Subho Majumdar | শুভব্রত মজুমদার

In an op-ed co-authored for Deccan Herald, Prof. Subho Majumdar, IIMB, assesses whether India can extend ISRO’s philosophy of “build it yourself; make it work” to the fast-paced terrain of #AI using the launch of IAIRO as a point of departure. 🔗Read here: deccanherald.com/opinion/isro-t…

@arienkaran @varunachar @PrinSciAdvOff ^This is exactly correct. Without alignment we don't have reliability, security, and safety. Just like shipping software without these things is unthinkable, so should be agentic systems and models that power them.

Respectfully — one last time, and then I'll leave this here. Yes, the policy mentions multimodal. Yes, companies are building. Nobody disputed that. The argument was about whether the scale, speed, and strategic coherence of what is being built match the moment. A policy mentioning multimodal and a sovereign multimodal capability is not the same thing. On agentic — you addressed the tooling. The concern was about the intelligence layer underneath it. Those are different questions. But the third point — "models don't need to be aligned" — that is where I'll stop and simply say: that single statement tells me we are not having a technical disagreement. We are having a values disagreement. Alignment is not a bureaucratic checkbox. It is the question of whether an AI system — operating at scale, embedded in governance, defence, healthcare, and financial infrastructure — behaves in accordance with the interests of the people it serves. Dismissing that as unnecessary is not a technical position. It is a dangerous one.

𝐀𝐬 𝐩𝐚𝐫𝐭 𝐨𝐟 𝐭𝐡𝐞 𝐨𝐧-𝐠𝐨𝐢𝐧𝐠 𝐀𝐈 𝐏𝐨𝐥𝐢𝐜𝐲 𝐖𝐡𝐢𝐭𝐞 𝐏𝐚𝐩𝐞𝐫 𝐒𝐞𝐫𝐢𝐞𝐬, 𝐭𝐡𝐞 𝐎𝐟𝐟𝐢𝐜𝐞 𝐨𝐟 𝐭𝐡𝐞 𝐏𝐫𝐢𝐧𝐜𝐢𝐩𝐚𝐥 𝐒𝐜𝐢𝐞𝐧𝐭𝐢𝐟𝐢𝐜 𝐀𝐝𝐯𝐢𝐬𝐞𝐫 𝐭𝐨 𝐭𝐡𝐞 𝐆𝐨𝐯𝐞𝐫𝐧𝐦𝐞𝐧𝐭 𝐨𝐟 𝐈𝐧𝐝𝐢𝐚 𝐫𝐞𝐥𝐞𝐚𝐬𝐞𝐬 𝐚 𝐰𝐡𝐢𝐭𝐞 𝐩𝐚𝐩𝐞𝐫 𝐨𝐧 “𝐀𝐝𝐯𝐚𝐧𝐜𝐢𝐧𝐠 𝐈𝐧𝐝𝐢𝐠𝐞𝐧𝐨𝐮𝐬 𝐅𝐨𝐮𝐧𝐝𝐚𝐭𝐢𝐨𝐧 𝐌𝐨𝐝𝐞𝐥𝐬. The versatility of Foundation Models makes them a critical layer of today’s AI ecosystem and a key area for innovation in India. Therefore, developing indigenous foundation models is a strategic priority. India’s objective is to harness foundation models for inclusive growth and public good, while ensuring they are governed in a manner consistent with the country’s values, legal framework, and security interests. This white paper provides an understanding of India’s approach to advancing indigenous foundation models through public–private collaboration and to governing these systems that support trust, accountability, and responsible adoption. The White Paper also provides details on India’s approach - which is centred on building indigenous capability across the foundation-model stack. Rather than relying on a single model, India is developing an ecosystem that combines (i) shared compute access, (ii) India-centric data and model repositories, and (iii) multiple model-building efforts across text, speech, multimodal, and sectoral systems. Read the White Paper here: psa.gov.in/CMS/web/sites/…

@BLRAviation @suhasinih It doesn't. Where's surmai fry and hyderabadi biryani?

@suhasinih This menu reflects the best of South and coastal India. What’s wrong about it? Have you eaten banana blossoms? It’s a staple. Heard about Kappa? Chakka? Or Kandhari?

Charred pineapple yogurt foam and banana blossom skewers?? Open secret in Delhi's diplomatic circles that visiting dignitaries now routinely return from such strange meals at official banquets and order Indian food from room service.

@abhay4798 @shravi_aj @SrinivasanSS52 As a recovering startup founder, I agree.

@shravi_aj @SrinivasanSS52 Actually initially they don't. Things change when they're able to scale and are successful.

I live in Bangalore, and let me tell you, it’s a bubble. It’s mostly the top 1% of people building things, taking care of their mind and body, solving problems, and genuinely working every day to improve their lifestyle. Every time I travel outside or go back home, I’m reminded how different the rest of the country feels. I meet people who make me realise how far we still are from understanding that taking care of yourself matters. Eating better. Moving your body. Setting fitness goals. Trying to grow. These things should be appreciated. Instead, they’re often mocked. People will eat all kinds of garbage daily because of zero self-control, and then laugh at others who actually put effort into eating clean and living better. My only point is: step outside this bubble once in a while. Go meet people outside your startup, gym, wellness echo chamber. You’ll realise how much unhealthy behaviour is normalised, and how deep-rooted these problems really are. This isn’t a quick fix problem. It’s decades of mindset change. By a large number of people. Slow, unglamorous work. That’s the reality.

In a recent article for Deccan Herald, Prof. @sbmisi, Decision Sciences area at #IIMB, sheds light on a double bind facing India’s civil services: 1300 vacant IAS vacancies and a perpetual churn of bureaucratic reshuffles 🔗Interested? Read here: deccanherald.com/india/how-vaca…

@thogge @amasad @ZohranKMamdani It's not. Demographic background is at best a confounding factor, but it's hardly causal in a "I'm X by culture so I'll vote for an X person" way.

@amasad @ZohranKMamdani i dont think its blind, and i dont think its the only criteria, but it seems very clearly true as a general observation. and yes, he was able to break out of that in a big way. extremely talented.

Congratulations to @ZohranKMamdani. We may differ on the economics, but I respect anyone who fights for their community with integrity and isn’t beholden to money or party machines.

@PrathameshD_8 Name and shame. This is unacceptable, especially from institutes of national importance.

I interned at an IIT and built the entire codebase for a research project. Just saw the paper published in a journal, with zero mention of me as an author or contributor. I did the work. They took the credit. This is beyond unfair. :D

Excited to be leading this workstream with @sbmisi!

🚨 AI Evals Crisis: Officially kicking off the Eval Science Workstream 🚨 We’re building a shared scientific foundation for evaluating AI systems, one that’s rigorous, open, and grounded in real-world & cross-disciplinary best practices👇 (1/2) evalevalai.com/research/2025/…

Red Teaming AI Red Teaming - arxiv.org/pdf/2507.05538… Red teaming has evolved from its origins in military applications to become a widely adopted methodology in cybersecurity and AI. In this paper, we take a critical look at the practice of AI red teaming. We argue that despite its current popularity in AI governance, there exists a significant gap between red teaming’s original intent as a critical thinking exercise and its narrow focus on discovering model-level flaws in the context of generative AI. Current AI red teaming efforts focus predominantly on individual model vulnerabilities while overlooking the broader sociotechnical systems and emergent behaviors that arise from complex interactions between models, users, and environments. To address this deficiency, we propose a comprehensive framework operationalizing red teaming in AI systems at two levels: macro-level system red teaming spanning the entire AI development lifecycle, and micro-level model red teaming. Drawing on cybersecurity experience and systems theory, we further propose a set of recommendations. In these, we emphasize that effective AI red teaming requires multifunctional teams that examine emergent risks, systemic vulnerabilities, and the interplay between technical and social factors. #AIRedTeaming #RedTeaming #AIsecurity #AIgovernance #SystemicRisks #EmergentBehavior #SociotechnicalSystems #ModelVulnerabilities #AIThreatModeling #MacroRedTeaming #MicroRedTeaming #CyberSecurity #ResponsibleAI #AIrisks #AdversarialTesting #AIresilience #SystemSecurity #MultidisciplinaryAI #AIdevelopment #AIoversightAsk

@icmlconf is around the corner! Are you presenting any papers or hot takes in Trustworthy ML? Share your work in this thread and we’ll retweet! 🚀

🚀 Technical practitioners & grads — join to build an LLM evaluation hub! Infra Goals: 🔧 Share evaluation outputs & params 📊 Query results across experiments Perfect for 🧰 hands-on folks ready to build tools the whole community can use Join the EvalEval Coalition here 👇

@gregisenberg If it's so easy, do it yourself.

I don't know why a bunch of band of vibe coders don't band together to vibe code free AI-powered alternatives to SaaS incumbents like Figma ($20B), Adobe ($16B ARR), DocuSign ($2.5B ARR) etc. This gotta be the most obvious $100M+/year startup idea of our time. So, legacy SaaS companies are sitting ducks. Their products were designed in a pre-AI world, with pre-AI economics. They have massive teams, expensive infrastructure, and shareholders demanding consistent growth. And more people than ever want to use these products, but they don't want to pay another monthly subscription. Meanwhile, a small team of vibe coders could band together tomorrow and build 90%+ of the functionality in weeks, not years and offer it for FREE to 99% of users. Thanks to tools like Replit, V0, Cursor, Bolt, Lovable etc and all these open-source models. And instead of charging monthly, you give it out for free, and charge power users per outcome. And you can layer on premium service to cash flow (I can talk about that on another tweet on what I'd do) What's holding people back? Is it the fear of taking on giants? The allure of building something "new" instead of "better"? Or people just like building on their own so they don't want to band together? I think the avg customer prefers a one-stop shop for all these free new AI saas powered tools. You don't need to raise VC if you don't want to. Every time I open one of these legacy products and see the same old interfaces with the same old limitations, I can't help but think we're at the beginning of a massive wealth transfer from incumbents to the teams bold enough to rebuild them from scratch. I'm not saying these incumbents are going to die btw. But I do think there's an opportunity (yeah a really big one) to peel off some of their sweet saas revenue and monetize it differently. Make it free for most. Nothing people hate more than a monthly subscription. Happy building.

@0xgaut Except that you're no Rick Rubin.

Can I code? No. Can I design? No. I have no technical ability and I know nothing about software. But I know how to prompt.

@emilyinvc I don't get it. Why is a coasting life bad?

to me, the greatest sadness of all is the banality of most people's lives i'll spend time with my non-elite friends, and it's clear that their lives are devoid of heroism and lore they lock in/clock in, but they fight no great battles, they dash no great foes they drop by a club or bar, but find no great loves there they travel the world, but it is by no means an adventure they return home, but they are not welcomed by a triumph – no, all they have are their succulents and the pile of dirty laundry they failed to do before they left

Along the lines of the stickers on my old laptop, my request to those who mean well: please be aware of your own privileges. please be an ally to those who don't share the same privileges. And never lose hope. Long live the revolution. /end

- moderate top-line economic growth - increasing inequality - significantly more rhetoric directed against disadvantaged communities and minorities 2/

Coming from a country that has become increasingly authoritarian since the regime change in 2014, the best case scenario for the US we'll see in the next 4 years is 1/

For those in the area, tomorrow (November 1, 2024) I’ll speak at two events at @JMU in Harrisonburg, VA. A roundtable hosted by computer science at 11:30 am and a talk hosted by philosophy at 4pm.

'Cultural prompting' feels quite on the nose. To build ethical AI, why not take care to actually curate diverse datasets and build better data pipelines? @CornellNews #EthicalAI #DiversityInTech news.cornell.edu/stories/2024/0…