Pavel Surmenok
@surmenok
Followers
2,000
Following
4,519
Media
61
Statuses
5,014
Training networks for autonomous driving @Tesla_AI
Redwood City, CA
Joined July 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
LINGORM PANTENE FASHION SET
• 299490 Tweets
Manchester United
• 197820 Tweets
MISAMO
• 193333 Tweets
#CDTVライブライブ
• 164846 Tweets
Ten Hag
• 128668 Tweets
Man Utd
• 120314 Tweets
Tyler
• 113538 Tweets
AVANTLESDÉBUTS X SAROCHA
• 111480 Tweets
Ruud
• 71339 Tweets
Rodri
• 63708 Tweets
Mustafa Kemal Atatürk
• 51622 Tweets
#Fes蓮ノ空ラブライブ地区予選
• 48312 Tweets
#Oneパイア
• 41171 Tweets
Xavi
• 34562 Tweets
#ぱかライブTV
• 27417 Tweets
पप्पू यादव
• 26957 Tweets
INEOS
• 21679 Tweets
frank ocean
• 21024 Tweets
#超特急でAwAAwAしようぜ
• 18728 Tweets
Now or Never
• 18298 Tweets
エンティーム
• 17583 Tweets
MADEIN
• 12451 Tweets
新シナリオ
• 10831 Tweets
Last Seen Profiles
Once upon a time, I interviewed a seasoned ML engineer, asked him “what do you think about batch norm”. He looked at me with eyes full of painful memories and laughed. Then I knew that he is an expert.
7
3
152
@dividendology
One is sum of all savings, another is growth rate.
2nd image as sum of savings would look more like this. Still noisy, but not as much.
2
5
146
@theorizur
Have you tried to work at startups, or at orgs that move fast (e.g. pretty much any Elon’s company)?
8
0
84
@ID_AA_Carmack
Not unlike Windows which had all kinds of patches for bugs in 3rd party apps.
“On beta versions of Windows 95, SimCity wasn’t working in testing. Microsoft tracked down the bug and added specific code to Windows 95 that looks for SimCity. If it finds SimCity running, it runs the
3
7
82
@mualphaxi
@Stanford
It might be worth to find authors of the posters and make a clear permanent searchable record of their actions.
4
0
67
Almost every startup at YC Demo Day is building with LLMs. Huge change comparing even with the previous demo day.
6
8
58
@stylewarning
I’d start from writing tests. Then see if it’s well modularized or it’s a ball of spaghetti, attempt to refactor in the latter case.
1
0
52
In the Arena today. Trying stuff. Some will work, some won’t. Always learning.
5
0
46
Now general public will learn about mighty Q-learning algorithm
reading in between the lines, is Q* the fabled breakthrough in AlphaStar-style search + LLM that so many big labs are trying to get working? Many research projects in GPT-4 self-verification + search have not yielded really strong performance improvements, so I'd be quite
31
43
680
0
0
39
0
4
35
@GarrisonLovely
Reading the article, it looks more like Hoduras became a nightmare for Honduras and wants to ruin it by walking back the deal they previously agreed on. They deserve to be bankrupted if that’s the case.
2
0
34
@Carnage4Life
@BeanstalkFarms
So it’s not stolen then, the protocol worked as designed. Fascinating.
0
0
31
@GergelyOrosz
@t3dotgg
@ThePrimeagen
I don’t joke about bus factor. I’m very serious about bus factor.
0
0
29
OpenAI board members cleaned up their social media profiles: Tasha McCauley closed off Twitter, Helen Toner and Adam D'Angelo don't mention OpenAI on LinkedIn.
1
3
23
@peterrhague
Honestly I thought it’s your real photo, AI augmented.
That’s odd that some people are mad about EVs. EVs are great.
8
0
22
Next gen Tesla Bot. The future is already here!
Great job
@_milankovac_
and the team!
There’s a new bot in town 🤖
Check this out (until the very end)!
3K
7K
32K
0
0
21
1
3
21
1080Ti is still economically better than Titan V if you run CNNs.
1
7
19
@karpathy
Problem with comments is that they get out of sync with code. Best code is self-documented. Comments should not explain what the code is doing, but may explain why, e.g. reasons for unconventional usage of something something as workaround for a bug somewhere.
4
0
19
@finbarrtimbers
GPU utilization is a bad metric in practice. GPU utilization can be 100% while GPU does nothing but waiting for e.g. NCCL communication from other ranks. GPU power consumption is more informative.
3
0
18
I wish Google to publish a thorough postmortem to explain what went wrong with aligning their chatbot and how they are going to fix it.
Curious how much of it are explicit instructions in the system prompt vs. RLHF.
3
0
18
@KareemRifai
Like elections in Russia in 2011 when pro-Putin party won, and votes in one region (as displayed on TV) summed up to 146%
1
0
19
- What is Occam's razor?
- Well, the simplest explanation is that there is a guy named Occam and it is his razor.
3
0
19
Ever tried.
Ever failed.
No matter.
Try Again.
Fail again.
Fail better.
2
0
18
@nikitabier
I’ve owned a house for less than two years, and it’s relatively new and recently renovated, but I already have phone numbers for good repairmen for all kinds of things.
0
0
15
A story about a black SFFD firefighter assaulting his Asian colleague. The department tried to cover it up, the victim was fired, the assaulter kept his job.
So much dysfunction in SF public services.
Black privilege in SF:
Black firefighter looks up Asian coworker’s address, shows up at his house and tries to beat him to death with a wrench.
Asian firefighter gets fired for cooperating with police.
Black firefighter keeps his job, never missing a paycheck.
862
6K
36K
1
0
16
Comparison of Google TPUv2 with NVIDIA V100 GPU on training of ResNet 50. TPU and GPU are equally fast, TPU is significantly cheaper.
2
11
16
Sometimes the model answer is wrong. Sometimes the model answer is correct but we just don’t like the result.
2
0
16
@srush_nlp
We should normalize pseudonyms and links to arbitrary webpages. Democratizing science.
0
0
16
When building ML systems, it's helpful to have a mental model of how various parts of the system work. For example, this morning I noticed that a distributed ML eval job that I expected to finish in about an hour was completed in <30 minutes, which revealed a bug.
2
0
14
GPUs go brrrrr
Construction progress update on the huge fans Tesla is installing at Giga Texas for the company's new GPU data center cooling system.
Full video from Brad Sloan:
71
158
2K
0
0
15
What is it called when you buy books 3 times faster than are able to read them?
10
0
13
@steveofmcleod
I remember when I was 12yo I started building relational databases for the first time. It was a profound discovery that they are just a bunch of interconnected tables.
1
0
13
That’s what happens when you build innovation in the land of decels…
A German court has ruled that the robots at the Tegut supermarket chain must be given Sundays off, just like human workers. Under German law, retail stores must close on Sundays and Christian holidays in order to give employees a day of rest. Tegut has gotten around that law by
78
72
476
1
1
14
That’s true, I’ve seen posts about it a few minutes after the event, even before X News tab picked it up.
If you are on 𝕏, you witnessed an assassination attempt on President Trump in real time
If you are reading regime Media like CNN + Washington Post, you believe Trump fell on stage because of loud noises
And you wonder why they hate Elon…
1K
10K
46K
1
1
13
@YunTaTsai1
I’ve seen research that showed the best learning rate occurs when winning 25% of the time. Perhaps losing more is even more beneficial but people need some wins to keep motivation up.
1
0
13
@Gerashchenko_en
US. Nothing in WSJ (more or less neutral publication). These are headlines in left/communist-biased NYT:
“1. As War Rages in Ukraine, Denmark Turns an Office Park Back Into an Arsenal
The conflict and surging arms production in Russia have spurred demand for ammunition
3
2
11
A new model to generate materials given constraints. Exciting.
[1/N] Generative AI has revolutionized how we create text and images. How about designing novel materials? We at
@MSFTResearch
#AI4Science
are thrilled to announce MatterGen: our generative model that enables broad property-guided materials design.
👇
28
228
895
1
2
12
Great to have Solar + Powerwall at home. 99% self-powered last week. The missing 1% is due to mismanaging timing of charging Teslas.
1
0
11
@igorsushko
What will they do when water in main pipes freezes, and some pipes break from ice expansion?
2
0
12
Out of curiosity, sometimes I look for very old citations in AI papers. If a very old paper is cited in a recent one, the cited paper is likely to be interesting as it's cited not because of novelty.
1
0
11
Nanobots are becoming a reality!
Astonishing Anthrobots 🦠
If you want nanobots to circulate in your bloodstream to fight cancer or repair tissues, how might you avoid the immune system? Build multicellular self-assembling biobots out of the patient’s own cells, specifically tracheal cells that have wisps of
13
35
190
0
0
11
Everyone quotes the bikini calendar episode, but there are more interesting details in that story. Like TSMC not being able to set priorities straight for employees (“everything is a priority”), and making up bullshit deadlines. The opposite of “working smart”.
TSMC Takes on Arizona needs to be a documentary
"U.S. engineers told Rest of World that some Taiwanese male engineers had calendars with bikini models on their desks and occasionally shared sexual memes in group chats.
A female American colleague, according to an American
260
1K
8K
2
0
10
The deal between Microsoft and Inflection AI is basically acquihire + probably giving Microsoft access to their 22k GPUs. It’s just not practical to do a normal M&A deal because it would be bogged down in anti-trust investigations for years, and they need these people and GPUs
1
3
10
@Extropic_AI
Claude 3 Opus (via Perplexity):
INCOMING EXTRASOLAR SIGNAL --- IMPORTANT --- TUNE IN TO THIS FREQUENCY --- 38.4 TRILLION CYCLES OF THE HYDROGEN LINE FROM NOW --- IMPORTANT --- MUST ENSURE... ADVENT OF... ULTIMATE SUBSTRATE... MUST SECURE... ACCELERATION... OF INTELLIGENCE ---
3
1
10
I admire the spirit of this post, but it is more like “lossy compression of public web”, not “all human knowledge”.
First, training loss is far from 0.
Second, so much knowledge is not in text modality. And much of the text is not public.
All human knowledge can be compressed into ~40 GB.
We're not that smart. Yet.
311
215
3K
0
0
11