@sarahookr ML 101 “Garbage IN, Garbage OUT”. Studying the data, setting up proper evals and appropriate baselines are crucial for algorithmic development. If someone does not understand data and develop intuitions, it’s quite hard to train an effective model.
@ShashiTharoor@Ram_Guha Once in a while, Shashi makes comments that make you doubt his IQ. This is one such moment. Not only has the pappu led his party to historic lows, he has no administrative experience at all. CM kf Gujrat had more than a decade of exp. Dont be intellectually vacuous!!
@snowmaker That comparison is intellectually lazy.
Raw LOC output was never the bottleneck in software engineering: correctness, architecture, debugging, maintainability, security, and product judgment are. If LOC were the metric, half the internet’s codebases would be masterpieces.
@vijayshekhar World models are still nascent.
Greg probably refers to Codex and computer use, apps powered by frontier LLMs.
If any company’s moat is the prompt, they will be eaten alive by the revenue hungry model providers.
@ShivAroor Shiv, dont you think the govt needs to do a better job at setting the correct narrative? You folks have seen this during op sindoor as well, if there is one thing the govt needs to get right is communication!
Fuel price rise in india is actually negligible compared to the West
@PadmajaJoshi@rahulkanwal@manogyaloiwal@SreyashiDey@TanushkaDutta@Pooja7arya It was an awesome coverage. Staying outside India, this felt like one of the few reports that actually captured the ground reality instead of the usual sanitized studio narrative. Your walks through those familiar streets made it feel deeply personal, nostalgic, and real.
@NandanNilekani@rvenk@EconomicTimes Totally dumb opinion !!
Same as saying India doesn’t need manufacturing but needs to ensure manufactured products are widely distributed..
@soumithchintala@dwarkesh_sp Interesting how Jensen Huang pointed out that frontier AI isn’t only chips, its a full-stack cocktail spanning compute, energy, data, and https://t.co/fsHtnC23ce many ways, constraints in one ingredient can be offset by scaling another