Pinned Tweet
I love how India is becoming a full blown fascist state. But not in ways that the west is accustomed to thinking.
IIT-Bombay reiterates guidelines to students
"Dont ask JEE Rank" strict punishment awaits. These are your young kids we are talking about.
i.e. ignore the elephant in the room.
DEIMAXX in IIT.
hindustantimes.com/india-new…
We'd never be able to do it because Cricketers use Mittens for baking...
Hence we can only do this.
youtube.com/watch?v=G89g3xTr…
Pradeep Gyanegowda scored 97% in 12th yet had to go to Ukraine to study MBBS because he was quota'd out & was k3lled during early days of the Ukraine/Rus war.
All news channels removed their microphones the moment his father mentions it.
Everyone knows the truth.
He's still worth around 2 billion USD with his investment and whatnot.
Sabeer Bhatia was co founder of Hotmail where he developed the business plan and helped raise investment. Jack Smith was real man behind idea for browser-accessible email called Hotmail.
Sabeer and Jack sold Hotmail to Microsoft for 400 millions
Jack joined Microsoft and Bhatia took the money.
Both bhatia and sabeer tried various other business after Hotmail, jack went on making money whereas bhatia without jack failed in all his future projects
Today he is running low on cash and he is fully dependent on X payouts.
To earn payouts he started posting against Modi and Adani.
A Friend of Sabeer bhatia told me if Bhatia was millionare, he would not have the time to post 50 tweets in a day..
Have you seen any millionaire or billionare so free that he keeps posting 50 tweets in a day
vikram retweeted
The deep frustration is evident from one man who has never failed to put money where his mouth is.
When we need a thousand Rajeevs, we are failing even one.
The entire Indian system is geared towards *prevention*.
*Prevent* all the bad things from happening, because recourse to cure is so poor. Why is there no cure? Courts and police are total failures - for instance, if somebody commits a financial fraud, it can take decades to prosecute them. Hence prevent the fraud - throw multiple OTPs at the user for a Rs 5000 fund transfer. Banks make you do KYC like fanatics because they are trying to *prevent* the last bad actor from getting away - if millions are to be tortured for it, so be it.
This prevention-first mindset throws sand in the wheels for everything. It makes life hell for consumers and small businesses.
Just imagine what the accumulated cost is of preventing all the bad things from happening - surely it shaves off many basis points from GDP growth when you stack up all of it.
And yet the cheats, criminals, troublemakers and frauds all find ways to do what they do anyway. Those who are just trying to run their business and get through life are burdened with this mountain of compliance and paperwork. None of this moves babus and ministers - they live in a parallel reality where somebody else works (at taxpayer expense of course) to remove these frictions from their existence.
Then they blithely bark at poor consumers and business operators: “ek form hi to hai, bhar do”. But of course it is never one form - every leech in the system demands a form and a submission. Stacked up, it goes into hundreds of pages of documents and dozens of tasks.
The associated parasite lobby of lawyers, consultants, company secretaries and chartered accountants makes merry. They lobby the government for more regulations and offer compliance services to hapless people. In exchange, they place the government elite’s nepo kids in jobs.
This parasite gang has a field day because more documents means more business for them. They are joined by economists and intellectuals, who want to join the leeching and bloodsucking bacchanalia. This intellectual cabal wants to snag a coveted role in the “world’s best” prize winner central bank or other sarkaari gig.
Productivity dies in India, businesses slow down and overall it is a slow-motion economic disaster. In isolation we feel happy to have 6-7% economic growth. Decades go on then somebody says - oh, how did Singapore, Vietnam and China do so well? We look at the delta and feel awkward. We justify it saying “we have democracy”. The super-cabal didn’t just prevent accidents or disasters like they claimed - they killed off and prevented growth and progress with it. Mega accidents and disasters happened anyway.
Enjoy the democracy. Praise the cockroaches and jholachhaps. Celebrate the thugs, parasites, leeches and bureaucrat-goons all around you.
Mera Bharat Mahaan!
It is so weird, and I hate to draw this analogy but whenever I see LLMs writing code or creating diagrams on LATEX. It feels like watching a Stockfish type chess engine working.
Its inelegant, its unintuitive, its an exhaustive search where the ends justify the means.
Some thoughts on AI:
Many people who say that LLMs are just next token predictors are missing a key point, which is that the LLM predictor is fit in a very high dimensional space (a billion or trillion dimensional space).
Think about predicting the curve y=ax²+bx+c. You need just 3 data points. Once you have the 3 points, you can predict any point on that curve with 100% accuracy. You can predict it for x=0, x=100, x=100000.
Language is fixed. Zipf's law tells you that the word "the" occurs 10% of the time. Meaning just one word is worth 1/10 of the English language. The entire English language can be represented in just 30,000 tokens.
Just like that, the space of tasks you can do on your computer is mathematically limited. There may be infinite combinations of commands on your terminal, but most of those combinations are pointless. There are only a few most commonly used commands. This "task space" is built using the language space as basis. It is a much larger space, optimization in this task space is the purpose of RLHF.
Same for code. Programming languages are deterministic. Design patterns keep repeating. The tokens if, int, return, while, for, void, {, } is probably 30% of C code.
The same goes for abstract math too. Many proofs and concepts have common ground.
The LLM can generate new knowledge by extrapolating in this space instead of interpolating, and then checking the extrapolated point with all the axioms and priors it knows (it is trained to do this).
But this is where it differs from a human. LLM extrapolation is very different from human extrapolation. A human extrapolates via intuition, or analogy that comes from physical experience. The LLM extrapolates by expanding outward into the task space. This can lead to incredible results like what we saw with the OpenAI ETH refutation paper. The LLM was able to brute force itself outward into a valley where a bizzare algorithm that has a massive pre factor (it's literally exponential for all practical purposes). This is also why mathematical results from LLMs are "ugly", they're not constructed via intuition, but via brute forcing. It's the difference between driving on a road versus driving through dirt patches to reach the same destination. A road route is memorable, you get to see many signs that tell you new places to go. A dirt road / off road route is going to be filled with trees and shrubs and you'll have no idea how you got there when you do. This is what Tao wanted to say when he said that humans solving a math problem uncover new theories that lead to new advancements, while LLMs just offroad drive to get there.
Much like how LLMs are breaking security, they're also "breaking math". OpenAI has discredited complexity theory by showing that you can brute force your way to an algorithm that will have an insane pre factor but will still be "subexponential". Likewise, the same goes for the Navier Stokes solution. The Agents managed to break Navier Stokes by extrapolating to a case which was bizzare but still legit, and it broke the Navier Stokes equation. This was always expected. NS does not describe reality. Real fluids don't have infinite velocities with finite energies. Neither are real fluids a continuum which NS assumes, all fluids are ultimately particles. Etc.
Another stream for corruption, bureaucratic paperwork, rent seeking & extraction.
vikram retweeted
India's female suicide rate is double the global average.
It is especially high in southern, more developed states.
Dandana et al (2023) find that it's mostly housewives, citing family problems. pdf.sciencedirectassets.com/…
I'd add, southern states also register high tolerance of gender based violence and divorce stigma,
Many women may feel trapped.
Everything, everywhere, all at once (blackpilled)
Read the set of tweets of @RMantri of the past 24 hrs.
It is the considered frustration of one who believed, continues to believe and put his money and skin in the process, seen success and failures in the country and wades through a broken system that continues to punish its strongest betters and innovators with the suspicion of a police treating a convict on parole.
Wouldn't this only make sense if NVIDIA literally stopped producing more GPUs?
Isn't a more likely theory that NVIDIA is deciding on future allocation based on purchase quantity of current gen chips? Even if there is no infra to support them for the next 36 months?
When you realize that "Celebrated Economist", "Famed Journalist" & "Renowned Historian" are all just globalist fiction.
One in four iPhones is made in India, but imported parts and costly subsidies complicate the success story. @andymukherjee70 takes apart an iPhone to explain why copying China's manufacturing model may not pay off 🎥
vikram retweeted
In case anyone missed this, OpenAI has proved the Kakeya conjecture in all dimensions.
Hong Wang got the Fields medal in 2026 for proving the Kakeya conjecture in dimension 3.
In fact OpenAI appears to have committed, in mathematics terms, what may be fairly termed a massacre, proving a far stronger result. It has proved the local smoothing conjecture, which sits at the very top of the tower of conjectures as described in the Quanta article in the reply.
I am not an expert in this area -- so happy to be corrected -- but the methods appear to be PDE based and basically bypasses a lot of work of the last two decades. This is the equivalent of Arbreans showing up to earth with an everything killer and explaining to us we have been doing warfare all wrong.
Announcing three major scientific discoveries, one each in mathematics (harmonic analysis), computer science (complexity theory), and physics (quantum field theory).
AI starting to help us improve our understanding of the universe! openai.com/index/building-sc…
They have been pretty clear of why they're anti-dehat. They clearly stated that they're anti tribalistic, inward looking and parochial.
I don't know why this is even such a hot button issue.
Boomer uncles need to stop self-inserting themselves into every little minutae.
vikram retweeted
Day-1 taunt host nation on an Indian owned steel plant
Day-2 ask Indian origin ppl to come back to build in a broken system
Day-3 Market Indian labour as indentured labour in so many words. :)
If there’s an Olympics for insensitive gaffes…
Idhula JD Vance sonnadhukku aiyayoo vera ammama vera!!
