Dev Nag
@devnag
Founder/CEO, @TryQueryPal. Previously Founder/CTO, @WavefrontHQ (funded by @Sequoia, acq by @vmware). Oregon-born, universe-raised
Vous pourriez aimer
(1/n) Since its publication in 2017, PPO has essentially become synonymous with RL. Today, we are excited to provide you with a better alternative - EPO.
lowkey i think ilya 30u30 needs an upgrade to 50u50 now btw if you go through all of these, you’ll know at least 70% of what matters today
In this thread I'll record some brief impressions from trying to use o3/o4-mini (the new OpenAI models) for mathematical tasks.
The financial services industry is uniquely positioned to advocate for collaborative AI leadership, given its vested interest in trust, transparency and global cooperation, writes @DevNag of @TryQueryPal, in @AmerBanker @BankThink.bit.ly/4jhAgvE
pretty mind-blowing fact I just learned about transformer language models: the positional embeddings don't really do anything. you can just get rid of them and the model still works just as well sounds impossible, doesn't it? turns out standard LLMs aren't actually…
Simply, no. I've been looking at my old results from doing RL with "verifiable" rewards (math puzzle games, python code to pass unit tests) starting from 2019 with GPT-1/2 to 2024 with Qwen Math Deepseek's success likely lies in the base models improving, the RL is constant
Is it feasible to do a true tabula rasa version of deepseek R1 zero, starting from an LLM with random weights, similar to alpha zero? Or is starting with an LLM which is pre trained on math required?
Another key reason people are spooked: around 2016ish we started seeing the *insane* power of purely self-improving Reinforcement Learning (RL) (think AlphaZero going from no knowledge to superhuman at chess in hours), and it was formative for a lot of folks, in terms of their…
The bandwidth of this single chip is the entire internet’s traffic??? WHAT We are no where close to seeing the top of intelligent systems
Jensen Huang shows off the NVIDIA GB200 NVL72: a data center superchip with 72 Blackwell GPUs, 1.4 exaFLOPS of compute and 130 trillion transistors
Ladies, has your Christmas vacation been ruined by the Deepseek launch? You may be entitled to compensation.
It's a bit sad and confusing that LLMs ("Large Language Models") have little to do with language; It's just historical. They are highly general purpose technology for statistical modeling of token streams. A better name would be Autoregressive Transformers or something. They…
the cathedrals in code and hardware are invisible to people. those who deeply understand that know the wonder that they hold in their hands to post this tweet instead of a slab of glass and metal.
Timing.
Left: Hierarchical model based RL with a large-scale pre-trained world model, auxiliary tasks and skill-discovery and a model for inverse kinematics. Right: PID
How the ‘Human Search Engine’ Trap Killed Productivity thenewstack.io/how-the-human-… @devnag @TryQueryPal #Sponsored #HumanSearchEngine #Productivity
thinking: as execution of ideas gets easier (thx to agents, APIs for every imaginable service, etc), ideas become more of the differentiator. good ideas aren’t derived solely from logic or patterns of the past, they’re also the exhaust of human experiences and traumas, mistakes…
🌟 Excited to be featured in the AI Tool Spotlight by @EverydayAI_! Check out the feature here: read.youreverydayai.com/p/sam-altman-c…. 🚀 Follow us for the latest updates and install QueryPal for FREE today at querypal.com!
🎉 Exciting News!! We're thrilled to announce that QueryPal clinched second place as Product of the Day on @ProductHunt and soared to the top as the #1 SaaS Product! Plus, we were featured in the Product Hunt newsletter, spreading our reach even further. A heartfelt THANK YOU to…
The last time Dev worked at a Pal company, it was PayPal 👀
Human attention is the scarcest and most valuable resource in the world. This is why we built QueryPal, launching today on Product Hunt (t.ly/6YXfm). It's an AI chatbot that automatically answers all those incoming repetitive questions at work. (1/10)
United States Tendances
- 1. #BaddiesUSA 61.6K posts
- 2. Rams 29.6K posts
- 3. TOP CALL 3,411 posts
- 4. #LAShortnSweet 22.2K posts
- 5. Scotty 9,957 posts
- 6. Cowboys 101K posts
- 7. Eagles 141K posts
- 8. Chip Kelly 8,790 posts
- 9. #centralwOrldXmasXFreenBecky 288K posts
- 10. SAROCHA REBECCA DISNEY AT CTW 303K posts
- 11. sabrina 62.8K posts
- 12. #ITWelcomeToDerry 15.8K posts
- 13. AI Alert 1,017 posts
- 14. Raiders 68.1K posts
- 15. Market Focus 2,263 posts
- 16. Vin Diesel 1,324 posts
- 17. #RHOP 12.4K posts
- 18. Ahna 7,368 posts
- 19. Stacey 24.6K posts
- 20. Stafford 15.3K posts
Vous pourriez aimer
-
Hemant Mohapatra
@MohapatraHemant -
Kritika Prakash
@kritipraks -
Victoria Krakovna
@vkrakovna -
Tara Viswanathan
@TaraViswanathan -
Josh Gordon
@random_forests -
Andrew Carr 🤸
@andrew_n_carr -
Miles Brundage
@Miles_Brundage -
Carlos Guestrin
@guestrin -
Aaron Defazio
@aaron_defazio -
Clement Pang
@panghy -
Nicola De Cao
@nicola_decao -
Dipanjan Das
@dipanjand -
Javier Ideami
@ideami -
Ross_Kukulinski.yaml
@rosskukulinski -
Bryan McCann
@BMarcusMcCann
Something went wrong.
Something went wrong.