Dhiraj

@Oommttl @anuragwho There is no need to go for mythos like model at the moment but we definitely should develop models as capable as Chinese are producing.

@anuragwho @IndustrlPolicy Your views?

There is a lot of chest thumping happening around this, claiming that India should also aim to build its own models, which just doesn't work. we are just nowhere capable of doing this today. Anthropic recently spoke about their new model called Mythos, which is setting a whole different benchmark on what a model should be, and it cost $10 billion to train that model. Reliance, which is our largest conglomerate in India, just did $10 billion of profit for the whole year. the big 4 tech all generate more than $100bn FCF. It is nonsensical to say that we should just pool in enough capital and make such a model happen! The Big 4s are spending $700 billion this year in capex for models. Even if we want to, there is no way for us to play in this game! What @NandanNilekani is saying is correct. We still have to have a meaningful role here in the capacity that is possible for us, where there is room for success for us, where we have the ability to solve our problems, train a massive population of our people into being AI-first, and that's where real value will emerge. Someday we will also have our models, but just to say today, maybe Infosys should invest and make it happen. India doesn't have that kind of money yet.

A 2005 state-designed worm designed to corrupt physics simulations sat undetected on VirusTotal for nearly a decade. Fast16, intercepted executable files at the kernel level and silently rewrote floating-point calculations to make them produce slightly wrong answers. Targets: high-precision engineering suites used for structural analysis, crash simulations, and physical process modeling, including LS-DYNA, a tool cited in reports on Iran's nuclear weapons research. The sabotage vector relied on deployment of the driver across a network via worm, corrupting calculations on every machine, and eliminating the possibility of cross-checking results against a clean system. Stuxnet got the documentary. Fast16 got twenty years of nothing. sentinelone.com/labs/fast16-my…

"A Harvard-led study published in Science found that OpenAI's reasoning AI model outperformed human doctors on emergency room diagnosis and clinical management tasks." ndtv.com/science/ai-dia…

Shouldn't a capable Chinese open-weight model and fine-tuned on targeted cybersecurity content be able to produce comparable offensive capabilities? And Chinese capabilities should be at par, my guess is.

Any model pretrained on internet which obviously includes a lot of cybersecurity material and with agentic capabilities with minimal censorship should already be capable of decent offensive cyber ops like Mythos? & specific fine-tuning on content could elevate capabilities.

"Xi urges more efforts to boost basic research... Realizing sci-tech strength through groundbreaking innovations...Xi emphasized the need to seize opportunities, rise to the challenges, prioritize basic research" chinadaily.com.cn/a/202605/01/WS…

Indians proved their technical prowess during Y2K that put Indian ITes on export map and it little to do anything with ambition to immigrate.

Kerala is not a good model anyway, yes state is beautiful but nothing more than that. Wrote about elite stem brain drain here: renaissant.substack.com/p/the-hidden-c… Economist article: economist.com/finance-and-ec…

"Can countries grow richer by exporting people, not goods?" For large countries - no. Exporting labor from the non-tradable sector eg construction workers,domestic helpers,caregivers etc is mostly fine as remittances flow back without meaningful loss to domestic productivity

Important new report from @SilveradoPolicy “China is the leading global producer of display cells and is projected to account for 75 percent of display production capacity in 2028.” ‘Displays are essential for defense applications, consumer electronics, and U.S. manufacturing. As a result, any disruption in the U.S. display supply chain could lead to a reduction in production in major manufacturing sectors, such as U.S. auto production, or limit the ability of U.S. industry to produce goods critical to national security.” reuters.com/world/china/ta…

China is using everything from AI to a massive database of DNA from 33,000 children in an effort to overtake the US as the world’s No. 1 drug developer. bloomberg.com/news/newslette…

The Indian government is negotiating with the US for access to Anthropic's Mythos AI model to protect critical infrastructure. livemint.com/technology/tec…

Google agreed to let the Defense Department use its artificial-intelligence tools in classified settings, becoming the third AI company to reach such a deal in recent weeks and accelerating the Pentagon’s push to embed the technology in its operations. on.wsj.com/4u7V9OO

1/ The U.S. faces a paradox: scale fast by inviting Chinese investment and manufacturing know-how into U.S. clean-energy sectors, or slow down to guard against security risks. A new @BrookingsInst series explores options for navigating this trade‑off. 🧵👇 brookings.edu/collection/bet…

UAE to leave Opec in blow to oil cartel ft.trib.al/7uAL6s8

China Says Hostile Foreign Forces Are Driving Its Youth to Slack Off—Country’s spy agency urges young people to work hard and resist ‘lying flat’ @hannahmiao_ wsj.com/world/china/ch… wsj.com/world/china/ch…

Good breakdown of Press Note 3's recent changes. Bottom line is that it has been largely defanged because pragmatism won over excess knee jerk jingo diktats and acceptance regarding china's supply edge that cant be wished away. carnegieindia.org/india/research…