You say that the times are bad. Be better yourselves, and the times will be better: You are the times. Technology must multiply human agency. DMs closed.

New York, USA
Joined October 2018
If you have an obsessive streak, this is the greatest time to be working on frontier technology. The problems are genuinely monumentally difficult, which makes progress feel amazing. Besides, it's quiet on the frontier. None of the circus of mainstream AI.
14
24
2
342
20,916
d5176f8f4c439d3014c6adce42814509c19c33ad119e1f4fe39baf59ffbd2c3577ec0fbf51ba73b913fc1b0d97a59c82046c4f9a2900df2c4449ff925b340a38
1
195
akhil retweeted
Economics of AI at the moment suggests there's $10K out there in someone's pocket that is worth more than the many hundreds of billions of dollars sloshing around because those dollars can get the necessary leverage over depleted physical resources and start a non-zero-sum game.
1
5
430
If you're a technologist, you have a duty to do more than critique. As Blake said: "I must create a system, or be enslaved by another man's. I will not reason and compare: my business is to create."
the technologists will not stop . they are incapable of it . every technology that is possible will be created
2
28
651
It's time to read CS Peirce. Let's see how much of a dent I can make in two volumes.
1
249
My criterion is: make sure you're working on the most important, civilization-critical technological problem, and review it often. This automatically rules out LLMs and a whole host of things that are DDoSing most people.
We’re hitting a period in time in tech where money is irrelevant. Important to re-evaluate all priors. The most expensive thing right now is to work with people that are just larping and wasting their time. You will actively get dumber and fall behind. This was always true but the clock cycles have collapsed to be much faster now.
2
1
47
2,607
This is a limited view of reasoning. Mathematical logic can be a blinding force when you think about knowledge production and its reflexive relationship to consciousness and cognition.
Ask your pretended logician whether there are any necessary reasonings of an essentially different character from mathematical reasonings. If he says no, you may hope he knows something about logic; but if he says “yes,” he is contradicting a well-established truth /1
1
4
277
I wonder if there are limits to symbolization.
139
This suggests our idea of intelligence is wrong.
Replying to @razibkhan
He may have well been one of the smartest people in the world for his birth year considering both physics Olympiad and Putnam honorable mention He switched into AI from biology at over 30 and was VP of research at OpenAI just a few years later…
1
260
Been reading Sloterdijk. Good. I'll be reading more like this.
2
255
It's curious that we believe nature evolves but don't think so about the universe. Anthropocentric bias + scale.
1
8
387
What is the relationship between gravity and computation?
2
6
295
There's an emerging coalition of catholic technologists. Many cradle catholics :)
Techbros believed they were above religion, so naturally they made their work their religion and ask you to fund their latest jihad/crusade against humanity.
2
3
312
These people are not technologists despite their STEM educations and sci-fi predilections. Not having the chops to spar with complex systems (including social, technical, political, economic, etc., aspects) is the tell. AI safety is the current manifestation.
Replying to @pitdesi
Y2K was the dumbest thing ever. It was easily mitigated, and companies had years to work on it. I remember VC backed companies “working” on Y2K, which evaporated 1/1/2000. Dario sees everything through the lens of computers, but there’s a whole world out there that just adapts when computers don’t work. AI isn’t a risk anymore than PCs or the Internet were. We will find problems … and fix them.
20
696
Hmm.
Ask your pretended logician whether there are any necessary reasonings of an essentially different character from mathematical reasonings. If he says no, you may hope he knows something about logic; but if he says “yes,” he is contradicting a well-established truth /1
1
217
I wonder if math is just like the alphabet in this regard.
to me it seems far simpler and much more likely that instead of the platonic notion of all math dwelling in some ideal eternal domain waiting to be discovered, math is the product of collections of logical statements devised by humans and has no special metaphysical status.
2
6
460
I wonder if new scientific paradigms also create new language games. (almost surely, yes)
4
266
Ha, just learned that Whitehead felt the same about space being defined in terms of points.
When you push on some theoretical cs questions, you realize that defining space in terms of points doesn't really make sense.
7
374
National Museums Scotland announced it had acquired a rare, 8thC, gold sword pommel #OTD in 2022. A detectorist found it at Blair Drummond, Stirling, in 2019. Encrusted with garnets, it has eyes and a beak on each end. 📸National Museums Scotland
6
134
8
683
7,940
Es gab einst viele berühmte deutsche Gräzisten. Die Synthese (ja gar Symbiose) von Deutschland und Griechenland wirkte noch bis ins frühe 20. Jahrhundert fort (Carathéodory), und wurde dann planmäßig zerstört. (U.a. durch Bildungsreformen; quasi-Abschaffung des Gymnasiums.)
Wenn die Deutschen schon mit der stupidesten Indifferenz den Zerfall ihrer Sprache über sich ergehen lassen, will ich wenigstens auf einen Griechen (Kondylis) aufmerksam machen, der Deutsch zu seiner philosophischen Wahlsprache erhoben hat und hier eine Liebeserklärung an die deutsche Sprache gibt (wie verständlich kann ein Text noch sein, in dem ein Ausdruck wie "geistige Ehrlichkeit" vorkommt?): "Wenn ich aber selber schreiben muss, kann ich nur meinen eigenen stilistischen Neigungen fol­gen, um ein eigenes Denken in die angemesse­nen Kanäle einzugießen. Und gerade die Strukturierung meines Denkens führt mich in die Nähe des transparenten und möglichst schlichten Stils, und zwar weil dann, wenn es sich um theoretische Texte handelt, die geistige Ehrlichkeit es verbietet, Mängel des Denkens mit Verblümungen und lexikalischen Seiltänzen zu überdecken. Insbesondere bietet die Sprache, in der ich ursprünglich meine Bücher schreibe, dem Kenner der klassischen Sprachen außerordentliche Vorteile: Die deutsche Sprache enthält weit mehr als andere Sprachen die syntaktischen Strukturen und lexikalischen Modi der altgriechischen und lateinischen Sprache. Insbesondere die fast unbeschränkten Möglichkeiten der Verbindung von Haupt- und Nebensätzen innerhalb längerer Satzperioden erlauben die präzise Formulierung mehrschichtiger Gedanken, wo in der syntaktischen Verbindung reliefartig die Verknüpfung und das Gewebe der Beziehungen zwischen den verschiedenen Dimensionen des Sinns ausgedrückt werden. In keiner anderen der mir bekannten Sprachen kann zum Beispiel so gut der Stil des Thukydides wiedergegeben werden. Meinen deutschen Lesern, die mich fragen, wie ich zwanglos und sogar mit Ansprüchen eines persönlichen Stils in einer fremden Sprache schreiben könne, antworte ich, dies sei größtenteils auf meine Vertrautheit mit den klassischen Sprachen zurückzuführen." (Panajotis Kondylis: In konkreter Lage. Gespräche) Was Kondylis hier über die Verwandtschaft des Deutschen und Altgriechischen sagt, sagte auch mein Altgriechischdozent gleich am Anfang der ersten Sitzung, bevor auch nur ein Wort übersetzt wurde. [S.E.]
3
4
1
12
1,307
Okay, so if you conclude that selling compute is more profitable than pursuing your research idea, who are you selling to? Either it's to nobody because everyone's in the same spot. Or you're selling to the people who are going to hand you a historic ass whooping.
Heard from a neolab friend: “Even if we fail to train the best model, we locked in GPU capacity early. H100/B200/B300 rental prices have since gone up 2–3x. We can just resell the compute and print money.” OK, so that’s the business model.
6
15
985
These idiots. God.
2
125