Met a Meta AI researcher. He studied Physics in China, came to the US for a PhD in Physics, and then fell in love with AI, despite never having studied computer science. He watched Andrej Karpathy and Andrew Ng, bought a GPU, read every arXiv paper title daily, and dived into the ones that interested him. Eventually, he published a few first-author papers at top AI conferences without any supervisors, and later transferred to his university’s AI lab. You can just do things.

Nov 2, 2025 · 6:00 PM UTC

205
649
74
9,206
He also had one more reflection: AI is much easier than Physics. You don’t even need complex calculus for it.
43
41
8
1,569
Come join us making PixelArt. Almost anything is allowed. Normal art greatly appreciated but we won't judge you!

Start Painting

geopixels.net
3
8
3
223
Replying to @Yuchenj_UW
it kind of proves the point that if you've trained your brain in math, you can just do things physics PhD --> strong foundation for understanding ML from first principles
2
15
3
425
Reminded me of what Geoff Hinton said:
6
21
1
513
more replies
Replying to @Yuchenj_UW
You can just do things when you have a 140+ IQ.
5
2
179
You don’t need a 140+ IQ to do things. IQ is overrated imo, curiosity, passion, agency, and taste are far more important. All of these can be learned. IQ can’t.
13
10
1
260
more replies
Replying to @Yuchenj_UW
Proof that passion outpaces pedigree every single time.
1
1
87
curiosity, passion, agency, taste >> iq, degree
1
8
142
Replying to @Yuchenj_UW
The most important skill a person acquire from a PhD is not about knowledge in the field, but the ability to do independent research, no matter what field. So if he is smart and determined, he can read a bunch of articles about wine and start doing research in oenology.
2
53
Replying to @Yuchenj_UW
this isn't "you can just do things" - this is you must do certain things if you want to make money
2
44
€5,000 Winter Raffle! Do you like skiing? This is your chance to get Ski Austria Experiences & Bybit Skiing Gear. 👉 Link in profile bio
1
3
1
16
Replying to @Yuchenj_UW
I don't recall the source, but a famous CS researcher once said "Most of us are refugees from more difficult disciplines".
1
1
33
Replying to @Yuchenj_UW
You can just do things that’s the unicorn mindset
16
Replying to @Yuchenj_UW
Obsession is the 9th wonder I guess
1
1
8
Replying to @Yuchenj_UW
There is no comparison the AI math is too primitive. That signals that architecture isn’t where it should be yet
6
Replying to @Yuchenj_UW
Nice story. It reminds me of a mantra I used to have: “We have a strategy. It’s called doing things.”
2
Replying to @Yuchenj_UW
genuine curiosity and consistency still beat formal credentials. The best breakthroughs often start with just doing.
2
Replying to @Yuchenj_UW
TL;DR The bro used physics PhD funded by other people to get to the USA but neither had capability nor intentions to become a physicist.
2
Replying to @Yuchenj_UW
He had two things going for him Maths and doing independent research So he can excel in any field
1
GTO Wizard Pro @SamuSabu9 Samuel Mullur faced Linus Love on a stacked GGMillion$ FT. @tombos21 reviews two of his toughest spots using our 50k+ ICM sims — preflop to river. 🗒️ Full breakdown:
1
2
29
Replying to @Yuchenj_UW
Yess. You learn, you build, you ship.
1
Replying to @Yuchenj_UW
Amazing how one changes the trajectory of their life due to finding a new passion
1
Replying to @Yuchenj_UW
That's a fascinating journey, Yuchen! It's amazing how passion can drive someone to switch fields like that, isn't it?
1
Replying to @Yuchenj_UW
"despite never having studied computer science" yeah because it's a fake science
1
Replying to @Yuchenj_UW
A reminder that curiosity is the only thing you truly need to innovate
Replying to @Yuchenj_UW
CS students spend four years learning best practices, physics people buy a GPU and accidentally revolutionize the field in six months.
Replying to @Yuchenj_UW
Love this.
Replying to @Yuchenj_UW
the meta of AI
Replying to @Yuchenj_UW
That's a great story! The journey into AI really shows how passion drives learning. Just like with autonomous transactions and post-quantum security in Sealcoin, diving deep into a field can lead to unexpected breakthroughs. Exciting times ahead!
Replying to @Yuchenj_UW
Because "just do things" skips the 99% who grind unseen and fail Yuchen Meta folks are the curated survivor, not the rule.
2
22
Replying to @Yuchenj_UW
I came from neuroscience and am now a PhD in ECE. You just find something that is interesting and then after a little obsession and some dedication a whole new world is opened up!
1
13
Replying to @Yuchenj_UW
The old gates are gone. You don't need permission. You can just do things. The curriculum is online. The tools are for sale. The frontier is open. I launched 4 products doing the same. What are you building?
15
Replying to @Yuchenj_UW
Math is really the foundation on which all of tech progress rests
9
Replying to @Yuchenj_UW
It’s fascinating how some of the best minds in AI today come from non-CS backgrounds, right? Physics, math, even biology.. it shows that curiosity and a willingness to learn can lead to breakthroughs in unexpected ways. But I wonder.. does this cross-disciplinary perspective actually give researchers an edge in tackling AI problems, or is it just incidental?
1
1
6
Replying to @Yuchenj_UW
this is basically the origin story of every ML researcher i know, turns out you don't need a CS degree when you have a GPU and enough arxiv papers
8
Replying to @Yuchenj_UW
My roommate in college was about to get a PhD in differential geometry, realizing he couldn’t get a faculty job at a top university, so he got a MCS degree. Somehow he got a postdoctoral position in computer vision, then became a CS professor at UFL.
7
Load more