I’ve gotten one B in my entire life.
Graduated Georgia Tech with a 3.97/4.0 GPA with a bachelors in Computer Engineering and a minor in Artificial Intelligence.
I slept 8-10 hours a night, never pulled all all-nighter
Was the president of my fraternity
And worked out 5-6 times a week.
Ask Me Anything.
Forget token budgets, what is your interns’ electron budgets (to create unlimited tokens)
they found the physical limits of our office space at @NVIDIAAI w one too many DGX stations
Fine, I'll say it
I believe all of you should have better, cheaper, intelligence
I believe we should continue to innovate and provide as much access to tools as possible
I believe we should enable our defenders of critical systems and infrastructure with the best tools
and that those who want to publish their research out in the open should continue to be able to do so, without restriction
Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community.
During the Hugging Face incident, closed AI blocked essential forensics. An open-weight frontier model helped contain the intrusion.
That’s why we created the Open Secure AI Alliance.
no one is saying that proprietary models (or software) shouldn’t exist
we are saying open weight models should ALSO exist
anyone who says otherwise is misconstruing the argument
@aryngupt it’s not about what i use.
open weight models are not only nice to have, they are required for use cases that require customization, data sovereignty, edge intelligence, etc
(defense, healthcare, robotics)
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
https://t.co/AUKzoQ5Ikb
The only AI company that works withe every other AI company.
When the world is abundant in intelligence, we all benefit.
It is already resulting in medical discoveries, innovations in robotics to keep humans out of harms way on the roads and in factories, and will help every knowledge worker in the 120T global economy get more done!
The best way to accelerate this? Work together, as an industry. Build on the work of others and contribute back your learnings.
@NVIDIAAI is the largest contributor on THE place to share Open Source AI artifacts, @huggingface .
This is the way. Intelligence for all.
The first-ever measured silicon numbers for @NVIDIA Vera Rubin NVL72 are in 😲
First measured performance shows 10x more tokens per megawatt than Blackwell.
No projections. Real results from live hardware.
“tomorrow is not guaranteed, live life while you are young”
frontal lobe development litmus test
the odds you are around tomorrow is very very high
in fact, the odds you will be alive to pull from retirement accounts is 90%
9 out of 10 times older, more tired you will wish you lived life while young AND invested for your future
in the 1/10 times where that is not true, you helped to take care of those who rely on you that you left behind
Open source creates the necessary environment for competition
Without competition, only incumbents have power.
We saw this first happen with Stable Diffusion.
Open AI released Dall-E. In the name of safety it would not render faces and it was extremely limited via waitlist. I only got access because I was a YC founder.
Then Stable Diffusion came out. It rendered faces. No wait list. Simply download the model and run it on your hardware.
In response to this competitive pressure, open AI released DALL-E 2. No waitlist. Renders faces, added in painting, etc
Same thing happened with reasoning models. OpenAI had o1, which at the time, seemed to give OpenAI an insurmountable advantage over everyone else.
Deepseek released a reasoning model, open weights, which led to a proliferation of reasoning models coming from research labs.
Same pattern again with Mythos and Fable.
Closed model, extremely limited. This time worse, because of nonsensical fear-mongering.
GLM 5.2, Kimi K3 both get released. Frontier Intelligence, open weights.
But now the calculus is different:
The fearmongering has inspired distrust in the closed-source model labs. If we can't trust continued access to the tools then why would we learn to use them? Closed source labs open sourcing harnesses is a great way to build trust.
The ecosystem is only healthy when we have both. Competition is crucial to keep the ecosystem fair. Open source is necessary for competition.
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
https://t.co/Ghebq5mG30
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
fragile organizations are not ready for the onslaught of AI slop that is starting to happen
movement != progress
too easy for people to turn off their brain and produce insane amounts of volume of (largely ineffective) work
this is a pretty lethal trifecta
1. wasting other people's time
2. employee's own salary
3. their consumption of AI tokens is not free
previously, the "quiet quitter" was net neutral. This person is actively harming an organization.