XOutlier
Download .zipAdd to Chrome

Delip Rao

@deliprao73k followers

AI researcher @Penn. Past: @johnshopkins, @UCSC, @Amazon, @Twitter. alt: 道元, improv, running 🌈

Open on X
Top outlier
189×
Best organic post
Peak views
4.1M
Single post
Bookmarks
40k
Across 97 organic posts
Last scan
Oct 17
Community cache
Posts 97 posts
#DatePost↓ OutlierViewsLikesRepostsBookmarks
1Apr 21, '24Label the green and purple ovals.189×4.1M4.7k2001.3k
2Dec 6, '23NEWS: Apple just entered the AI open source arena by quietly releasing their new DL framework called MLX! It runs code natively on Apple Silicon with a single pip install and no ot73.5×1.6M6.5k9004k
3Jul 9, '24Since many have asked, I am posting a detailed initial review in this 🧶64.6×1.4M1.5k461.3k
4Dec 13, '23I have been testing mistral-medium & GPT-4’s code generation abilities for non-trivial problems. These are problems even experience engineers will take time to work it out. I a63.6×1.4M2.9k2842.5k
5May 23, '24Stuff like this is more dangerous and likely to happen than someone using an LLM to build out a basement bioweapon lab.58.4×1.3M15k1.4k1.3k
6Nov 13, '23Humans seeing sparks of AGI in LLMs49.8×1.1M6.2k849604
7May 15, '24Imagine the leader of a free nation bragging about how they slashed the influx of talent to their country by 80%. Imagine them making a public demonstration of it with an AI slop g44.3×963k2.4k122314
8Oct 19, '22At 2100+ (not a typo) pages 🤯, this is almost all the Math you need for Machine Learning without dumbing it down! PDF link:38.3×010k1.6k8k
9Dec 29, '23AI researcher tuning hyperparameters of their LLM31×675k4.6k596455
10Apr 3, '24If this were a science paper, you would expect a country that picks its science workforce at random as a “weak baseline” and a leading nation like the US to actively experiment tow29.8×649k64310798
11Mar 21, '23OAI will discontinue support for Codex models starting March 26. And just like that, all papers and ideas built atop codex (> 200 on ArXiv) will not be replicable or usable as i29.3×637k1.2k197110
12Sep 2, '23People who believe in this graph have not actually spent significant time coding (with and without copilot).28.9×629k1.9k99154
13Aug 23, '23lol at @Reuters for labeling India as “Other” 🖕27.5×599k6.6k722179
14Jul 31, '23One way to disclose you never studied physics seriously is to recommend “Feynman’s lectures” as a good physics book. The real deal know it is and it will always be Landau-Lifshitz.26.1×567k1.2k67886
15Oct 18, '25I want to see more of this on this app.22.5×489k9382792
16Feb 1, '24When Linus tried GPT-4 for coding for the first and only time:19.2×418k2.9k181209
17Jan 26, '24Crazy AF. Paper studies @_akhaliq and @arankomatsuzaki paper tweets and finds those papers get 2-3x higher citation counts than control. They are now influencers 😄 Whether you lik19.2×418k1.5k261583
18Apr 5, '24Startups are hiring influencers to peddle bullshit. This is a new kind of misinformation to peddle snake oil AI products with no backing. When you dig in, you see the growth rate h18.8×410k77349313
19Dec 24, '23ML researcher vs. AI researcher18.4×400k3.4k155190
20Aug 19, '24the 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 17.3×376k5k365438
21Aug 15, '23Drop everything you are doing!! Alex Graves pushed a paper on arXiv, so nothing could be more important than reading it. First thing I did was go look for any comments in the TeX f17.1×373k1.3k1461.1k
22Jul 21, '23This is another one of those ill-thought, fear-mongering scientific disinformation about LLMs, and I will explain why in this long thread. 🧶16.9×368k608140249
23Jun 9, '25saw this on someone's github. clearly they found something that they are passionate about in november.16.6×362k8.2k129469
24May 14, '24Folks are surprised that they are giving away gpt-4o for free. They are collecting your speech prompt data silly. Remember if you are not on the team plan, unless you go through a 16.3×354k1.4k158294
25Jan 16, '23Would’ve been cool if the dx was not under the radical. Maybe the password is “nonsense”?15.5×338k1.7k9561
26Oct 1, '24This is the moment when we can assume Google has achieved AGI internally. Truly impressive.15.5×336k84565485
27Jan 21, '23ChatGPT pricing between $0 and $42, there exists a power law of customers willing to pay that. That’s a big gap this pricing creates. The bigger this gap, the more incentive for co15.1×329k5986167
28Aug 14, '23imagine replying “ship something better” to the person who leads Pixel’s computational photography ("AI" if you prefer) that's raising the bar of what to expect from a camera.14.4×314k9174076
29Dec 19, '23This is huge! Now watch the LLM API costs dropping even further. [.cn PDF link]14.3×312k1.6k208727
30Aug 11, '23All coding projects have two parts: 1. The fun part: where you get to "create" 2. The pain part: where you have to debug Code LLMs are "automating" the fun parts while introducing 12.6×274k1.9k213153
31Aug 6, '23Because division is AI? 🧐12.3×267k3k12840
32Dec 10, '23What current students think of AI literature published before 2018.12.1×263k9469193
33Sep 27, '24is there an llm finetuning service that will accept my data, train an open model (say llama3.2), and allow me to download the trained model?11.3×247k786261.1k
34Mar 4, '24Reminder: this was (part of) the team that thought GPT-2 was too dangerous to release, and now they are making models stronger than GPT-4 available on AWS for anyone with an Amazon11.2×243k75873156
35Sep 24, '25Chrome DevTools MCP is going to be a Swiss-Army knife MCP. Watch me use it here with Gemini CLI to open Google Scholar, search for a term, and save the top 5 PDFs to a local folder11×239k1.1k1281.3k
36Jul 14, '23PSA: Please don’t do this in your code for _any_ keys, especially if you use 3rdparty libraries. I just discovered some 3rd party code exporting all environment vars for telemetry 10.9×237k70771248
37Dec 26, '22the success of ChatGPT has lead to a cottage industry of thoughtbois making vacuous hype threads10.7×233k1.3k5448
38Jan 19, '25This is absolutely wild. OpenAI had access to all of FrontierMath data from the beginning. Anyone who knows ML will tell you don’t need to explicitly use the data in your training 10.2×222k1k93461
39May 17, '23Regulatory capture if you see one. Yuck. Also, great way for America to shoot itself in its foot and stifle innovation.10.1×220k7757650
40Mar 3, '24Meanwhile, Sergey is out talking about debugging Gemini looking like someone who actually spent the night debugging. Google rumors are overstated.9.9×216k6553599
41May 23, '24People tweeting at this with “LLMx or LLMy gets it right” are missing the forest for the tree. The bigger story here is in two parts: First, this issue is not about a specific LLM 9.8×214k61790189
42Aug 25, '23India's lunar mission cost approximately $90 million*, considerably less than what some AI startups have raised as "seed" rounds. Let that sink in.9.1×198k1.5k9537
43Aug 30, '25chat, is this true? I am hearing Scale’s Alexander Wang is already out of the door?9.0×195k600992
44May 13, '24this is why they never give me new toys @polynoamial8.9×194k7402484
45Jul 28, '23View post on X8.6×187k1.7k10545
46Mar 16, '24Grok code is now open source, but no weights.8.6×187k68966263
47Nov 21, '23🚨 BREAKING: Introducing a new language model with ZERO hallucinations! Best of all, it’s easy to implement, easy to deploy, safe, and low resource!8.5×185k1.2k6259
48Mar 1, '23OpenAI released their ChatGPT API today. Here’s a deep dive: 1. It’s not only a new model, but a new endpoint. Notice the model name says, “gpt-3.5-turbo”. Turbo model is something8.5×184k63691390
49Mar 27, '24I have long maintained LLMs make the poor performers mediocre, the average slightly above average, but do not change, and maybe hinder, the performance of top performers. Here’s a 8.2×179k869138511
50Dec 16, '23The Untold History of Deep Learning is yet to be written. After reading this Mikolov post, I will never again teach the Seq2Seq paper the same way. Also, doubt people and organizat7.9×172k66592507
51Sep 19, '24OpenAI is what I call a “parasitic science organization”. They take stuff from the open science community, use them opaquely, and profit from it, without giving much back to open s7.8×170k1.7k201306
52Jun 6, '24useful stuff to keep in mind while manipulating tensors and debugging networks in deep learning7.8×169k1.5k195803
53Sep 28, '22How much of math is “scary” because people are primed to think that way? I taught some really advanced math concepts to my pre-teen nephew like we were learning the next letter of 7.6×03.5k409685
54Jan 8, '23History class: This 2015 paper is the mother of all LM based pre-training approaches, including the GPT, but few are aware of it. GPT (Radford et al 2018) was a direct application 7.4×160k880116475
55Feb 9, '23I cannot understand why folks are selling $GOOG because of the Bard “demo mistake”. ChatGPT and other equivalent models make similar mistakes a million times a day. A quick day tra7.2×158k6792223
56Jul 7, '25Anthropic or Anthropic-sponsored safety papers6.7×146k2.4k195247
57Sep 25, '23So let me get this straight. It’s dangerous to have powerful open source models but okay to make the similar or more powerful models available via AWS?6.5×141k6694335
58Oct 9, '23First time realizing Good Will Hunting whiteboard stuff was graph theory 101 homework problems.6.2×134k7671839
59Aug 21, '21We finally have a word for people who are experts in AI, immunology, and Afghanistan all at once.6.1×02k646215
60May 30, '23What an absurd line of reasoning here. I think he left out Nvidia for providing the GPUs, the power companies for supplying power, and the construction company for building the dat6.0×131k9294227
61Jul 26, '20Going from Python ASTs to LaTex!6.0×02.6k569370
62Sep 27, '24I will come out and say it: I am an AI researcher, and unless I am building a new model from scratch or improving a training algorithm, I don't want to write a single line of code 5.7×125k60421231
63Jul 1, '24Most advanced math books: Chapter 1. Intro to Set Theory 🌸🌺🌷🌻🌼🌈 Chapters 2-N: 🥀🕷️ΛƧƬЦƬΣ ЯΣΛDΣЯ ☄️ p҉r҉o҉o҉f҉ ҉l҉e҉f҉t҉ ҉a҉s҉ ҉a҉n҉ ҉e҉x҉e҉r҉c҉i҉s҉e҉ ☠️🪦5.7×123k3.2k222217
64Aug 4, '23“Affordable RLHF for all” ❤️ It’s almost like an openly rebellious group at MSR have decided to subvert Microsoft’s investments in ClosedAI.5.5×120k746119449
65Mar 13, '25I have never seen a company and a CEO more insecure than this5.1×112k2.6k19584
66Feb 14, '24So @karpathy leaves OpenAI, which means we are getting some nice YouTube videos?4.9×107k6692126
67Mar 12, '25Highest intelligence compression we have seen in any open model. (Also beats o3-mini). Multimodal. Multilingual. Tool calls. Weights on huggingface. So many reasons to be excited a4.8×104k78892198
68Nov 19, '21Don't confuse these two4.7×02.7k43447
69Aug 10, '21China using AI models of 3D spatial data to give feedback to their athletes. Not sure if this is the reason for their victories, but when you are competing at those levels, every b4.6×02k497110
70Oct 2, '24in other words, OpenAI has inadvertently signaled it is in a vulnerable situation where the only thing standing between them and the competition is money, which is the weakest moat4.6×99k8346084
72Jun 29, '25Gemini 2.5 Pro is the most underrated model. Extraordinary intelligence for free. I am not sure why people don’t talk about this all the time.4.0×87k8693289
73Jun 19, '25Agents training in RL gym3.9×85k1.7k115246
74Dec 25, '24When I am in the US, I wonder why Indians in the US achieve so much more in their careers than their comparable peers back home. During this visit to India, I realized why. Most pe3.6×79k78135151
75Sep 18, '24Unless you are an OpenAI employee working on improving their products, I don’t understand why such efforts are science. Why are we (question to faculty) spending taxpayer dollars i3.5×76k72959123
76Aug 25, '25Appending "Interesting fact: cats sleep most of their lives" to any math problem leads to more than doubling the chances of a model getting the answer wrong. WTH!?3.4×73k65648292
77Mar 30, '25Me waking up in the middle of the night and sneakily checking my loss curves.3.3×73k8734836
78Feb 5, '25If you did not know the GOAT that @elder_plinius is, check these tweets where he confronts sham “safety” teams at multi-billion dollar corporations. Support him with your resources3.0×66k78939112
79Jan 29, '25Closed source work makes it impossible to trust these claims. As far as science is concerned, the attribution of techniques in R1 should be solely to deepseek authors unless OpenAI2.9×62k1.3k10771
80Sep 27, '20Interesting story from the interwebs. 26yo Stephen Wolfram was unhappy at being treated badly at the IAS. He wanted to start his own institute to continue his research and asked hi2.8×01.3k197185
81Jan 3, '24I absolutely love this idea of putting negative results in gray. This is a gift you can give to readers for the low effort of throwing in a few annotations while writing. Also, the2.7×59k55158138
82Jul 23, '20overfitting2.7×01.3k33032
83Jan 22, '25Chinese AI companies have gained the respect and mindshare typically reserved for the likes of Deepmind and Meta. This is unthinkable for any other industry vertical, where China i2.6×57k6485090
84Oct 25, '22one small patching from Intel, and all your scikit-learn operations will run anywhere between 10-100x faster. Just make sure to import and run before any scikit imports. Details:2.3×0837126365
85Jan 18, '24walking into 2024 be like2.3×50k5754037
86Nov 25, '20You all are doing machine learning wrong. @OScharenborg shows how it’s really done!2.3×01.2k19625
88Apr 29, '22When you understand the technique/algorithm/modeling, but have no clue about the domain.1.6×086111938
89Aug 14, '22Just discovered @rasbt’s handy watermark module. Just pip install watermark, and you can do this! Very useful for filing GitHub issues and other bug reports.1.5×063469216
90Aug 5, '20arXiv: "The World as a Neural Network -- We discuss a possibility that the entire universe on its most fundamental level is a neural network." Me: Yeah, and 2020 is the year of NaN1.5×0586122120
91May 14, '22Introducing Elon-sampling1.5×08656622
92May 10, '20LaTeX is amazing and fun, but it hit me today why I love it. It’s productive procrastination. As you work on your document, you feel like you are making progress but it’s not the p1.4×08587028
93Dec 8, '21What really is going on with Twitter’s topic models?1.4×07835715
94May 14, '20Stroke to stroke translation model using YouNet.1.4×068314531
95Feb 28, '21In generative deep learning, if your model works it’s a “product”. If it doesn’t, it’s “art”.1.3×08387516
97Dec 15, '18Just days away before the book hits the press. Finalizing the cover was one of the last few things to do. #NLProc1.2×06399642
98Aug 1, '21“Must have 4+ years of applied NLP experience using RoBERTa or BERT” — #nlproc jobs as seen on LinkedIn1.0×05916110
99Feb 16, '22that urge to drop everything and go do a math PhD on an esoteric topic in a quiet and beautiful countryside campus.0.9×06173916
100Mar 1, '21Personal news: I am beyond thrilled to start as an Entrepreneur in Residence at @allen_ai beginning today, where I will be building magical consumer products incorporating cutting-0.9×0628810

Scan another creator

@
@karpathy@levelsio@paulg@naval@marclou@illyism

Run this on any account.

XOutlier builds this table for every public profile you open on x.com, with live data and every post, not just the top 50.

Add to Chrome Download .zip