All the Biomass of Earth, in One Graphic 🌍️ From the Archive: https://t.co/4XwMMvdeg6 https://t.co/IfpBuZSpPp
— Visual Capitalist (@VisualCap) Mar 31, 2024
from Twitter https://twitter.com/VisualCap
March 31, 2024 at 02:00AM
via IFTTT
All the Biomass of Earth, in One Graphic 🌍️ From the Archive: https://t.co/4XwMMvdeg6 https://t.co/IfpBuZSpPp
— Visual Capitalist (@VisualCap) Mar 31, 2024
Increasingly relevant book. https://t.co/JUQaiuDtIz https://t.co/fkDH6uia1m https://t.co/F238JS7TtN
— Marc Andreessen 🇺🇸 (@pmarca) Mar 31, 2024
Honestly, #DomoAI does an incredible job animating an image from a reference video. Their new models are seriously badass.🔥🤯 https://t.co/kIlqCNC0SS
— Pierrick Chevallier | IA (@CharaspowerAI) Mar 31, 2024
The AI tsunami is coming. @ycombinator's W24 mega-batch of 157 AI startups touching 20+ different sectors gives us a glimpse of where AI is headed this year. Here's a breakdown of the top 6 themes 🧵👇 https://t.co/5GM20kzurt
— Chrys Bader (@chrysb) Mar 31, 2024
https://t.co/T31iO7jqHH
— Physics In History (@PhysInHistory) Mar 30, 2024
Marc Andreessen is a prominent entrepreneur, investor, and software engineer. He advises that carrying a book and a notebook everywhere is an excellent idea and recommends buying paper copies of all books you plan to ever read. 25 books recommended by @pmarca 🧵 https://t.co/mLWdYAcQAA
— Reads with Ravi (@readswithravi) Mar 31, 2024
Claude 3 Prompt: you are going to act as a twitter influencer desperate for views. I want you to follow the following formula: I'm obsessed with How to (A) So I spent (B) + Hours studying how (C) (A) Here are (D) if you want (E) A = Something important and vague B = random #… https://t.co/JgyoIc16BT https://t.co/qldxfUY2il
— Riley Brown (@rileybrown_ai) Mar 30, 2024
Sources: Microsoft and OpenAI are a planning a new "Stargate" US-based supercomputer that could cost $100B to power OpenAI products (The Information) https://t.co/2XrytXCp92 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/qmnz9sGDs2
— Techmeme (@Techmeme) Mar 29, 2024
I’ll finally have a short break from travel soon. I can do either: (1) a blogpost on how using DSPy well answers most questions I see around building effective LM systems, (2) a DSPy program that shows what more complex compositions allow you to do. What would you like to see?
— Omar Khattab (@lateinteraction) Mar 29, 2024
Yann LeCun @ylecun delivered a lecture on Objective-Driven AI. He began with a reality check: "Machine Learning falls short compared to humans and animals!" Here's his insight on constructing AI systems that learn, reason, plan, and prioritize safety: 1/5 https://t.co/W0UOXAznPw
— Ksenia Se (@Kseniase_) Mar 29, 2024
Very interesting paper on using LLM generated code at Meta. 73% of recommendations passed review and is in production. WHAT. https://t.co/cna1o8BX0L
— Nick Dobos (@NickADobos) Mar 30, 2024
Explaining black box machine learning models is critical to gaining leadership's buy-in and trust. Here's 6 months of research on Explainable ML in 6 minutes (Business Case included). Let's go! 1. Explainable Machine Learning (ML): Refers to techniques that make the outputs and… https://t.co/hLwhRbBorz https://t.co/NYf04r2Mil
— 🔥 Matt Dancho (Business Science) 🔥 (@mdancho84) Mar 29, 2024
@NandoDF I think the concept of recovering a world model purely by prediction is definitely possible and has some similarities with Takens's theorem in dynamical systems https://t.co/gQ093MJBhp
— Abhishek Singh (@tremblerz) Mar 29, 2024
Claude 3 can now create flowcharts and diagrams. It is super simple and takes a few seconds. Here is how you can do it: https://t.co/ARhlBjQsL6
— Alvaro Cintas (@dr_cintas) Mar 28, 2024
Imagine not paying for all these AI tools anymore, having everything under one roof. Here's https://t.co/H89wi8Js3M – the one tool that brings it all together. → Generate images → Create content → Solve complex maths problems → Code → Chat with Claude 3, GPT-4, Gemini Pro… https://t.co/fHbFJY6QFD https://t.co/22WHOBtCXG
— Robin Delta (@heyrobinai) Mar 22, 2024
How good is GPT-4-Vision at extracting text from images? I wanted to find the limit - but I found weirdness instead Most surprising: GPT-4V performance varies depending on the *structure* of text it sees Let me explain A set of images with progressively more text was… https://t.co/mZIdICf2ne https://t.co/CNm1WMByyN
— Greg Kamradt (@GregKamradt) Mar 28, 2024
🚨#PromptShare🚨 Infographic diagram Let's switch to an Infographic diagram mode #prompt, a very useful token that will enhance all kinds of characters, vehicles, or objects. A character sheet of [subject] with various angles and full body views. Infographic diagram, details… https://t.co/ICYa7rce8K https://t.co/OU6LMlslwi
— Pierrick Chevallier | IA (@CharaspowerAI) Mar 28, 2024
Copy AI is launching a new Go-To-Market (GTM) Platform. This requires doing research and market research to analyze potential folks to reach out to and get a good sense of the competitive landscape. They chose to use the Perplexity APIs for this, and we're also providing 6 months… https://t.co/77sldhbwsc
— Aravind Srinivas (@AravSrinivas) Mar 28, 2024
We built a tool that generates AI clones of people and @ycombinator used it in their official podcast today! 🎉 Check out the clip below to see AI clones of @garrytan, Jared (@snowmaker), @harjtaggar, and @sdianahu talking about AI. Very meta Here's a 🧵on how we built it: https://t.co/QCAQV5Qzmk
— Lina Colucci, PhD (@lina_colucci) Mar 28, 2024
Updated list of diffusion model tutorial sources, from my lecture on diffusion. https://t.co/7M4bMi1xjt https://t.co/cZADri0O3X
— Kosta Derpanis (@CSProfKGD) Mar 28, 2024
“I don't need to change the world overnight, I am gonna change the world over the next 50 years”—Jensen, NVIDIA CEO, 2003 https://t.co/FjfpfmQnze
— Brian Roemmele (@BrianRoemmele) Mar 28, 2024
This is how I know Big Tech could cut 80% of their headcount and perform better. https://t.co/VGWshu5EhN
— Marc Andreessen 🇺🇸 (@pmarca) Mar 27, 2024
These things are amazing especially the Claude models Lumentis can now do Mermaid diagrams and Latex for less than 10 cents with Haiku - go check it out https://t.co/jylXz7Zb31 https://t.co/dsmSw0rHKg
— Hrishi (@hrishioa) Mar 26, 2024
Why Google *couldn’t* make GPT-3? answer from David Luan, a former VP Eng of OpenAI (~30th hire at OpenAI) and former LLM tech lead at Google Brain https://t.co/st5UpfeK4P
— Kourosh (@kouroshshafi) Mar 26, 2024
Seems like the very Erik Meijer himself left Meta to... build a new DSPy? Well, we welcome some competition when he's done! Been strange being the *only* programming model that compiles high-level code to optimized LM parameters, i.e. instructions, examples, and weight updates. https://t.co/Hmq3JJMMdo https://t.co/ndoc3NEVI3
— Omar Khattab (@lateinteraction) Mar 26, 2024
Claude Opus is extremely impressive for doing research that finds, cites, and summarizes research papers. Highly recommend picking a topic and trying to steer it that direction. Have it write outlines that include data, stats, probabilities etc and cite papers directly.
— Patrick OShaughnessy (@patrick_oshag) Mar 26, 2024
On Shein: - Their clothes are 65% polyester (double Zara & H&M!) - - Laundering polyester is responsible for ~35% of the microplastics in the ocean - They ship between 2-3 billion items a year - Their emission grew by 52% last year - They've 10x'ed their lobbying spend in… https://t.co/WAXnDMTSgB https://t.co/WfYfYrVihO
— Kyla Scanlon (@kylascan) Mar 25, 2024
Meet Devika: An Open-Source AI Software Engineer that Aims to be a Competitive Alternative to Devin by Cognition AI Quick read: https://t.co/8Sqn1X6fQn Github: https://t.co/WfkRF7bND7 #ArtificialInteligence https://t.co/SUYSLKVw3y
— Marktechpost AI Research News ⚡ (@Marktechpost) Mar 25, 2024
United States v. Apple Apple is being sued by the DOJ, but most of the complaints aren't about the App Store. I think, though, Apple's approach to the App Store is what led to this case. https://t.co/RQXhqqddqS
— Stratechery (@stratechery) Mar 25, 2024
Apple Researchers Propose a Multimodal AI Approach to Device-Directed Speech Detection with Large Language Models Quick read: https://t.co/2weLZ1oMiE Paper: https://t.co/cz4QrgZvgb #ArtificialIntelligence https://t.co/EPwdMtcMII
— Marktechpost AI Research News ⚡ (@Marktechpost) Mar 25, 2024
It's been cool checking out Devin, an "AI software engineer". But I do find it super ironic that with all of this insane futuristic tech, Devin still debugs code by putting in print statements and looking at the output. https://t.co/kXu4jZ7D1L
— Vijay Pande (@vijaypande) Mar 25, 2024
Gemini 1.5 Pro can read large Github repos (225 files and 727,000 tokens in my test) and answer questions with links to source files! This might devaluate programmers' value, especially seasoned ones. $GOOG https://t.co/j5J8UAZZn9
— Hongcheng (@hzhu_) Mar 24, 2024
"Los Angeles is, crucially, not really a 'place' - it is viewed by even its residents as simply a kind of loosey-goosey, intangible mirage. Because the city doesn’t really (literally or figuratively) have a true center, it’s kind of a thing that any kind of dream can be imposed… https://t.co/s6emjupCa2
— Andy Weissman (@aweissman) Mar 23, 2024
Jotted some thoughts about how AI opens up new killer app possibilities which can be wedges to systems of record https://t.co/XdXbUsfN9u
— Jack Altman (@jaltma) Mar 23, 2024
Hayao Miyazaki & Akira Kurosawa discuss the importance of storyboards in filmmaking. https://t.co/C9bH8ErlDQ
— DepressedBergman (@DannyDrinksWine) Mar 23, 2024
My top 5 picks that I believe will dominate the rest of this decade 👇 1. $NVDA | Semiconductors • They're set to cement their status as the core of the AI & gaming revolutions, making their GPU technology indispensable for the future tech ecosystem. 2. $CRWD | Cybersecurity… https://t.co/Qo7NIS8UHW
— Shay Boloor (@StockSavvyShay) Mar 23, 2024
Devin Alterative: Open Source AI Software Engineer 🔥 Introducing Devika 🚀 Run Devika on Your Computer 👩🏫 Step-by-Step Tutorial 🧠 AI-Powered Features 🛠️ Easy Installation 🌐 It's Free 🔗 Ollama Integration Subscribe: https://t.co/RTY3pSVFGl YT: https://t.co/mKP8wUZ0tm… https://t.co/TzLN1HZ2iC https://t.co/K2idA5v5Lb
— Mervin Praison (@MervinPraison) Mar 23, 2024
I learned a lot from this: https://t.co/WpOqK8tW48. (If you liked Nonlinear Dynamics and Chaos, @stevenstrogatz is quoted on the back saying that it's a "masterpiece" that's "nothing short of revolutionary".) https://t.co/pu0z5HxShc
— Patrick Collison (@patrickc) Mar 24, 2024
“Spend less time in conference rooms and more time just trying to make your product as amazing as possible.” ~ Elon Musk https://t.co/1YazQrtBbi
— Jon Erlichman (@JonErlichman) Mar 23, 2024
👨🎓Advanced RAG Series: Generation and Evaluation In the fifth part of this series, we look at techniques for: 🔊 Generation (CRAG, Self-RAG, RRR) 🧮Evaluation (RAGAs, Langsmith, DeepEval) Another awesome blog by @divyanshu_van https://t.co/CeVgksyhnT https://t.co/xpswUsh09H
— LangChain (@LangChainAI) Mar 23, 2024
This book is 🔥 BEYOND BELIEF. Highly recommend 10/10. @BennSteil https://t.co/CXqRnePQay https://t.co/phJpBOT32Z
— Marc Andreessen 🇺🇸 (@pmarca) Mar 22, 2024
Working on an episode about Quentin Tarantino and loved this gem: No wonder you can’t do it —You acquiesce to defeat before you even begin. https://t.co/L1gn0pU2V5
— David Senra (@FoundersPodcast) Mar 22, 2024
Devin is wild AI software engineer. People have been getting wild results like taking website building requests on Reddit and trying to charge people🤯 8 examples: https://t.co/WLzWu40BO4
— Min Choi (@minchoi) Mar 22, 2024
Steve’s favorite quote: We are what we repeatedly do. Excellence, then, is not an act, but a habit. Spotify: https://t.co/f5SCgXqRS2 Apple: https://t.co/awe1d3pyuw
— David Senra (@FoundersPodcast) Mar 22, 2024
Steve Jobs on why we should spend our time making great things: “One of the ways that I believe people express their appreciation to the rest of humanity is to make something wonderful and put it out there. And you never meet the people. You never shake their hands. You… https://t.co/D3aqUuWSNt https://t.co/80mNQaDQgz
— David Senra (@FoundersPodcast) Mar 22, 2024
Introducing `claude-investor` 📈 The first Claude 3 investment analyst agent. Just provide an industry, and it will: - Find financial data/news for key companies - Analyze sentiment/trends for each - Rank stocks by investment potential + price targets And it's open-source! https://t.co/7i3sjwio5z https://t.co/XuOCLcGMcM
— Matt Shumer (@mattshumer_) Mar 22, 2024
I KNEW IT https://t.co/mfYrz6rpGN
— Marc Andreessen 🇺🇸 (@pmarca) Mar 21, 2024
Maestro made this for me in about two minutes lol https://t.co/B3h0AZHDZl https://t.co/AhTjfgPvXu
— Pietro Schirano (@skirano) Mar 20, 2024
manual data labeling is (almost) dead 1,500,000 images auto-annotated within 2 weeks of release. now, we also support automatic segmentation labeling. ↓ read more about open-source models that power this feature https://t.co/HjRUdPG4pA
— SkalskiP (@skalskip92) Mar 19, 2024
"If you want to hire great people and have them stay working for you, you have to let them make a lot of decisions - you have to be run be ideas, not hierarchy. The best ideas have to win !" — Steve Jobs, D8 conference, 2010
— Steve Jobs Stories (@SJobs_Stories) Mar 19, 2024
📝 This pattern of having a large planning-model break down a larger task into subtasks -- and then "right-size" smaller models or services to accomplish each subtask -- is very, very powerful. Imagine being able to delegate a simple task to a small OSS model, and more complex… https://t.co/1zAmXVO3x2 https://t.co/QFPqF2qWj0
— 👩💻 Paige Bailey (@DynamicWebPaige) Mar 19, 2024
Claude-ception: Teaching Claude3 to prompt engineer itself Claude3 Opus is excellent at prompt engineering. @alexalbert__ recently laid out a nice workflow: write an prompt, run it on test cases, grade responses, let Claude3 Opus use grades to improve prompt, & repeat.… https://t.co/QkdjG4hWRI https://t.co/FVNpBZHxeV
— LangChain (@LangChainAI) Mar 19, 2024
Open source observability for your RAG pipeline in 2 lines of code with @langfuse! The richer your application, the more important observability becomes. We're delighted to launch this integration, which combines ➡️ Tracing ➡️ Prompt management ➡️ Evaluation all into one… https://t.co/44PyU85MNZ https://t.co/vONQff1oPb
— LlamaIndex 🦙 (@llama_index) Mar 18, 2024
3-18-2024 ($) •Apple Releases AI Research Paper •Apple + Gemini? https://t.co/a9X5cCni80
— Stratechery (@stratechery) Mar 18, 2024
When I was first discovering AI agents, I came across the "anatomy of autonomy" visual by @swyx. The definition of agents is still fluid, so I am trying to make my own visual. What do you think is missing? 🧠 Brain = database (e.g., @pinecone) + LLM (@OpenAI, @MistralAI,...)… https://t.co/1RCsg5CbI3 https://t.co/FFdGtlbXLb
— Tereza Tizkova (@tereza_tizkova) Mar 18, 2024
Prepare yourself for some new @nvidia GPUs being announced today at GTC with the name of Blackwell (B-series). Source: https://t.co/8Ti2i2XWmV TL;DR; - 192GB of memory, with a potential increase to 288GB per GPU. - B100 air-cooled 700W can even slide into existing servers that… https://t.co/D91eT12FWl
— Philipp Schmid (@_philschmid) Mar 18, 2024
The math behind the Apple Logo [🎞️ mat.h_director] https://t.co/exliTr6hwX
— Massimo (@Rainmaker1973) Mar 18, 2024
Create Illustrated Storybooks Instantly with Crew AI Agents! With the power of Groq @GroqInc ✨ Quick Storybooks 🖌️ AI Illustrations 🛠️ Crew AI Custom Tools 🚀 @ollama integration 🤖 AI-Powered Subscribe: https://t.co/RTY3pSVFGl YT: https://t.co/RjbyDlQINd @joaomdmoura… https://t.co/kagHrjAVj0 https://t.co/52KlRuAGsT
— Mervin Praison (@MervinPraison) Mar 17, 2024
Look at this. Steven Hao gave Devin (the AI software engineer) access to his account, and Devin is just… doing his job for him Devin is doing what best human devs do: 1) He finds an issue 2) He reads through the documentation but can’t figure it out 3) BUT Devin doesn't just… https://t.co/wliVV0c8Z2 https://t.co/769pI31cZE https://t.co/VtPMx0r5Dr
— AI Notkilleveryoneism Memes ⏸️ (@AISafetyMemes) Mar 17, 2024
here's your DEEP DIVE into @grok's architecture! I just went through the https://t.co/8Y5cjeImg6, for this 314B open source behemoth with *no strings attached*. 👇🧵 https://t.co/CraHKGqILe
— Andrew Kean Gao (@itsandrewgao) Mar 17, 2024
Clarity AI is open source and you can host it yourself and run it for free, or run it on Replicate If you want to use it without writing any code, I have my own hosted and paid version at https://t.co/buhETYiU02 which helps support my work as an AI dev It's 25% cheaper than… https://t.co/iX7IRAqx2a https://t.co/8DsCRPHAC6
— philipp1337x (@philz1337x) Mar 16, 2024
Devin, take this ComfyUI workflow and write it in pure Python. Optimise for inference speed and good quality output. Send me a push notification when you’re done.
— fofr (@fofrAI) Mar 16, 2024
find a collaborator who pushes you to be better
— Greg Brockman (@gdb) Mar 17, 2024
SpaceX is radically reducing the cost to send stuff into orbit. https://t.co/NyGBidBnel https://t.co/FqGIpjrk2S
— Timothy B. Lee (@binarybits) Mar 15, 2024
Devin has some competition. Maisa KPU is focused on complex problem-solving and reasoning. The benchmarks they are posting surpass the frontier. It is launching in beta today. https://t.co/GlWPTGn4RM https://t.co/meLBX0DMI8
— Andrew Curran (@AndrewCurran_) Mar 15, 2024
We’re sharing our technical report for Devin’s results on SWE-bench: https://t.co/9Aoz0MLx7x Highlights in 🧵 https://t.co/8nI7cqsmSY
— Cognition (@cognition_labs) Mar 15, 2024
Sources: TikTok's US revenue hit a record $16B in 2023; ByteDance revenue grew 40% YoY to $120B in 2023, with a net profit of $28B (Financial Times) https://t.co/D5shROGLRx 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/mqGdB3E15R
— Techmeme (@Techmeme) Mar 15, 2024
🌀Check out RAFT: Retrieval-Aware Fine Tuning! A simple technique to prepare data for fine-tuning LLMs for in-domain RAG, i.e., question-answering on your set of documents 📄 Exciting collaboration with @berkeley_ai 🤝 @Azure 🤝 @AIatMeta MSFT-Meta blog:… https://t.co/5uTO6i4dBm https://t.co/KdYPh5Hapl
— Shishir Patil (@shishirpatil_) Mar 15, 2024
Max Levchin on how Peter Thiel reacted to losing his entire investment in Luke Nosek’s failed startup When assembling the found team for PayPal, Max Levchin wanted to recruit Luke Nosek. But he was worried at how Peter Thiel would react given that Peter lost his entire… https://t.co/HSvTvktuNH https://t.co/zTOjyR7veZ
— Startup Archive (@StartupArchive_) Mar 15, 2024
Partly true! Greatness also requires intelligence; it's true, and we should speak truth. Smart people tend to deprecate their own intelligence; instead we should credit it, and build AI to enhance everyone's intelligence. https://t.co/uCKJci9aU5
— Marc Andreessen 🇺🇸 (@pmarca) Mar 14, 2024
A great investor is a great rejector . You have to read this ! Source : https://t.co/2OnWw2Gf8c https://t.co/Ibew7jKZvR
— Finding Compounders (@F_Compounders) Mar 14, 2024
Scattered Mixture-of-Experts Implementation - Presents ScatterMoE, an implementation of Sparse Mixture-of-Experts on GPU - Enables a higher throughput and lower memory footprint repo: https://t.co/kJZjY6Q4O7 abs: https://t.co/xxIx1a2vu2 https://t.co/IqWxmNln6W
— Aran Komatsuzaki (@arankomatsuzaki) Mar 14, 2024
Your best people consistently make problems go away. Your worst people consistently create problems that you need to fix. When you do the inner work to release your personal desire to fix things, then you can fire fast and lead with maximum accountability and minimum problems.
— bradford cross (@bradfordcross) Mar 14, 2024
Claude 3 Haiku processes images in 1.6k tokens. It corresponds to 40x40 patches, and I would guess patches of 8x8 using a traditional VQGAN so image input at 320x320 px which seems reasonable. https://t.co/1ccNyBsHvQ https://t.co/EhWp6mCNL4
— Boris Dayma 🖍️ (@borisdayma) Mar 13, 2024
I ❤️ @cohere . They opened another huge dense embedding dataset. This time with their new v3 model. This is invaluable in performance & scale testing over real data. Thank you! https://t.co/gJ1Cv94iRD
— Benjamin Trent (@benwtrent) Mar 13, 2024
Vision Pro is an over-engineered "devkit" Hardware bleeds genius & audacity but software story is disheartening. What we got wrong at Oculus that Apple got right Why Meta could finally have its Android moment // very thoughtful / insightful post from @hbarra… https://t.co/L5wBgGur1b
— Steven Sinofsky (@stevesi) Mar 13, 2024
A couple notes from my @cognition_labs story today. I first met the CEO @ScottWu46 and his brother @WuNeal when they were teenagers in sport coding competitions. People talked about them with reverential awe. They helped make the US competitive against Russia, China and others.… https://t.co/RFtQxLeEDb
— Ashlee Vance (@ashleevance) Mar 13, 2024
Stripe's 2023 annual letter: https://t.co/1NdjDdHCtU.
— Patrick Collison (@patrickc) Mar 13, 2024
i never believe recorded demos so I reached out to the @cognition_labs team for early access to try for myself and got it! will be sharing my unfiltered opinions on #devin here. 🧵🧵 1/n https://t.co/fotGHh6KlU
— Andrew Kean Gao (@itsandrewgao) Mar 12, 2024
We're excited to publish a definitive set of @AnthropicAI cookbooks for building RAG and agents with Claude. Learn how to build basic RAG (top-k VectorStoreIndex), to advanced RAG capabilities (routing, query decomposition), to sophisticated document agents, to multi-modal.… https://t.co/FTovjzzI8t https://t.co/ZlecVkr1Nu
— Jerry Liu (@jerryjliu0) Mar 11, 2024
Midjourney's new character consistency feature changed the whole game. I mean, look at this 😮💨🤌 Testing some very interesting prompts and workflows, that I'll share later today https://t.co/U0J9Fc16tF
— Halim Alrasihi (@HalimAlrasihi) Mar 12, 2024
# automating software engineering In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like: 1. first the human performs all driving actions… https://t.co/0mPCh8amba https://t.co/u2EXfxcV6e
— Andrej Karpathy (@karpathy) Mar 12, 2024
3-11-2024 •Congress Considering TikTok Ban •TikTok's War •What Now? https://t.co/pyy1as7q3T
— Stratechery (@stratechery) Mar 11, 2024
ICYMI from Friday. Our best DSPy prompt optimizer yet for complex language programs! (though more coming soon) With great animated descriptions of how it works, step by step, and a free (cached) google Colab example: https://t.co/naltuCI04P
— Omar Khattab (@lateinteraction) Mar 11, 2024
face + style – using IPAdapter for style, InstantID for consistency, prompt for steering and depth controlnet for continuity with original. This also uses DreamshaperXL Lightning, so these images are in just 4 steps. Will share details soon. https://t.co/uo8Ap1JBF0 https://t.co/s5VQKcnQTJ
— fofr (@fofrAI) Mar 11, 2024
Google presents: Stealing Part of a Production Language Model - Extracts the projection matrix of OpenAI’s ada and babbage LMs for <$20 - Confirms that their hidden dim is 1024 and 2048, respectively - Also recovers the exact hidden dim size of gpt-3.5-turbo… https://t.co/XiDzNGesZd https://t.co/8MxEC7BSEa
— Aran Komatsuzaki (@arankomatsuzaki) Mar 12, 2024
76% of Shopify revenue came from financial services last year. Embedded finance is undefeated as the best way to drive for SaaS businesses if they solve adjacent customer problems https://t.co/PFmVsmB1pN
— Simon Taylor (@sytaylor) Mar 11, 2024
What is a Mixture-of-Experts (MoE)? A Mixture of Experts (MoE) is a machine learning framework that resembles a team of specialists, each adept at handling different aspects of a complex task. It's like dividing a large problem into smaller, more manageable parts and assigning… https://t.co/QjerpGzGur https://t.co/nE60uURNKR
— Akshay 🚀 (@akshay_pachaar) Mar 11, 2024
When Cornelius Vanderbilt died he had more money than the U.S. Treasury This story is wild! New episode available now! https://t.co/BSHwWPYMYy
— David Senra (@FoundersPodcast) Mar 11, 2024
Just tested Midjourney's "Character Consistency" feature (-- cref). It's promising! I don't think it beats a good #LoRA (yet). Especially when paired w. IP Adapter. Below are some quick tests I ran on Scenario (high-def pics below + more updates coming👇👇) What do you think? https://t.co/UebcdcgDgT
— Emm (@emmanuel_2m) Mar 12, 2024
Sunday morning read: PersonaLLM: Investigating the Ability of Large Language Models to Express Personality Traits 🎭 PersonaLLM is a new work exploring personas through prompting the language with the Big 5 personality traits: Extraversion, Agreeableness, Openness,… https://t.co/aNA5vnedmk https://t.co/NXa00jJwBs
— Erika Cardenas (@ecardenas300) Mar 10, 2024
RAG over Complex PDFs V2 📑 If you’re looking for a definitive tutorial on solving RAG over your messy, complicated PDFs (with messy formatting, fonts, tables, charts), this is the tutorial for you. In these settings, naive RAG doesn’t work ⛔️. This video by @AIMakerspace is a… https://t.co/pOJy5AkxZa https://t.co/V4iryJ6DDI
— LlamaIndex 🦙 (@llama_index) Mar 10, 2024
One of my fav blog posts I've read in a while "An Opinionated Guide to ML Research" by @johnschulman2 (openai cofounder) Goes over: - honing your taste - improving your general knowledge to do great work - falling in trap of working on unimportant problems https://t.co/us4oYWlRVH
— peter! 🥷 (@pwang_szn) Mar 10, 2024
From a Jim Carrey speech: “My father could have been a great comedian, but he didn't believe that that was possible for him. And so he made a conservative choice. Instead, he got a safe job as an accountant. And when I was 12 years old, he was let go from that safe job. And our… https://t.co/oodNHu0a66
— Kevin Smith (@KevinBenSmith) Mar 9, 2024
Did a basic implementation first, using typed predictors. Will generalize it later (general def of agents, vector stores, tools etc) Explanatory article in second tweet. Link to github in third tweet. https://t.co/n2jLQjhEFo https://t.co/PR2mMT3ijp
— Franck SN (@ndzfs) Mar 10, 2024
A minimally intrusive DSPy logger. A first step towards building a self-tuning LM pipeline. Creating realistic training examples is a pain! Here, I can deploy the app internally and start collecting real usage examples with feedback. Exciting! https://t.co/NH7pcxKoBs
— Dhar Rawal (@RawalDhar) Mar 9, 2024
Interviews with Garry Tan, Paul Graham, YC partners, and others about Tan reshaping YC in a "return to roots", as some founders complain about politicization (@alexrkonrad / Forbes) https://t.co/3go5KboAPw 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/rJj93lB35n
— Techmeme (@Techmeme) Mar 9, 2024
Got a pipeline with **multiple prompts**, like a DSPy program? What's the right way to jointly optimize these prompts? Introducing MIPRO, a Multi-prompt Instruction Proposal Optimizer. We integrated MIPRO into DSPy. It can deliver +11% gains over existing DSPy optimizers! 🧵👇 https://t.co/hFfqSsELyT
— Krista Opsahl-Ong (@kristahopsalong) Mar 8, 2024
Masked self-attention is the key building block that allows LLMs to learn rich relationships and patterns between the words of a sentence. Let’s build it together from scratch… The big picture: Large language models are based upon a deep neural network architecture called a… https://t.co/8FTKvDlNbl https://t.co/EjshqtHJcO
— Cameron R. Wolfe, Ph.D. (@cwolferesearch) Mar 8, 2024
This discussion with Jensen Huang at the Stanford economic summit this last week is a must watch. https://t.co/7eR68YO9e5
— Ben Bajarin (@BenBajarin) Mar 9, 2024
My notes from Larry Gagosian (Billionaire Art Dealer) turned into maxims: 1. The best way to raise the price of something is to say that you would never sell it. 2. The competitive drive of self-made billionaires does not go into remission once they've made a fortune. 3.… https://t.co/NHT8kEm8kr https://t.co/appHREDfQn
— David Senra (@FoundersPodcast) Nov 1, 2023
Jensen Huang: Do irreplaceable work. Work on things which without you, wouldn't / couldn't get done by anyone else. https://t.co/hsTNPMsdGU
— BUILD OR DIE (@BUILD_OR_DIE) Mar 7, 2024
If you don't like the default Claude response style, you can use a "priming prompt" to ask for a different response style or format. Here's an example of a priming prompt that gets Claude to be more conversational. https://t.co/bOfOuhdCLE
— Amanda Askell (@AmandaAskell) Mar 8, 2024
So I finished this beautifully written book by @francoisfleuret, highly recommend, like you wouldn't believe this, but if you know some statistics you can skim the whole thing and get a (obviously incomplete) very rough shape understanding of how all this works in like an hour https://t.co/km2ZR1TK3S
— alz (@AnthonyLeeZhang) Mar 8, 2024
3/n Paper link: https://t.co/Bg1LOd4WoZ
— Carlos E. Perez (@IntuitMachine) Mar 8, 2024
We’re excited to feature a novel hierarchical code splitting 🧑💻✂️ technique that allows you to build advanced RAG/agents for code understanding. Our `CodeHierarchyNodeParser` (big thanks to ryanpeach 🙌) converts a big code file into a hierarchy. The scope body of a long… https://t.co/rIqRYOaLxv https://t.co/dZZTnWAY7Z
— LlamaIndex 🦙 (@llama_index) Mar 8, 2024
Link to paper: https://t.co/RC5SuHXiGH
— Carlos E. Perez (@IntuitMachine) Mar 8, 2024
Paper: https://t.co/tPbtAinfvF
— elvis (@omarsar0) Mar 7, 2024
A new Amazon review dataset just dropped on HF: Bridging Language and Items for Retrieval and Recommendation https://t.co/kfiJ8tetp0 https://t.co/ORAPC1BoxZ https://t.co/LEbqkHxvY2
— Jo Kristian Bergum (@jobergum) Mar 7, 2024
Alright, this is the last one, before I release my workflow for creating authentic images with your AI characters! 👫 Which one of these is AI and which one is real? 👀 Choose you number(s) 👇 https://t.co/mE3QtbvZsq https://t.co/tLmejkaf1e
— Halim Alrasihi (@HalimAlrasihi) Mar 7, 2024
At 19, Sam Altman dropped out of Stanford and raised $30M to found Loopt Here's his pitch: https://t.co/msKn4cRYLo
— PitchDeckGuy (@BetterPitchGuy) Mar 6, 2024
RAG for AI-Generated Content Cool survey paper providing an overview of RAG used in different generation scenarios like code, image, audio,... I like the taxonomy of RAG enhancements and it seems to mention a lot of key RAG papers. https://t.co/WxQXS1IRQt https://t.co/73wi0nKEx8
— elvis (@omarsar0) Mar 6, 2024
RAG with Persona 🎭 If you ask the same question to multiple people, there is a strong chance each person will have a different response. Here’s a new demo on RAG with Persona using DSPy, @cohere, and @weaviate_io! When building or using chatbots, it’s important to get a… https://t.co/smDCTHsZ68 https://t.co/qXq3FCIOGW
— Erika Cardenas (@ecardenas300) Mar 6, 2024
Here is an incredible Claude 3 prompt for engineers. Use it to speed up any code by identifying inefficiencies and rectifying them: --- <prompt_explanation> You are a world expert in making code run faster. You use any resource you can to do so. Given some code, first, explain… https://t.co/mpNL1nXjHm
— Matt Shumer (@mattshumer_) Mar 6, 2024
A fairly concise breakdown of our thoughts on the economic case for gen AI. What problems it's clearly suitable for. What problems it clearly isn't. And what still needs to be answered. https://t.co/PdugtcPa7W
— martin_casado (@martin_casado) Mar 6, 2024
$CRWD CEO George Kurtz throwing more shade at $PANW: "It is the customers trapped in stitched-together pseudo platforms that are suffering from budget fatigue. Not CrowdStrike." https://t.co/5pekbto2tH
— Brad Freeman (@StockMarketNerd) Mar 5, 2024
I delivered my usual chaotic high energy style of presentation about the @GroqInc architecture. There was some interest in my Haskell Haste DSL for programming our LPUs via our GTen IR. I talked a bit about how our LLM deployments are powered by our chips and toolflow. https://t.co/ouf0wTscTM https://t.co/WlYdj5tSBu
— Satnam Singh (@satnam6502) Mar 3, 2024
“Basically you iterate until you achieve product/market fit, and then you accelerate… If you have the runway, you’re not dead. You iterate until you survive and thrive. And then you become big.” https://t.co/FgbfuKzcrC
— Mike McGuiness (@mikemcg0) Mar 5, 2024
I’ve been reading Vaclav Smil forever. This is yet another amazing write up from him on the extreme challenge of a second large energy transition TLDR, it is going to take way, way longer than anyone wants it to. Must read: https://t.co/t8e8A7zJNB
— Patrick OShaughnessy (@patrick_oshag) Mar 6, 2024
Well done by @jon_prosser. I appreciate him pointing out how jaded most tech reviewers are and because of that entirely miss the forest from the trees. Quality production and observations about where this category can develop. https://t.co/6JFLsrEELL
— Ben Bajarin (@BenBajarin) Mar 5, 2024
There's a difference between what men like to read and what women like to read. https://t.co/YZABb3Ujdj
— i/o (@eyeslasho) Mar 5, 2024
The latest ML/AI research (Part 1): ▪️ 1-bit LLMs ▪️ bGPT ▪️ ChatMusician ▪️ StructLM ▪️ StarCoder2 ▪️ FUSECHAT and more 🧵 https://t.co/6vxgwYXs6r
— TuringPost (@TheTuringPost) Mar 5, 2024
Damn. Claude just kicked the crap out of GPT-4 for a basic web programming question I needed a quick answer to. Insta-switch. https://t.co/rCpqNvYZeI
— gfodor.id (@gfodor) Mar 5, 2024
🅰️🔧 Anthropic Tool Calling 🅰️🔧 We've been hacking away with the impressive new @AnthropicAI Claude-3 models, and shipped an experimental wrapper for consistently structured Pydantic/JSON outputs in both Python and JS! Official tool calling support is on the horizon, but since… https://t.co/lldeDTb3GC https://t.co/tD6qXVz3Q1
— LangChain (@LangChainAI) Mar 5, 2024
Can GPT-4-Vision Play Texas Hold'em Poker? I used Multimodal Gamer to let GPT-4-Vision control my mouse and browser to play. After a few hands, GPT-4-Vision's pot was up, and it had a couple of wins. https://t.co/4h0EE1YWFD
— Josh Bickett (@josh_bickett) Mar 4, 2024
I'm convinced. Claude is a now a serious competitor not just on a technical level but an emotional one too https://t.co/4yH29RjEuz
— Kevin Fischer — soul/acc (@KevinAFischer) Mar 4, 2024
With Claude 3 released today, I updated my token-based pricing providers sheet.📊 Including @OpenAI, @GoogleDeepMind, @AnthropicAI, @MistralAI and some open model providers. 👉 https://t.co/o9V36negQs 👨🏫 Note: To understand the TCO between the providers, you also need to look… https://t.co/lr7sbvt25u https://t.co/gdpkjVZ5Gv
— Philipp Schmid (@_philschmid) Mar 4, 2024
Trying Claude 3 today by @AnthropicAI for https://t.co/ZrwuHw4Gz9 At first glance it feels slightly better than GPT4 and way better than Mistral etc. The first thing I notice is that its responses seem MUCH more human than LLMs before it "I know it may not feel like it right… https://t.co/eo5dLSnhkI https://t.co/nJIKhEmEG2 https://t.co/8E8BkM2Zy9
— @levelsio (@levelsio) Mar 4, 2024
More here: https://t.co/6GZvHUbHpR
— elvis (@omarsar0) Mar 4, 2024
JUST IN: Anthropic announces Claude 3! Here is what you need to know: • Claude 3 consists of a family of three models (Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus) • Claude 3 Opus (the strongest model) seems to outperform GPT-4 on common benchmarks like MMLU and… https://t.co/QO3FToJHdm https://t.co/t2BNES5i1X
— elvis (@omarsar0) Mar 4, 2024
Wow, Claude 3 is incredible. https://t.co/7qYBWvKf7z
— Matt Shumer (@mattshumer_) Mar 4, 2024
Got it to work. Pass a Pydantic type and a text, receive JSON retrievals from the text. Yeah, DSPy is very cool. https://t.co/4pGMFxng6X https://t.co/kwlG5vOWSR
— Cainã Costa (@cfcosta_) Mar 3, 2024
"You see here I have a sad face ;-( , since tokenization is my least favorite part of working with LLMS, but unfortunately, this is necessary to understand" There are many reasons why people would consider @karpathy as a 🐐 AI Engineer. Tackling essential aspects, even if they… https://t.co/EsphbdtWcc https://t.co/bpSgsrujkG
— Itamar Friedman (@itamar_mar) Mar 3, 2024
I've been generating RPG players by the hundreds thanks to DSPy. And all I need is to define a pydantic class like so: https://t.co/WBsgyONt7T
— Franck SN (@ndzfs) Mar 3, 2024
A lot of good ways to split your data for RAG are dependent on the file type, for instance a Markdown-aware splitter (that can also parse out tables), or a JSON splitter. For pure unstructured text semantic chunking is a good approach. Because you can't make proper use of the… https://t.co/N3aGQZ13iW https://t.co/wGibj6mGQC
— Jerry Liu (@jerryjliu0) Mar 4, 2024
A Comprehensive Guide to Semantic Chunking for RAG 📑 A better approach to naive sentence/token chunking is to group semantically similar information, leading to better retrieval + more complete synthesis. This post by Florian June is a comprehensive guide towards semantic… https://t.co/YogzlUYjfm https://t.co/VzNtLRxsAf
— LlamaIndex 🦙 (@llama_index) Mar 3, 2024
I did a double take when I saw this. There are caveats, but this is still a remarkable thing to consider. https://t.co/JrmwvAifxK
— John Carmack (@ID_AA_Carmack) Mar 2, 2024
₿𝗥𝗘𝗔𝗞𝗜𝗡𝗚: Watch Michael @Saylor give first speech after his company broke records hodling $12,000,000,000 worth of bitcoin. https://t.co/yU4eiVBiBt
— Documenting ₿itcoin 📄 (@DocumentingBTC) Mar 3, 2024
A look at Apple's hubris that doomed the electric car project, spring hardware updates without a dedicated event, and Dan Riccio's retirement plans (@markgurman / Bloomberg) https://t.co/7RhRKA65gl 📫 Subscribe: https://t.co/OyWeKSRpIM https://t.co/hl14UFkwqv
— Techmeme (@Techmeme) Mar 3, 2024
When I reread this, I can't believe how many people were outraged by it when I wrote it in 2016: https://t.co/lpeUO5ZSrM
— Paul Graham (@paulg) Mar 3, 2024
Klarna has made headlines with its AI assistant (powered by OpenAI) handling two-thirds (2.3 Million) of Klarna's customer service chats in the last months. 💬 According to Klarna, the AI Assistant: 💸 Can handle refunds, returns, payment-related, cancellations and more 🚀… https://t.co/NGG8pvfKUA https://t.co/sTbS2O1Dxe
— Philipp Schmid (@_philschmid) Mar 3, 2024
ChatGPT is no longer the Top G. Microsoft just released Copilot Pro, and it's simply amazing! Here are the 10 Copilot features you can't miss in 2024:👇 https://t.co/R63ZeBOvXk
— Robin Delta (@heyrobinai) Mar 3, 2024
New to DSPy 🚀 Adding a Synthesizer class to generate synthetic data based on examples you provide! Provide 3-5 examples from a dataset to the Synthesizer and it'll understand the task, generate data for input fields, and the output fields. Works great with gpt-4-preview! You… https://t.co/YG2gaAaAPn https://t.co/vxBcSN0ptD
— Herumb Shandilya (@krypticmouse) Mar 2, 2024
I am going even further. Now, with a bit of meta-programming, I can dynamically create signatures on the fly and generate synthetic datasets based on pydantic models. Crazy nice to see it in action 🔥 I propose to add the synthetic_data_generation function in the utils of DSPy.… https://t.co/ZxbQcE8FY9 https://t.co/FVJ9REzpiB https://t.co/46Bxc6pyDS
— Franck SN (@ndzfs) Mar 2, 2024
Looks like both @krypticmouse and @ndzfs have been independently developing synthetic dataset generators for DSPy optimizers. So today we got two PRs for this, with very different approaches! Community feedback welcome; maybe we can merge the two into one robust version? https://t.co/ws4oUykGod
— Omar Khattab (@lateinteraction) Mar 2, 2024