Dan @DanielCardena
A.I. Learner / A.I. Consultant. Experience working on Virtual Assistants. Ex: Google plus.google.com/u/1/communitie… Bay area, CA Joined June 2009-
Tweets2K
-
Followers162
-
Following129
-
Likes26K
This is how it works (stack): Hardware: SeeedStudio XIAO NRF52840 Sense: seeedstudio.com/Seeed-XIAO-BLE… Battery: a.co/d/2UbB7Sz Software: - 2x Flutter Apps (built in flutterflow so you can modify without code) - OpenAI for chatting/summary - whisper for processing audio…
Found this insane Google doc for "Keeping up with AGI" And claiming to be updated automatically every 5 minutes. Really brilliant 🔥 docs.google.com/document/d/e/2…
@winglian @LMStudioAI Yeah, it's super relevant now that we have an excellent Llama 3 Instruct. It feels like we might (slowly) move away from these general-purpose datasets as the performance boost they provide becomes less and less obvious.
💾 LLM Datasets LLM development is increasingly moving towards curating high-quality datasets, as shown by Llama 3. I've compiled a collection of fine-tuning datasets along with advice and tools for creating your own. 💻 GitHub: github.com/mlabonne/llm-d…
I've written up my study group lectures on implementing Transformers in PyTorch into a blog series: Creating Transformers from Scratch: - Part 1: The Attention Mechanism benjaminwarner.dev/2023/07/01/att… - Part 2: The Rest of the Transformer benjaminwarner.dev/2023/07/28/res…
⚔️ Closed-source vs. Open-weight LLMs The gap between the best-performing closed-source and open-source models is also narrowing in terms of Arena ELO (chat.lmsys.org/?leaderboard). We now face a 6 to 10-month lag, rather than years as was the case when GPT-4 was released. Let's…
Llama 3 8B & 70B models are just the beginning of what we’re working to release for Llama 3. Our largest models currently in the works are 400B+ parameters and while they’re still in active development, we’re excited about how this work is trending.
💥After months of work, we're thrilled to introduce ⚗️distilabel 1.0.0! 🚀More flexible, robust, and powerful. 🙌 Let's empower the community to build the most impactful datasets for Open Source AI! Blogpost: argilla.io/blog/introduci… Github: github.com/argilla-io/dis…
Alongside our Mixtral 8x22B release, we are releasing our tokenizers, which go beyond the usual text <-> tokens, adding parsing of tools and structured conversation. Repo: github.com/mistralai/mist… Guide: docs.mistral.ai/guides/tokeniz…
Mixtral-8x22B outperforms all open models across reasoning, knowledge, code and math capabilities while offering the best performance/cost ratio. Mixtral-8x22B-v0.1 on HF: huggingface.co/mistralai/Mixt… Mixtral-8x22B-Instruct-v0.1 on HF: huggingface.co/mistralai/Mixt…
👉 Meet Bobby Tables, but for LLMs! 🔥 An exploit case study with Google NotebookML. 🧐 embracethered.com/blog/posts/202… #llm #infosec #ml
Most people don't know that Lightning Studios offer: - free persistent storage - free persistent environments - unlimited background execution - VSCode, PyCharm, (any IDE) integration Set up your Studio environment once and reuse it again any time 🤯🤯
The main theme I got out of this video from the Sora authors: scale is all you need for AGI
The main theme I got out of this video from the Sora authors: scale is all you need for AGI https://t.co/WA9wOpQCed
Introducing MatX: we design hardware tailored for LLMs, to deliver an order of magnitude more computing power so AI labs can make their models an order of magnitude smarter. Our hardware would make it possible to train GPT-4 and run ChatGPT, but on the budget of a small startup.…
a16z just released an enterprise AI report after speaking to dozens of Fortune 500 companies. Thanks @sarahdingwang for this. Here are 8 insightful slides: 1. For production use cases, OpenAI still has dominant market share.
If you are doing LLM (>1B) training runs, you ought to do these 3 things: 1. Use SwiGLU 2. Use ALiBi 3. Use µP Why? Your training will be almost 3X faster! You can do 3 runs for the price of 1. You can go for much bigger model or train longer. There is no excuse. 1/3
Change is accelerating. The world is changing faster than it has ever changed in its past. Few people realize this. The pace of tech innovation is so much faster than in the past.
Change is accelerating. The world is changing faster than it has ever changed in its past. Few people realize this. The pace of tech innovation is so much faster than in the past.
Fully automated vulnerability research is changing the cybersecurity landscape Claude 3 Opus is capable of reading source code and identifying complex security vulnerabilities used by APTs. But scaling is still a challenge. Demo: claude.ai/share/ddc7ff37… This is beginner-level…
Here is probably the most useful Claude 3 prompt I've written. Use it to you help make engineering decisions in unfamiliar territory: --- <role>You are an engineering wizard, experienced at solving complex problems across various disciplines. Your knowledge is both wide and…
Introducing the latest version of our Hermes series of models, Hermes 2 Pro 7B. This latest version improves several capabilities, using an updated and cleaned version of the Hermes 2 dataset, and is now trained on a diverse and rich set of function calling and JSON mode…
JoYeates @83Y30oxcq9t6j6
1 Followers 339 FollowingNellyCecillia @2ETs0T0Mr2d6j0
3 Followers 267 FollowingDeonnaMcspedon @DeonnaMcsp49804
24 Followers 972 FollowingEthel💎💎💎 @Ethelsingslng
1K Followers 263 Following There's nothing terrible about it, I'm capable enough to do it and do it wellDnerseyd @dnerseyd47464
0 Followers 343 FollowingJulie @Julie0801541883
91 Followers 3K FollowingAliyahmommi @aliyahmomm72134
18 Followers 1K FollowingTeashaez @TeashaezjhAL1
3 Followers 150 FollowingGary Black🚀 @gary_black_OO
76 Followers 4K Following Managing Partner, The Future Fund LLC, SEC registered investment adviser. CEO/CIO/PM. https://t.co/FaOYjbSi4j garydblack. Disclosure: https://t.co/qsirMmZ4asGary Black @garrybllack00
53 Followers 3K Following Managing Partner, The Future Fund LLC, SEC registered investment adviser. CEO/CIO/PM. https://t.co/QBTlFsY9v0. Disclosure: https://t.co/NFcsppBf9oShosh @ShoshMmxM94
3 Followers 227 FollowingPetPortal @PetPortal_ai
164 Followers 920 Following Chat with your Pet using AI! https://t.co/wjCAREUzqL Free Beta open now!just Rob @rsealsvyzion
1K Followers 5K Following it’s about the quiet work | AI and mma nerd | diligently thoughtful | I’m likable I have a 4.9 Uber passenger ratingEva McMillan❤ @EvasTeslaScopes
853 Followers 7K Following Residential Real Estate in FL, Waterfront & Beachside Specialist. First Tesla S Plaid owner in FL, Tesla Shareholder. #FSDBeta tester. Fluent in 6 languages.fafa.👩🏻💻 @DS_Fafa_
29K Followers 18K Following 👩🏻💻 Senior Customer Insights Director. 👋 I tweet mostly about #DataScience with ☕️Frank Pinnola 🇺�.. @frank_pinnola
690 Followers 856 Following fullstack + ml in healthtech || prev eng @getchargeapp CPE/CS + hackathons @followstevens || 🐇🏔Reshe @Reshe59979
163 Followers 5K FollowingPierre Ferragu @p_ferragu__
50 Followers 3K Following Commenting on technology - investment recommendations.Elon Reeve Musk @EMusk66794
105 Followers 1K Following Founder, CEO, and chief engineer of SpaceX, CEO and product architect of Tesla, Inc. Owner and CTO of X, formerly Twitter President of the Musk FoundationBen Artuso @ben_artuso
55 Followers 129 Following fullstack engineer + devrel @ https://t.co/45xJnyl1jS 🐂 🦀Dataplus @DataplusShu
10 Followers 58 Following Crowdsourcing Platform with High-quality and Efficient Data Solutions for AI training at a wide rangePierre Ferragu @p_feerragu
17 Followers 394 Following Commenting on technology - no investment recommendations.Alexander Naumenko @AlexanderNaume2
714 Followers 3K Following Theory of intelligence, #AGI, #NLP paper - https://t.co/zS8V7Ymwiq book - https://t.co/PJiAS8iMWxLady Luck @LadyLuckend
845 Followers 1K Following This journey in the human world, over mountains and ridges, is just for myself...penguin🐧 @tashinsmith22
9 Followers 294 Following I love life, I like outdoor activities, sunbathing🌞, diving🤿, swimming🏊♀️, hiking🎧, golf⛳️, fishing🎣, and all kinds of food🍖🥩🥠🍜ParallaxEffect @ParallaxAngle
77 Followers 356 FollowingMax Zanoga @zanoga
214 Followers 948 Following 🚀 AI Enthusiast | 🔍 Exploring the frontiers of LLMs & ML | 📚 Lifelong learner & tech optimistTereza Tizkova @tereza_tizkova
3K Followers 984 Following Founding DevRel @e2b_dev | Mathematics Grad. | I like cheeseInvesting Paths @investingpaths
156 Followers 650 Following TA | Learner | Chart Enthusiast | F&O Trader | Investor | Managing Risk & Capital Allocation For a Living | No Paid Tips | Sebi UnregisteredShounda @Shounda25978
46 Followers 2K Followingshare9 @innovateinput
0 Followers 4K FollowingSteve @stevebutnotjobs
102 Followers 2K FollowingSteve Hou (“CONSUME.. @stevehouf1
794 Followers 6K Following Research @Bloomberg Indices, formerly @aqrcapital , @umichEcon PhD, @ETH and @UVA , opinions my own.Thuse @Thuse396745
11 Followers 1K FollowingLevi Githaiga @CodeTitanium
6 Followers 402 Following🆁🅾🅼🅰🅽 .. @ScienceGeekAI
695 Followers 4K Following #Warsaw - AI combined with ML | 🙌🏻 | 🥃 | #AI | scientist 🧠🤯🗣| ᎥᏁ🅵Ꮎ🅱ᏒᎾᏦᎬᏒAlexandra Howard @HowardAlex75904
104 Followers 3K FollowingMaggie @JoanaDharma
10 Followers 492 FollowingReeves @AskReeves
105 Followers 434 FollowingFlowers from the futu.. @futuristflower
13K Followers 114 Following In the unity of all, under night's solemn sky, she stands, sacred and forlorn, with a cunty gleam in her eye.Matt Shumer @mattshumer_
51K Followers 1K Following CEO @HyperWriteAI, @OthersideAI - I make AIs do the impossible.Paul Christiano @paulfchristiano
2K Followers 0 FollowingLLM360 @llm360
1K Followers 50 Following A framework for open-source LLMs to foster transparency, trust, and collaborative research.Teslascope @teslascope
59K Followers 134 Following An elegant view of everything about your Tesla vehicle. We are the worldwide drivers' platform, the highest-rated vehicle companion, and always here to help. 🚘oxen @oxen_ai
307 Followers 86 Following Build World-Class AI Datasets. Together. You a nerd? Check out our weekly #AI #ML Paper Club Arxiv Dives too! Hugs and MoosUnsloth AI @UnslothAI
3K Followers 257 Following Making AI & LLMs more accessible + faster for everyone! 🦥 Github: https://t.co/2kXqhhvLsb Discord: https://t.co/1Gmc1SDEljVladimír Vondruš @czmosra
1K Followers 712 Following Graphics dev, creator of m.css and #MagnumEngine. Sometimes using physical and virtual cameras to produce questionable art. DMs open, ask me things!Tesla AI @Tesla_AI
150K Followers 14 Following Developing & deploying autonomy at scale in vehicles, robots & moreneurallambda @neurallambda
222 Followers 164 Following λ/acc. building opensource AGI. the gap is reasoning. reasoning is just the ability to manipulate symbols regardless of their meaning.Skunkworks AI @skunkworks_ai
3K Followers 7 Following Accelerating Open-Source AI https://t.co/B5v2ohlIbH https://t.co/9TNVZeJYjd no website for our groupRWKV @RWKV_AI
2K Followers 3 Following AI model built by the community, for everyone in this world Part of the Linux Foundation, Apache 2 licensed An RNN scaled to 14B params with GPT-level of perfJames Betker @neonbjb
2K Followers 6 FollowingObsidian @obsdmd
131K Followers 3 Following The private and flexible writing app that adapts to the way you think. For help and deeper discussions, join our community: https://t.co/QsDArfFSa3Chris Gorgolewski @chrisgorgo
7K Followers 1K Following Product manager, software engineer, researcher. Previously at: @googleanalytics, @GoogleAI, @kaggle, @StanfordPsych, @MPI_CBS. Opinions are my own.Fastmail @Fastmail
14K Followers 378 Following Privacy-first email, calendars, & contacts that make your life easy. Masked Email w/ @1Password. Email experts for 20+ years. Also [email protected]Adam Sadovsky @asadovsky
561 Followers 296 Following Distinguished Software Engineer / Senior Director, GeminiMaxime Labonne @maximelabonne
13K Followers 439 Following Staff ML Scientist @LiquidAI_ • Author of Hands-On Graph Neural Networks https://t.co/Q8victWUmRsync.labs (YC W24) @synclabs_so
3K Followers 0 Following an api to create, modify, and understand humans in video | AI to turn videos into living storiesDevendra Chaplot @dchaplot
8K Followers 365 Following Building next-gen AI at @MistralAI. Past: Research Scientist at Facebook AI Research. Ph.D. @SCSatCMU, BTech @iitbombay CS.Lesaun @LesaunH
2K Followers 2K Following exists (whoa!) — assigns a nontrivial probability of achieving LEV — wants to leverage superintelligent-level AI.CHAI: AI Platform @chai_research
138K Followers 84 Following CHAI: AI Platform for chat. Our Mission is to crowdsource the leap to AGI. Small Team, Big ImpactBing Xu @bingxu_
667 Followers 33 Following CEO & Co-Founder @HippoML_com, Initiator of AITemplate, XGBoost-Python/Julia, CXXNet, MXNet, MXNet.js.Crossbar @crossbarinc
502 Followers 309 Following Crossbar is the leader in ReRAM technology, enabling kilobytes to terabytes of always-on data storage.Dipam Chakraborty @__dipam__
194 Followers 660 Following Machine Learning Engineer at @h2oai 🤖 Tweets about topics that interest me in Deep Learning/Reinforcement LearningMistral AI @MistralAI
91K Followers 0 Following Fast, open-source and secure language models. Join us https://t.co/INALdNGvCPKepler @Kepler_L2
7K Followers 47 FollowingWeyaxi @Weyaxi
2K Followers 2K FollowingJohn (Zhiyao) Ma @johnma2006
286 Followers 62 FollowingSheldon Axler @AxlerLinear
12K Followers 29 Following Emphasis here on 3 of my books: (1) Linear Algebra Done Right; (2) Measure, Integration & Real Analysis; (3) Harmonic Function Theory (with Bourdon & Ramey).Harvey @harvey__ai
2K Followers 2 Following Building the most secure platform for legal and professional services. Raised $100M from @OpenAI, @Sequoia, @KleinerPerkins. Hiring: https://t.co/vfPpAcdyI3Ted Xiao @xiao_ted
11K Followers 682 Following I teach robots to be smarter @GoogleDeepMind. Tweets about robot learning, scaling, and large models. Opinions my own.Teven Le Scao @Fluke_Ellington
2K Followers 551 Following Researcher @MistralAI, producer @ my bedroom, no BLOOM slander authorized on this accountkenshin9000 @kenshin9000_
6K Followers 311 Following Working on Computer Vision and AI Safety. Twitter browsing account.Greg Schoeninger @gregschoeninger
424 Followers 1K Following CEO @ https://t.co/N4n4dXwLhC 🐂 usually plowing through data to build practical AI/ML, sometimes in my camper van with my dogI was credibly corrected in my DMs. 400 will be better than GPT-4 in every way.
I was genuinely impressed with how bad L1 regularization really is... And it's used *everywhere* in ML as some sort of silver bullet.
The shrinkage bias caused by L1 and cross-validation do not mix well, forcing an unnecessary trade-off between promoting sparsity and reducing bias. In the end, your reconstructed network ends up with a bunch of fake edges. 👎 Besides being slow and annoying to use... 6/N
The shrinkage bias caused by L1 and cross-validation do not mix well, forcing an unnecessary trade-off between promoting sparsity and reducing bias. In the end, your reconstructed network ends up with a bunch of fake edges. 👎 Besides being slow and annoying to use... 6/N
L1 regularization for sparse solutions - as usually taught - is actually terrible in practice! I’m always surprised how few people know this. To get good results, retrain with the sparsity pattern found from the initial L1 run, but without the regularizer. Works much better.
I was genuinely impressed with how bad L1 regularization really is... And it's used *everywhere* in ML as some sort of silver bullet.
In short, implicit sparsity is the key in understanding deep learning. The benefit of getting closer to L0 sparsity, is that it allows representation learning / selection from significantly larger candidates. imho, this is the reason why deeper DL works better. The end. 7/7
Even though L1 sparsity results in more bias than L0 sparsity, it is in some sense *exactly* accounted for in the number of non-zero elements. For Lasso, the effective degree-of-freedom is *equal* to the E[# of nonzero weights]. This is magic to me when I first saw it. 3/n
First, there is no straightforward L0 beats L1 or L1 beats L0. L1 (lasso) works better in low-signal-to-noise (SnR) regime, while L0 (subset selection) works better in high-SnR regime. The best of both-world is what's known as "relaxed lasso". 2/n
The ongoing L1-regularization on X is a deja vu of the Bertsimas-Tibshirani fight over best-subset selection vs lasso back in 2017, which resulted in the following paper. A short thread to cover a few not so well-known facts. 1/n
~Aug 2023: impossible to find a cluster of even 256 A100s May 2024: can find 512 H100s w next week delivery I guess everyone over bought?
@cnunezimages A super reliable, light, reusable heat shield is the biggest technical challenge remaining for Starship
This is the most Belgian thing. Likely smart guy. Works as warehouse operator. Cracks century unsolved cyphers as a hobby. I have 5-6 friends from school, pretty sure all smarter than me, doing random stuff in life (one police officer for example) and simply being content☺️
This is our home. We have nowhere else to go. We don’t want sharia. We don’t want blasphemy laws. We don’t want our women covered. We want our country back.
It is clear that elements of the Right in the UK are embittered at the rising population of Muslim and or ethnic communities in major cities. Let's be real, if you are waiting for a policy or party that will remove 1.4 million Muslims British Citizens from London or 350 thousand…
Maiden launch of Falcon Heavy, the most powerful operational rocket on Earth, with the first dual rocket landing
Great to know!! Hardware iteration at light speed.
@DrJimFan @Tesla_Optimus The new Optimus hand later this year will have 22 DoF
@Ravarora1 @joerogan Covid & Trump broke Sam Harris’s brain
Sam Harris is upset that the majority of @joerogan’s audience believes 1) Covid was a “non-issue,” 2) it was “not much worst than the flu” 3) Covid deaths were substantially inflated - “lots of people died with Covid, not from Covid” 4) “many, many people died from the vaccine.”…
Congrats to @Tesla_Optimus team on another stellar update! The video gives us a peek at their human data collection farm, which I believe is Optimus' biggest lead. What does it take to build such a pipeline? Optimus nailed all of the following: 1. Optimus hands are among the…
@elonmusk @DrJimFan @Tesla_Optimus The new Optimus hand, set to be released later this year, will have 22 Degrees of Freedom (DoF). This means the hand can move and rotate in 22 different ways, allowing for a high degree of dexterity and precision in its movements. This is a significant improvement over the…
@DrJimFan @Tesla_Optimus The new Optimus hand later this year will have 22 DoF
I was surprised & honored by a wonderful hand-made gift from Ned last night after the San Antonio Tesla Club Star Wars themed Light Show! My coverage of Giga Texas over the years has allowed me to connect with so many amazing people from all over, and its gestures like this…