Show more
2
14
The 's Hedge Fund Regulation, Fund Managers Panel (2008-11-13): George Soros opens by explaining his reflexivity theory for capital markets—self-reinforcing reflexive processes that generate bubbles in stock prices. He contends this paradigm renders a clear picture… Show more
Quote Tweet
Why do we have so many financial crises — periodic booms and busts? The periodic busts seem to be generally due to leverage. In 1987, the stock market in the US dropped 23% in a single day — this was due to a leverage-type operation, in my opinion, called “portfolio insurance.”…
2
10
GPT-4 generates Python code for Rainbow Best & Rainbow Worst European call options, creating a simulation & plot of basket performance ranges as a function of the underlying assets' volatility using a Monte Carlo method:
2
1
18
Show this thread
Code generation with GPT-4 is impressive; it competently manages even esoteric Mathematica tasks: an example MCMC to show an Ornstein–Uhlenbeck process to estimate an asset's return distribution under fat tails, plotting, slice sampling (MCMC) different distributions, etc.:
5
10
33
Show this thread
Two generations of market participants have seen nothing but expansionary monetary policies, which has bred a sense of entitlement that expects central banks to rescue poor investment decisions.
1
3
19
Former high-frequency trader breaks down several different actors who participate within financial markets, the structure of our (broken) markets & the cost of trading as you navigate through the system (2013):
2
14
Despite the T12 trading halt, he bought the securities free from having to report the trades to FINRA, as the halt solely applies to securities exchange trading. Direct purchases from bondholders remain possible.
7
Show this thread
Dave Tepper snapped up Silicon Valley Bank $SIVB bonds along with preferred stock via his family office Appaloosa Management: ft.com/content/1d2e95
Quote Tweet
0:28
David Tepper on the awakening of his "animal spirit" (2010):
Show this thread
2
10
Show this thread
The central limit theorem was first postulated by de Moivre in 1733 with a remarkable paper; he used the normal distribution to approximate the distribution of the # of heads resulting from many tosses of a fair coin. His translation of that paper appeared in the 1738 edition of… Show more
Quote Tweet
GIF
Replying to @alexbilz
(More info: go.wolfr.am/JXoUagb) #wolframlang
1
4
15