Eric Chen

2.2K posts

Eric Chen banner
Eric Chen

Eric Chen

@below_ocean

ce

bay Katılım Mayıs 2016
235 Takip Edilen161 Takipçiler
Eric Chen
Eric Chen@below_ocean·
learning boomer cpp cutlass has been quite the awful experience, too much for my tiny brain
English
0
0
0
11
Eric Chen
Eric Chen@below_ocean·
i know for older nvidia chips like a100 LDMatrix instructions optimize 1-stride K layouts for gemm (M, N-major for A, B) but benchmarking on pytorch by creating (m, k) (n, k) contiguous A, B has the worst results on 256x256 and 8192x8192. Anyone know why?
English
0
0
0
18
Eric Chen
Eric Chen@below_ocean·
never never never rely on AI for any CUDA or GPU stuff at all it will find any way to explain any discrepancy it faces and is often wrong i can't tell by using it to learn CUDA whether it's been an overall benefit or harm because of how much it's wasted my time
English
0
0
0
11
Eric Chen
Eric Chen@below_ocean·
@bitcloud You're so naive aren't you. How is a neural network searching the knowledge space via gradient descent not the same as this organic neuron activating to external stimuli? My ClaudeGPTProMax would like a word
English
0
0
0
145
Eric Chen
Eric Chen@below_ocean·
@xwanyex the hardest question for a rationalist is why because just keep asking why and their brain blows up or ask if it's ok to sacrifice one human life for every ant stepped on and I guarantee you that's about as explainable as why "causing suffering is bad" if you're "rational"
English
0
0
0
91
wanye
wanye@xwanyex·
If you're a rationalist, then not only can you do this, you already do. For example, the rationalist must believe that we are but insignificant dust, that all of morality is constructed and contingent, that there is no final accounting, that someday you will die and it just literally will not matter what kind of life you lived. Yet almost all psychologically normal rationalists feel like murdering a child in cold blood is wrong. They aren't making tradeoffs. It's not a rational decision. It's not philosophical. They of course will *justify* this feeling with various kinds of moral arguments. But that's not really why they think murder is wrong; they *feel* as strongly as they feel anything that it's capital-W WRONG to kill children, that this is a fact of the universe, that it can't be any other way. They organize their entire lives around beliefs like this, which not only drive their actions, but determine how they conceive of themselves and others. The average rationalist really truly believes that torturing children for no reason is wrong in some way that's baked into the universe. And those who don't, those who are intellectually honest enough to have followed me all the way to this point, who are willing to bite the bullet and say that torturing children isn't actually capital-W Wrong in the way I clearly mean here all *live as though they believe it's true*.
Sin Tax@sin__tax

@xwanyex No, I can't just choose to believe something I don't actually believe. I genuinely would if I could - I believe it would be good for me! Maybe some people can. I find the notion baffling.

English
83
11
317
45.5K
Eric Chen
Eric Chen@below_ocean·
@skdh you sign up to youtube to share math with the world and Big Youtube is just going to yoink your video and make you fight a 3 year battle to earn like $300 because some random blud claims you copied his AI generated paper that shares 2 words with your script
English
0
0
0
57
Sabine Hossenfelder
Sabine Hossenfelder@skdh·
I received a false copyright claim on one of my videos and YouTube removed the video because of this. It's a video about the Riemann Hypothesis. The claim comes from some person who submits a link to their paper about "The Continuity Engine: A Formally Verified Framework Prime Resonance Unification with Medical, Physical, Mathematical Evidence" with links to two unpublished papers that are completely unrelated to my video content. It's obviously some crackpot work, I receive dozens of those a day. YouTube took the video down based on this false claim. The only way they allow me to react to this requires me to submit my personal contact information to some random crank on the internet. Alternatively, I am supposed to hire a lawyer (!!) on my own costs, to track down some random guy from whom I then have to extract my up-front expenses. I have complained to YouTube support about this multiple times. No success, the video is still down. This procedure is completely unacceptable. It allows random people to try and blackmail me into responding to them. I have no time for this bullshit and no patience either. Frankly the only sensible course of action forward that I see is to sue YouTube for facilitating fraudulent DCMA claims. @YouTubeCreators @YouTube
English
432
802
8.7K
459.5K
Phil Trubey
Phil Trubey@PTrubey·
Sorry, but it just had to be done.
English
121
169
3.4K
319.1K
Eric Chen
Eric Chen@below_ocean·
@xwanyex prompted you? oh so this whole time you've just been an AI
English
0
0
1
105
wanye
wanye@xwanyex·
This line from Mathew Yglesias‘s Substack today prompted me to want to say again that it’s important to remember that everything proponents of AI believe about AI is based on thought experiments, not on having developed something first and then noticed something about its character. It may be that they developed these ideas based on thought experiments and then as they created new tools it turned out that those new tools perfectly validated all their predictions. But you should really consider the possibility that they’ve just always believed these things and were always going to believe them no matter what the new tools could or couldn’t do. The fact that they believed these things before any tools even existed tells you that it cannot logically be the case that their views were developed in response to the tools. This alone can’t really answer any questions for you, but it’s an important thing to keep in mind, if only for basic human psychology reasons and nothing else.
wanye tweet media
English
27
18
306
31.7K
Eric Chen
Eric Chen@below_ocean·
@kitten_beloved most googs who joined meta as of late have said there is no AI push in Google like there is at meta
English
0
0
1
92
Kitten 🐈
Kitten 🐈@kitten_beloved·
Everybody is chimping out about the deepmind guy dunking on yegge but I would bet money he is substantially correct, in fact I would be shocked to learn that coding agent workflows were in wide deployment at Google Their entire engineering culture is antithetical to it
Kitten 🐈 tweet media
Steve Yegge@Steve_Yegge

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.

English
47
13
385
57.9K
PNW Conservative
PNW Conservative@PNWConservative·
Imagine how much your life must suck to spend your time making something like this….
English
2.1K
1.4K
29.6K
15.6M
Eric Chen
Eric Chen@below_ocean·
@arcticinstincts challenge: write without writing about being asian level: impossible
English
0
0
4
1.3K
David Sun
David Sun@arcticinstincts·
Checking in on the latest Asian American diaspora DEI grievance literature
David Sun tweet media
English
120
79
2.1K
222.3K
David Senra
David Senra@davidsenra·
How a final round interview wth Tony Xu (@t_xu) at DoorDash goes: "I'm going to give you 20 minutes and you can ask me any question that you want. But after the 20 minutes expires, I'm going to give you $20 that you can use to go and acquire 100 customers for us, and you have eight hours to do so. But here's also a plane ticket, in case you want to quit the interview now and just move on and find somewhere else to work. So much of what we were trying to test for early on is someone who's going to do something to go and collect information, as opposed to someone who's going to collect data, scrape information from the internet and do some magical analysis on it. What if none of that information existed? You have to go and do things in order to actually collect information."
David Senra@davidsenra

My conversation with Tony Xu (@t_xu), co-founder & CEO of @DoorDash. 0:00 DoorDash MVP in 43 Minutes 1:39 How Delivery Worked in 2013 3:17 Small Business Roots and Insight 5:48 Why Restaurants First 8:24 Palo Alto vs San Francisco 11:03 Early Customers and Unit Economics 15:22 YC Summer Three Questions 19:50 The Hidden Complexity of Delivery 22:02 Competing on Invisible Details 23:54 Chaos Data and Experiment Loops 30:58 Trust Reset Every Day 31:30 Stanford Game Meltdown and Refunds 34:41 Scaling Through Experiments 37:37 Customer North Star Metrics 40:10 CEO Customer Support Habit 42:55 Anecdotes Versus Data 46:52 Eternal Mission Local Economies 50:09 Turning Data Into Merchant Growth 59:12 New Products Beyond Delivery 1:01:14 Autonomous Delivery Strategy 1:05:06 Hiring Rhodes Scholar Navy SEALs 1:12:46 Driver Switch Experiment 1:13:42 Who Delivers and Why 1:15:33 Hiring for Action 1:18:07 Earned Secrets via Experiments 1:20:01 Money vs Problem Solving 1:21:18 Thousand Days of Hell 1:26:04 Staying Sane as CEO 1:30:07 Ignore the Stock Price 1:31:44 Two Operating Systems 1:35:17 Internal Venture Stage Gates 1:38:17 Learning from Founder Peers 1:42:29 Jiu Jitsu Lessons 1:44:37 AI Changes the Loop 1:47:01 Data Needs Action 1:48:24 Closing Thoughts Includes paid partnerships.

English
46
29
873
395.2K
Eric Chen
Eric Chen@below_ocean·
@negligible_cap only a rejection and not a full scale IRS investigation? I think they have a few more years
English
0
0
1
627
Negligible Capital
Negligible Capital@negligible_cap·
Yesterday I gave Claude Cowork everything it needed to do my taxes and sent it loose on TurboTax Claude absolutely flew through the process. Incredibly confident, very few follow up questions It’s so over for accountants
Negligible Capital tweet media
English
129
214
14.1K
1.9M
Eric Chen
Eric Chen@below_ocean·
my agi benchmark for robotics will be when they can take a face mask out of its package and put it perfectly on someones face
English
0
0
1
33
Eric Chen
Eric Chen@below_ocean·
@StefanFSchubert @jburnmurdoch are they central in a vacuum or in general responses? I notice the response to any topic that triggers a search tends to be polluted by whichever articles it reads
English
0
0
0
57
Stefan Schubert
Stefan Schubert@StefanFSchubert·
While social media is polarising, evidence suggests AI may nudge people towards the centre. This holds true of all studied models. Grok is more right-leaning than other models, but also has depolarising effects. By @jburnmurdoch.
Stefan Schubert tweet media
English
234
1K
6.2K
1.2M
Eric Chen
Eric Chen@below_ocean·
@Yuchenj_UW like...individually? You'd have to be prompting Opus 1M with each finger and toe all day long
English
0
0
1
43
Yuchen Jin
Yuchen Jin@Yuchenj_UW·
Friends at both big tech and startups tell me they’re spending more than $1000 per day on Claude Code or Codex tokens. That’s $365,000/year. We’re not far from companies spending more on LLM tokens than on human employees.
English
321
109
2.3K
217.2K
Eric Chen
Eric Chen@below_ocean·
@NathanpmYoung it's like people convince themselves they can only be knowledgeable about something if their pedigree says as much. The next time her milk spoils in the fridge she's gonna have to consult a milkologist to determine whether to toss it
English
0
0
0
6
Eric Chen
Eric Chen@below_ocean·
@NathanpmYoung how are you gonna put an appeal to your authority and then use the same thing to say why you're not qualified to answer
English
1
0
0
37
Eric Chen
Eric Chen@below_ocean·
@karpathy it's rough to do any ML research without being super careful with wording because it can literally argue any interpretation of your results which is pretty bad mostly because they find it hard to forget things they've concluded
English
0
0
0
9
Andrej Karpathy
Andrej Karpathy@karpathy·
- Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - Wow, feeling great, it’s so convincing! - Fun idea let’s ask it to argue the opposite. - LLM demolishes the entire argument and convinces me that the opposite is in fact true. - lol The LLMs may elicit an opinion when asked but are extremely competent in arguing almost any direction. This is actually super useful as a tool for forming your own opinions, just make sure to ask different directions and be careful with the sycophancy.
English
1.8K
2.4K
31.3K
3.5M