Calc Consulting

Making bigger and bigger steam engines never produced thermodynamics. Thermodynamics produced engines that made the steam engine obsolete. Engineering frequently precedes scientific understanding, but once the science is understood, the engineering leaps forward by orders of magnitude, and the old products become museum pieces. Examples include: Simple electricity & telegraphy → Maxwell’s theory of electromagnetism → radio, radar, modern electronics, photonics Practical telegraphy, telephony & radio → Shannon’s information theory → digital communications, error-correcting codes, the Internet Early powered flight (Wright Flyer era) → Aerodynamics & boundary-layer theory → monoplanes, jets, supersonic aircraft Vacuum tubes & early solid-state work → Quantum mechanics + band theory of solids → transistors and the entire semiconductor industry Nuclear reactors & atomic bombs → Nuclear shell model → advanced reactors, nuclear medicine Empirical organic chemistry & industrial dyes → Quantum chemistry & molecular orbital theory → rational drug design, modern catalysis, advanced materials We are just beginning to explore neural networks. The Night is Large

Making bigger and bigger steam engines never produced thermodynamics. Thermodynamics produced engines that made the steam engine obsolete. Engineering frequently precedes scientific understanding, but once the science is understood, the engineering leaps forward by orders of magnitude, and the old products become museum pieces.

we have had AGI for at least 2 years

@garrytan Because the communists never cheat x.com/typebulbit/sta…

@GeneralMCNews whoa!!!

BREAKING: Mark Zuckerberg's employee at Meta, Josh Felker, who created "digital experiences for kids," was arrested after admitting he tried to sext a 13-year-old during an undercover sting.

@HelloVyom yeah, that’s pretty Baller

Another proof that you need to leave India to make it big in life. No name college in India to CTO at OpenAI

@rohanpaul_ai I was asked to be a extra witness on this case, but I refuse to do it because I knew they were guilty

Download the weights from HF right now

During the Cold War, the US and its allies actively banned or severely restricted the export of advanced technology to the Soviet Union and its bloc. This was a core part of our containment strategy. And it worked. The same thing can and may be done now, in reverse, to contain the CCP threat. I would think that if Trump does ban these models, it's not just going to be in the US, bit that there would also be severe penalties on any country using Chineese AI models.This could include: - Targeted sanctions and secondary sanctions on governments, agencies, and major firms - Tariffs and trade restrictions - Export/import controls and technology-sharing bans with violators. - Immigration and visa restrictions - Broader diplomatic and, where necessary, military signaling The US and the West need a modern CCP containment strategy that combines hard power penalties paired with softer influence campaigns and strategic investments at home.

The only way this works is with a full decoupling of the two economies. We also have to make allies and partners choose a side. Otherwise, there will be so many holes as to make this idea a joke. It would be the same as the $800 tariff exemption policy which meant nearly everything shipped from warehouses in Mexico and Canada.

@6Silence3 ?

@CalcCon missing him by months is the brutal part heavier tail than anything in that histogram

To truly reach AGI, we have to go beyond engineering hacks and establish the scientific principles behind why deep learning works. One of the most amazing things we have discovered is that the best layers of the best deep neural networks all seem to converge with a universal signature This work dates back to papers we published in JMLR and Nature Communications back in 2021, for work actually done in 2018. Hard to beleive its been this long It had been clear to me from day one that these signatures resembled a univerisal critical exponent from Renormalization Group theory. And after several years of work, and numerous publications and talks, and a lot of sweat equity, I think I have finally cracked this and can provide a complete Wilsonian RG view of what's going on in learning (My only regret is that my PhD advisor passed away earl this year before I could share it with him, as he was one of the greats of RG theory.)

@gfodor 😂

the stochastic parrots are getting pretty lucky

@miles_commodore 😍

My cousin Eartha Kitt played cat woman in the third and final season of Batman.

@AlecMacGillis He lost some of the best years of his life for creep and his investors

This is pretty bleak.

@WombatPM @JundeMorsenWu @pfau The vast majority of full-time PhD students (both domestic and international) receive full tuition remission (they pay $0 or very little in actual tuition).

@CalcCon @JundeMorsenWu @pfau The tuition for foreign students in Graduate STEM depends on the program, but lab facility’s, materials, research programs all come out of research funding. And there are only so many slots available. Non STEM, and non Tier 1 programs, foreign students are a cash cow.

I’m an AI PhD at Oxford wanted to move to the US for years, but it never worked out. Maybe I can share why. To me, it feels like the US is making it harder and harder for Chinese talent to come or stay. First, Trump signed PP10043, which meant I couldn’t go to the US for any graduate school. That’s why I had to turn down my Stanford offer and come to Oxford. Later, I qualified for an EB-1A (the extraordinary ability green card). But because I was born in China, I’m stuck in a country-specific backlog that could take another 4–5 years. Meanwhile, on the other side, I’m getting 3–4 emails from China almost every week. They offer high salaries, free housing, generous research funding, and even offered an astonishing amount of money just for me to come back and have a conversation. I’ve always loved the American spirit. I truly believe my abilities could have a bigger impact there. But honestly, the whole process has been incredibly frustrating. All of this makes me wonder whether I should just give up. Why am I spending so much energy trying to go to a country that doesn’t seem to want me, when another country is doing everything it can to bring me back?

@WombatPM @JundeMorsenWu @pfau And that research funding comes from where ? NSF, NIH, DOE, etc And don't forget, US corporations and wealthy individuals get a tax break for funding such work as well. We even fund research in China, like the Wuhan lab.

@adamzwasserman The approach here is not to try to characterize the outputs of an LLM, but to understand the fundamental phyiscs of learning.

@CalcCon To truly reach AGI, all we have to do is define AGI in reachable terms.

Mostly correct, incentives matter But in a STEM PhD program, the US taxpayer or some company pays And where does that money go? Overhead to university administratora who discriminate against the best American students and push policies like math equity on our kids Even with the Supreme Court decision, they are still discriminating against our kids in secret And of course they charge ridiculous tuition plunging American kids into a modern indentured servitude with useless degrees in a market that’s flooded with foreign talent

@CalcCon @JundeMorsenWu @pfau Its incentives. America has been able to attract top talent from across the world. America has also been able to attract foreign money. Universities do a little from column A to get generational minds, and a little from column B because foreign students pay full rate.

@quxiaoyin All true. I argue, however, that their biggest mistake was what certainly looks like from the outside, conspiring to commit espionage with the mix-up with Hegseth and the DOW Incredible arrogance and stupidity rolled into one

Anthropic had the biggest miscalculation in AI 2026: 1. They underestimated the progress others made, and thought they can win forever. 2. Thus, they assumed their customers will continue to tolerate their terrible policies (pricing, random usage limits, bad data retention, arbitrary access removal etc.) and cocky PR because they are AGI. They didn't invest in building customer trust because they thought they didn't need to. 3. They also assumed by simply painting open-weight as dangerous and unsafe, they can persuade enterprises to not touch them and persuade government to ban them. 4. They knew user data in Claude Code is how they win, yet their bad user policy/pricing gave aways their already-sticky users to Codex/Grok/etc, enabling others to capture equally valuable user data on their own. 5. Meanwhile, they didn't invest enough in owning their compute, instead putting itself in a vulnerable position at the mercy of @elonmusk, their competitor. When a strong open-weight provider catches up(@thinkymachines, Kimi, GLM etc.), people suddenly realized Claude has 1. no pricing advantage 2. no compute advantage 3. no branding advantage (already broke enterprise trust, pissed off prosumers, and consumers don't know Claude exists) Anthropic needs some serious strategy pivots. The status quo won't work at all for them. They still have the highest caliber talents, so I am still hopefully they will change for the better. But simply a better model than Fable isn't going to be enough to turn things around IMO.

Not having any data or any compute, back in 2015/2016 I started out by just looking at the eigenvalues of the layer weight matrices I soon realized that the best layers are power law distributed, with what was looking like a universal critical exponent That screamed SOC, I knew about both the work on sandpiles at Chicago when I was there in the 90s, and Jack Cowan's RG work on the percolation universality class for his famous model of spiking neurons A decade later, I think I have figured out what is a reasonable description of an RG flow for the eigenvalues, including a proper description of the proper gauge condition, and a way to implement it as an optimizer extension Comments welcome weightwatcher.ai/rg_theory_webp…

@CalcCon @adamzwasserman the concept self organized critical assembly is one of the one of the foundational principles in my project github.com/ShawnaRStaff/C…

@ShawnaStaffReal @adamzwasserman Cool! It is also one of the foundational ideas where the weightwatcher project started

@MFinChina @pfau @JundeMorsenWu Those of us who are armed.

@CalcCon @pfau @JundeMorsenWu Who is "we"?