@GauravMLi
iAccount based inUnited States!
About this account
- Account based in
- United States
- Connected via
- India App Store
! X says this location may be affected by a proxy or VPN.
Account-level information from X, not a live location or the device used for a specific post.
Senior Staff Tech Lead Manager @ Google Research. Author of the Efficient Deep Learning book. Opinions are my own, obviously. Memento mori.
Bay Area, California.
Joined June 2019
- Tweets636
- Following1K
- Followers1K
- Likes22.4K
Pinned Tweet
Excited to share a survey of the vast landscape of efficiency in deep learning on how to make your deep learning models smaller, faster, and better!
arxiv.org/abs/2106.08962 (1/n)
I'm not entirely sure how to navigate the trend of technical design docs being vibed.
The information density, the odd writing tone, and the standardized presentation style across all the vibed docs generated from the same model are making technical artifacts human un-readable
One of the things I admire about Google is how people love to dunk on it, and the company just keeps putting in the work, and lets the results speak for themselves.
I was chatting with my buddy at Google, who's been a tech director there for about 20 years, about their AI adoption. Craziest convo I've had all year.
The TL;DR is that Google engineering appears to have the same AI adoption footprint as John Deere, the tractor company. Most of the industry has the same internal adoption curve: 20% agentic power users, 20% outright refusers, 60% still using Cursor or equivalent chat tool. It turns out Google has this curve too.
But why is Google so... average? How is it that a handful of companies are taking off like a spaceship, and the rest, including Google, are mired in inaction?
My buddy's observation was key here: There has been an industry-wide hiring freeze for 18+ months, during which time nobody has been moving jobs. So there are no clued-in people coming in from the outside to tell Google how far behind they are, how utterly mediocre they have become as an eng org.
He says the problem is that they can't use Claude Code because it's the enemy, and Gemini has never been good enough to capture people's workflows like Claude has, so basically agentic coding just never really took off inside Google. They're all just plodding along, completely oblivious to what's happening out there right now.
Not only is Google not able to do anything about it, they don't seem to be aware of the problem at all. I'm having major flashbacks to fifty years ago as a kid at the La Brea Tar Pits, asking, "why can't they just climb out?"
My Google friend and I had this conversation over a month ago. I didn't share it because I wanted to look around a bit, and see if it's really as bad as all that. I've been talking to people from dozens of companies since then. And yeah. It's as bad as all that.
Google is about average. Some companies at the bottom have near-zero AI adoption and can't even get budget for AI. They may have moats and high walls, but the horde is coming for them all the same.
And then there are a few companies I've met recently who are *amazingly* leaned in to AI adoption. One category-leader company just cancelled IntelliJ for a thousand engineers. That's an incredibly bold move, one of many they're making towards agentic adoption. In my opinion, that company is setting themselves up for a _huge_ W.
As for the rest, well, it's the Great Siloing. Everyone's flying blind. With nobody moving companies, no company knows where they stand on the AI adoption curve. Nobody knows how they're doing compared to everyone else.
Half of them just check a box: "We enabled {Copilot/Cursor} for everyone!" Cue smug celebrations. They think this is like getting SOC2 compliance, just a thing they turn on and now it's "solved." And they don't realize that they've done effectively nothing at all.
All because of a hiring freeze.
Google Home's voice recognition needs to improve a bit. My parents wanted to listen to some Indian spiritual music, and Google Home ended up playing Enter Sandman. Not kidding.
It is such a joy to co-read dense papers with Gemini!
Jumping to random points, recapping something defined several sections earlier / if a particular idea was tried in prior work / getting a quick refresher on a topic that I have either not learnt or have foggy memory of, etc.
Gaurav Menghani retweeted
Nvidia played a 30+ yr game but not enough has been written about the 10+ yr game Google has played with TPUs. I still remember the 2016 Google i/o when it was launched. The internal funding for it was basically cobbled together from other failed projects that had leftover $s, and the core TPU team ran hard for 15mos from concept to tapeout which is incredibly fast. Even the core reason why TPUs were built in the first place was interesting - to support voice-first use cases on Android e.g. maps. Google was building for a voice first world a long time ago.
My view on specialized ASICs remains the same: I think general purpose intelligence needs general purpose compute. But ASICs will continue to play a critical role in everything else specialized.
Gaurav Menghani retweeted
A fun conversation with the amazing @shripati and his team at @Primevp_in! Thanks a lot for hosting me.
What happens when one of @GoogleDeepMind's top scientists sits down to unpack AI’s past, present & future?
The full episode with @jainprateek_ is here. 🎙
Topics you can’t miss:
🔹 Deep learning → transformers → generative AI
🔹 India’s once-in-a-generation chance to lead in deep AI research
🔹 The next big bottlenecks — and opportunities — in this space
▶️ Watch the full episode 👉 📺 YouTube: youtu.be/7pCILXw2_UY
🎧 Spotify: open.spotify.com/episode/7zb…
Hosted by @shripati
#ArtificialIntelligence #DeepLearning #GenerativeAI #GoogleDeepMind #PrimeVenturePartnersPodcast
New post on 'Noam Notation' or Shape Suffixes. Just a quick way to make modeling code slightly more readable. Give it a read, and leave a comment if I missed something / got something wrong.
blog.gaurav.ai/noam-notation…
Doing something new: I will be writing about algorithmic efficiency techniques every other week or so. Starting off with the the dynamic duo of KV Caching and KV Sharing. Feel free to give it a read, and drop a comment if I missed something.
blog.gaurav.ai/2025/08/05/kv…
Gaurav Menghani retweeted
Lina Khan deciding if you should be permitted to sell your startup or be forced to deliver additional shareholder value.
Finally took the time to transition my personal website and blog to GitHub Pages along with the domain. Also turned off my private server after nearly 11 years of managing everything by myself. Not only is GitHub Pages convenient, it is free! What a delightful experience :)
Gaurav Menghani retweeted
I don’t think he’s ever told the story, but it’s worth telling. When we were selling @Behance to Adobe many years ago, @scottbelsky made a spreadsheet of every employee (32 of us at the time) and personally negotiated each persons title, salary and incentive structure, and made that part of the overall deal terms. I heard the phone calls where he went to bat for each individual.
He not only didn’t have to do this, but it actually complicated some of the other factors in the deal. It changed the trajectory of so many people’s lives, including my own.
2 years later, 100% of that original team was still at Adobe. Even today, a dozen years later, many of the core members are still there, building.
I was inspired by it then, and I'm inspired by it now.
Gaurav Menghani retweeted
Want to learn about the research behind Gemma 3n?
Altup - arxiv.org/abs/2301.13310
LAuReL - arxiv.org/abs/2411.07501
MatFormer - arxiv.org/abs/2310.07707
Activation sparsity - arxiv.org/abs/2506.06644
Universal Speech Model - arxiv.org/abs/2303.01037
Blog - developers.googleblog.com/en…
Proud to have contributed to Gemma3N via LAuReL (arxiv.org/abs/2411.07501).
developers.googleblog.com/en…