All 12 modules have been written. I wrote up some thoughts behind this. kcsrk.info/ocaml/oxcaml/teac…

It is easy to pick on lack of hard-tech because it's so glaring. But consequences of abundant intelligence will make freedom to innovate, to build business, to take things to market and move money easily, for the need for infra (whether roads or judges) - much much more important - in ways that are not glaring right now. Consider an analog: - Before invention of motorized transport - human movement over land was limited by speed and capacity of horses/bullocks. - That meant some parity across nations in moving people/goods. horses just needed dirt tracks. everyone had dirt tracks (except rain-forests). - Once motorized transport came in - we suddenly needed asphalt and rails. cross country highways. These were largely public works. Those who built them were able to take true advantage of motorized transport. Something similar will happen with AI - Earlier the bottleneck in 'production' of many things was the intelligence and human time required to build it. A business plan, designers/architects, hiring, putting an office/factory together, coding/executing, marketing etc. etc. - What super-abundant intelligence will do is compress some of these steps to near zero. (and we haven't even entered the era of abundant robots). - When some stages get compressed - the bottleneck will shift to remaining stages. The overhead of GST/MCA filings, commuting, tax audits etc may have been 1/10th of an earlier execution path length. As path length shorten - maybe they become 1/5th. (see Amdahl's Law) Countries where regulatory burdens are minimal and who can provide smooth infrastructure businesses need, will prosper - as new products/services are launched with breathtaking speed. Countries who are stuck in bureaucratic past, with poor infrastructure - will find that their economic gap from more business friendly countries - is increasing. Because their overheads are becoming even more of a bottleneck than before. All this is not visible and in the news. What is in the news is the LLM, the GPU. These are the Steam Engine, the Mercedes, the Ford of our era. But what will matter equally eventually, or perhaps even more, is the Rail and the Asphalt.

the FUD on training economics in India is borderline harmful You don’t need multi billion clusters.. american open source leader has trained their 400B model by raising ~$50M (they have since then raised a bigger round) we should celebrate competence not jingoism

In a few years you’ll be able to have a chip lab in your hostel room

anyone over the age of 40 or has made money by labour arbitrage should automatically be disqualified from influencing AI policy in India if you have not spent >5% of your revenue for R&D in any organisation that you’ve headed once again you are disqualified

Sorry sir. With due respect, Indian tech titans sat on their hands for almost a decade now and even professed that India can just build around applications and build its moat. Now when sovereign imperatives demand, the same tech titans seeking the driver’s seat is untenable.

In light of the news of the US export licence on Anthropic's Fable/Mythos, I'm re-upping my op-ed, in which I argued that India must seek digital sovereignty in the form of open AI models because access to proprietary models can be revoked at any time. pranesh.in/press/mythos-self…

CEA V. Anantha Nageswaran writes about the multiple possible causes for India's underspending in R&D. Significantly increasing this spending is critical for India. We need to look at all these factors and address them.

The cynic in me feels like this is orchestrated by Anthropic to get more publicity/hype. Or someone found a way to break it or it’s not as good as they claim. Pretty sure the “order” will be rescinded after the IPO.🙄

Now we know why they named it fable and myth 😀

I made a PCB business card that measures your heart rate: CARDio

We fabricated and characterised our first semiconductor devices @hackerfabindia today- a PN junction(right) diode and a MOSCAP(left)- both using tools built entirely in house. There’s obviously a lot of improvement to be made to our process, but it’s only upward from here!

Every few months, India falls in love with a new sector. Semiconductors are the future. Then green hydrogen. Then drones. Then electric vehicles. - Policy announcements with billion-dollar incentives. - States competing for factories. - Television debates buzzing with the word “ecosystem”. But industrial success is not a reality show where governments pick winners and wait for applause. Industrial policy works best when governments look beyond the excitement of emerging sectors and: - start obsessing over the plumbing underneath them - start focussing on building durable national capability In the piece below, @anujg argues — India’s Industrial Future Depends on the Boring Stuff swarajyamag.com/commentary/i…

The way things will turn out (most probably): everybody tries to go "fast" with AI, nobody gains any advantage (what you can do with AI your competition can also do, provided they have built the proper engineering/management structure). But if you don't go fast, you will be left behind. Burn tokens so that you are not left behind. Nobody wins other than the frontier labs (and the consultants who will help you in "AI transformation" 😀)

Still remember my "engineering workshop" days ... the CS students felt it to be beneath their level and an utter waste of time, after all we were the elite people who would sit in AC rooms and operate computers, not get their hands dirty doing sheet metal work or welding. The workshop staff also had similar feelings and were more than willing to complete the assignments for us, especially for the girl students 😀It was also fairly easy to pass the exam, as you just had to demonstrate some minimal familiarity with the tools. The workshop would have been considerably more useful had the faculty enlightened us about the interdisciplinary nature of engineering and technology and also structured it around small meaningful projects. Coming to the present: in the light of the massive transformation of our technology landscape (from one that is driven by low end software services to one that is based on modern manufacturing and R&D), the Engineering Workshop will have to be re-imagined and brought into a central place in the engineering curriculum. Are the decision makers listening?

What I genuinely don't understand is this: the LLM's are built and tuned according to theoretical frameworks that humans have constructed. So where exactly is the secret sauce? Does an element of unpredictability magically lead to latent capabilities that are beyond our understanding? Maybe we don't yet have mathematical apparatus sophisticated enough to analyse highly unpredictable natural systems, but don't we have the tools required to analyze algorithmic machinery that we ourselves have created? If we don't have the tools to understand the capabilities of these systems, how are the frontier labs able to increase the capabilities of these systems dramatically? Where is this element of magic that is "beyond our understanding"?

A unique UG program at @IITKanpur! Meant especially for young ethical hackers. (Some of them are in the news 😊.) Admission through a hackathon. Specialized coursework with two year long internship at security agencies. We aim to produce cyber warriors of the future.

30k laptops may not last 4 years. A desktop build with a low end ryzen cpu (integrated gpu), 16gb ram and nvme ssd should be around the same figure and will perform amazingly well with any Linux distro (and you can even cut down RAM to 8gb if needed). Whatever you are buying/building, if you are a CS / ECE student, make sure you run Linux on it and make it your primary computing environment!